Privacy and Security Notice
Email exchange relevant to the 120Hz Task Force
Email exchange relevant to the 120Hz Task Force
(Some header information removed to keep this simple....)
On Wed, 29 Sep 1999, W. Des Ramsay wrote:
> Hi Brian,
>
> I was thinking about your comments on the data taking schemes. It's
> true that integrating for exactly 1/30 s should cancel 60 Hz and all
> its harmonics. If I understood Roger's comments at the meeting
> correctly, I think he was concerned that if for some reason we don't
> cancel the noise completely, (perhaps the gate is not exactly 1/30 s,
> or the line noise is changing during the gate) then we may not know
> it. We would just see a batch of strange data and not know the reason.
>
> The fact that we have a continual phase slip relative to the line
> provides a way out of this. Presumably, if some line noise leaks
> through into our data, the magnitude of the contamination will depend
> on the phase of the noise relative to the integration gate. Over the
> long term, the phase slip ensures that all phases are equally
> populated, so such noise will cancel out. If, in addition, we include
> in the data stream a scaler indicating the phase of each read relative
> to the 60 Hz line, then we can look for correlations between the data
> and the line phase. We do this in E497 and have not seen any such
> correlations.
>
> In E497 we integrate for exactly 1/60 s and the phase slip is quite
> slow. We can select any desired phase slip, but have been using 360
> degrees in 20 minutes. The phase slip for G-zero will probably have
> to be much faster, I guess 360 degrees in a few seconds or less -- is
> that right? Nevertheless, if the time between reads is chosen right,
> all phases will eventually get populated.
>
> I suppose the advantage of the 1/120 s scheme is that, in addition to
> cancelling line noise, we can measure how much we cancelled. At the
> very least it may be satisfying to know we weren't cancelling huge
> amounts of line noise.
>
> It may be an irrelevant point, but it also occurs to me that in the
> 1/120 s scheme, assuming that the integration periods are shorter that
> 1/120 s but are separated by exactly 1/120 second, only the
> fundamental and ODD harmonics of 60 Hz will cancel in a given 1/30 s
> spin state. EVEN harmonics have to wait (and be constant) for two spin
> states (1/15 s) to cancel.
>
> Des
>
Date: Wed, 29 Sep 1999 22:38:08 +0000 (GMT)
From: Doug Beck
To: "W. Des Ramsay"
Subject: Re: G-zero data aquisition schemes
Dear Des,
I agree with what you have written - I think the basic idea is the one of
satisfaction we are not having to cancel a large amount of 60 Hz noise.
This is why I suggested running only part of the time in the 120 Hz mode.
There is a cost if we use the Orsay electronics - as I understand it -
because of the data rate we wouldn't be able to get the full buddy
spectrum (just a number with which to compute deadtime like the other
electronics). I am not sure we will want to save all these buddy spectra
either, but they would be nice to have for some of the data-taking period.
I believe we can also consider flipping the spin at 120 Hz as well or at
60 Hz (or, if Brian has his way at 360 Hz). This, thinking quickly, I
belive reduces the problem you raise in the last paragraph.
I am confident that our task force will sift through the various
combinations to find the most useful ones for the experiment.
Cheers,
Doug
Douglas H. Beck
Professor, Department of Physics
University of Illinois at Urbana-Champaign
1110 West Green Street
Urbana, IL 61801-3003
(217) 244-7994
dhbeck@uiuc.edu
Date: Wed, 29 Sep 1999 22:31:04 -0500
To: "W. Des Ramsay"
From: "Roger D. Carlini"
Subject: Re: G-zero data aquisition schemes
Hello All:
Some comments: 30Hz is generally a quite frequency in the noise
spectrum of accelerators - Jlab included. If you hang a spectrum
analyzer on just about any signal you can observe this. Below about
20Hz you start seeing mechanical vibrations. A 30Hz reversal locked
to the line also cancels any 60Hz noise to the extent that your
system is linear over the magnitude and frequency of the noise.
You don't want to reverse at 60Hz for obvious reasons. However, the
noise spectrum has harmonics (sharp peaks) at 120, 240, 360Hz,... The
general guidelines I have followed for parity runs are that if you
want to reverse at a frequency higher than the line it should be very
high - say ~1Khz and not be locked to the line. This gets you well
above the line harmonics and again in a region where the overall
noise is small. Try looking at spectrum of this noise at Jlab or you
favorite accelerator. The down side of these very fast reversals is
that you might reduce the precision of each measurement to such an
extent that least count effects come into play and prevent you from
operating near counting statistics.
The basic reason I push for a reversal at 30Hz with a 4x oversample
is to monitor that any 60Hz noise does not exceed the linearity
limits of our detector + electronics. What I know about the HAPPEX
runs is that with the high polarization source they stay just under
the level where a false parity signal systematic correction is
required. We want to do an experiment ~10x more precise and do not
have their simple linear analog electronics.
Best Regards
Roger
Date: Thu, 30 Sep 1999 21:34:30 -0400
From: Brian Quinn (bquinn@cmu.edu)
To: pate@nmsu.EDU, pitt@vt.edu, CARLINI@JLAB.org, LRYLEE@CSV02.TRIUMF.CA,
VANOERS@CSV02.TRIUMF.CA, RAMS@reg.triumf.ca, beck@nialas.npl.uiuc.edu,
clark@ERNEST.PHYS.CMU.EDU, QUINN@ERNEST.PHYS.CMU.EDU
Subject: 60 Hz noise
Hi Guys,
I certainly agree with Des, that it would be a very good idea to keep track
of our phase relative to line for each 1/30 s macropulse. With a 100 us
dead-spot to allow for polarity reversal, the phase will slip at ~2 degrees
per macropulse or 64.6 degrees per second. So Des has it right, we'll sample
all starting phases every 5.6 seconds.
I'm afraid Doug misunderstood me or just mis-typed in saying that I would
advocate flipping the spin at 360 Hz. I would certainly stick with 1/30 s
beam macropulses (with an ADDED 100 us to allow for possible polarity
reversal). In fact I'm not an advocate of readout at 120 Hz, 360 Hz, or
720 Hz. I just pointed out that the argument for 120 Hz running (ie READOUT,
not spin flipping) quickly becomes an argument for 720 Hz running once you
realize that 360 Hz is usually a large noise component, and try to oversample
that as well.
I think we all agree that maximal noise cancellation is achieved if we
accumulate for exactly 1/30 s. As Des points out, if we break up that
interval by stopping data taking briefly every 1/120 s, then we undermine the
cancellation of the even harmonics of 60 Hz. Fortunately, if we must readout
at 120 Hz, we should be able to keep these added interruptions under 1 us
since we will NOT be flipping the spin (nor moving the Orsay data to the back
DSP) during the added 3 extra breaks between the 100 us spin-flipping break.
Now the big question is: do we need to do any 120 Hz running to measure our
sensitivity to 60 Hz noise, or is the phase-slipping sufficient?
In fact, I'm not sure what signal we would look for relative to the phase.
I suppose we would be looking for a wider-than-root-N distribution to the
number of counts measured for 1/30 s starting at some phase of 60 Hz which
isn't seen for intervals which start at other phases of 60 Hz. I'm not sure
I can imagine anything that would cause such an effect, though. Perhaps, as
Des suggests, if our '1/30 s' is not EXACTLY two cycles of the power company's
phase, then it is possible. Alternately you need to imagine that the
electronics has some sort of hysteresis which allows it to 'remember' the
pattern of what the line voltage has done since last time the electronics was
read out.
But I'm also not sure what we do with anything we measure at 120 Hz readout
frequency. Suppose, we see some big effect. Suppose (for some starting phase
of the 1/30 s macropulse w.r.t. 60 Hz ) we see that the first and third 1/120
s intervals have 10 % more data than the second and fourth intervals. What
does that tell us about the 'linearity limits of our detector + electronics'?
And how, in heaven's name, does it cause a 'false parity signal'???? I
suppose, if the charge monitors indicated that the beam was constant during the
intervals, we could say that our detection threshold was effectively bouncing
around at 60 Hz. That wouldn't be a good thing. But would it be a bad thing?
Can somebody explain to me why such a variation in effective threshold would
in itself either induce a false parity signal (when we're RANDOMLY changing
helicity) or how it would even degrade the error on the extracted asymmetry?
I can see how it could modify other effects, such as helicity-correlated beam
intensity variation. But I don't see how it would cause effects beyond what
we would have predicted if we had been blind to the 60 Hz variations.
(Presumably we will MEASURE the dependence of count rate on beam parameters in
the presence of the same 60 Hz effects that will exist during data taking.)
On the flip side... suppose we see NO difference between even and odd
sub-intervals of the 1/30 s macropulse. What does that tell us? We can't set
any meaningful upper limits on things such as effective threshold variation.
We can just say that the happen at frequencies which are harmoics of 120 Hz.
Best regards,
Brian
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Fri, 1 Oct 1999 12:08:08 -0500
To: Brian Quinn (bquinn@cmu.edu)
From: "Roger D. Carlini"
Subject: Re: 60 Hz noise
Hello:
An extreme example where you could see the effect of a non-linear
detector/readout. The helicity flips at 30Hz, all the + pulses are
absolutely constant in current, all the (-) pulses have high
frequency noise with peak-to-peak variations100x the average current.
Assume the true integrals of these two pulses are exactly the same.
However, if you try to count events the high instantaneous rates will
cause you to miss some events in the (-) pulse if the bandwidth of
your detector + electronics is not high enough. So the difference
over the sum (ie. parity signal) will be non-zero. This happens in
both analog and digital systems. This sort of effect generally does
not trouble anyone doing a ~.1% experiment, but if we try to hold the
system to 10-7 or 10-8 these sorts of effects can bite us.
Now as for a "random reveral" at 30Hz curing all ills - on should
keep in mind that it is not a random reversal that is flat in time.
We would be reversing only at multiples of 30Hz and not at any time
between 0 and infinity. So a fourier analysis of this so called
random reversal will still show a component at 30Hz. This sort of
reversal arrangement works well for HAPPEX primarily (I believe)
because the Jlab beam is so darn good anyway. If we had a real noisy
beam it might not provide enough suppression.
Finally, reversing at 30Hz and oversampling at 120Hz is on solid
grounds with respect to being able to detect small components at 60Hz
and 120 HZ. It should prove a good (maybe critical) tool when we try
to under stand the data at the <10-7 asymmetry level. Also, if the
readout is at 120Hz there will be no 30Hz component to leak back into
the electrronics.
Best Regards
Roger
Date: Fri, 1 Oct 1999 17:21:47 -0400
From: Brian Quinn (bquinn@cmu.edu)
To: pate@NMSU.EDU, pitt@vt.edu, CARLINI@JLAB.ORG, LRYLEE@CSV02.TRIUMF.CA,
VANOERS@CSV02.TRIUMF.CA, RAMS@reg.triumf.ca, beck@nialas.npl.uiuc.edu,
clark@ERNEST.PHYS.CMU.EDU, QUINN@ERNEST.PHYS.CMU.EDU
CC: QUINN@UQUARK.PHYS.CMU.EDU
Subject: Aren't we talking about line-induced effects??
Hi Rodger,
...and all the other poor souls trapped on this mail list... let me know
if you want to drop out of this dialogue,
It seems like the discussion is widening to lots of non-60 Hz related
questions. I certainly don't think 1/30 s readout and random-reversal
protects against all possible false asymmeteries, and I didn't mean to imply
that it does. But we're talking specifically about 120 Hz readout here, which
is targeted at line-induced effects.
Your first point is about **helicity-correlated** beam intensity structure.
I agree they would be bad things to have.. whether they're at 60 Hz, 120 Hz,
1kHz, 33 MHz or non-periodic. But that seems like something which is
addressed much more directly by the buddy scalers (that's exactly what they
are for) than by 120 Hz readout. You MIGHT see such an effect by 120 Hz
readout if it happened to be at 60 Hz or 120 Hz. But you would see a
different correlation indicated in the buddies for the different helicities
regardless of the source or frequency of the problem. So I don't see this as
an argument for 120 Hz running, nor as an answer to my (serious) question of
how a periodic line-induced variation in performance of the
detectors/electronics/beam can cause a false asymmetery. Sure it can if it's
helicity correlated. Anything (room temperature, line voltage, seismic
activity) can cause a false asymmetry if it's helicity correlated. And we
sure have to fight to control and monitor helicity-correlated variations in
parameters. But I don't see a connection between that and 120 Hz running.
On the 2nd point. Again, I don't think 30 Hz random reversal cures all
ills. I just think it makes us immune to false asymmetries from 60 Hz
flucutations and reduces overall sensitivity to 60 Hz noise. It's a small
point, but I sure don't think the fourier analyis of beam helicity will have
a 30 Hz component. It's the integral from 0 to large of sin(omega t) *
a random number. If that's not zero, is it positive or negative??
I agree you'll be able to detect small 60 Hz components. I still don't know
what we do with that info. For example.. how big a 60 Hz variation in
count rate is acceptable?? 1%? 10**-6? How do you decide?
I didn't understand the last point about a 30 Hz component leaking into
the electronics. What effect does it have that the electronics is being
gated on and off at 30 Hz? How does it hurt us?
Have a good weekend,
Brian
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Sun, 3 Oct 1999 13:25:51 -0500
To: Brian Quinn (bquinn@cmu.edu)
From: "Roger D. Carlini"
Subject: Re: Aren't we talking about line-induced effects??
Hello:
I will try to address some of you statements. Yes we are talking
about line-induced effects. These can be carried both by the beam
itself (bad power supply wiggles the beam at 60, 120, 240, 360,...
Hz.) or coupled into the electronics directly. In either case these
are the bad components which consume the linear bandwith of your
system. So it is a good ideal to monitor them on all recordable
signals. The key word here is all.
Your wrong on the issue of what a fourier analysis of a so called
"random reveral" at 30Hz which produced by a roll of the dice each
1/30th of a second. This signal will indeed have a large component at
30Hz. In fact what you will see on a spectrum analyzer is a large
peak at 30Hz, one at 15Hz, 7.5Hz...., but Nothing faster that 30Hz of
course. What your effectively generating is some reversal at 30Hz,
some at 15Hz, some at 7.5Hz, etc. by this method of reversing the
helicity. Try it for yourself.
If you did not randomize the 30Hz reversal any 30Hz used as a gate in
the readout electronics would be 100% correlated with the helicity.
This minor voltage increase let loose in the racks will slightly
shift discriminator thresholds and also mess up precision analog beam
current measurments. Again, this is a very real effect and an
experiment killer when trying to do experiments at the <10-7 level.
Ask Vanoers!
Regards
Roger
> Date: Fri, 1 Oct 1999 14:52:46 +0200 (MET DST)
> From: ARVIEUX Jacques
> To: Doug Beck
> Subject: summary report of vancouver meeting?
>
> Hi Doug
>
> Do you have the intention to write a summary of the Vancouver meeting. We
> would appreciate some official statement about the specss for the
> 120 Hz data read-out (we have a simple solution for which there would be
> no limitation on the Orsay electronics: we will implement 4 sets of
> spectra within one 30 Hz period, and then transfer them at all together at
> the end during the 200 microsec devoted to helicity flipping).
>
> We would also appreciate some statements about the limitation to 4 MHz of
> the 4-fold counting rate. I'll have Monday the latest estimate by Emmanuel
> for the rates due to the target (elastic and background). We will write a
> report soon after. By the way, if the problem is with the neutrals we are
> all in trouble with the single counter rate in any (fw or bw) set of
> counters.
>
> If you don't find the time to write a full summary report, could you still
> please take a few minutes to answer the above questions.
>
> Cheers
>
> Jacques
>
>
> ====================================================================
> Dr Jacques Arvieux
> Institut de Physique Nucleaire d'Orsay (IPN-Orsay)
> Universite Paris-Sud
> Bat. 101B, piece 027
> F-91406 ORSAY-Cedex, FRANCE
> Tel: 33-1-6915-6727
> FAX: -6470
> email: arvieux@ipno.in2p3.fr
> ====================================================================
Date: Mon, 4 Oct 1999 20:28:57 +0000 (GMT)
From: Doug Beck
To: ARVIEUX Jacques
cc: dhbeck@uiuc.edu, pate@nmsu.edu, Phil Roos ,
Serge Kox , Allison Lung ,
willy@uinpla.npl.uiuc.edu
Subject: Re: summary report of vancouver meeting? (fwd)
Dear Jacques,
I didn't plan to write a summary of the Vancouver meeting.
On the 120 Hz readout - what I heard was agreement that it was a good idea
- the exact configuration (flipping spin at 120 Hz - this is a distinct
possibility, measuring for exactly 1/120 s or (1/120 - delta) s, or what
fraction of the time should be allocated to this mode) will be determined
after a recommendation by the task force chaired by Steve Pate. You
should let him and me know if there is particular further information you
need for the Orsay electronics development.
On the 4 MHz - what I heard at the detector meeting was that 4 MHz of
coincidences between front and back detectors would not be acceptable -
that we would have to run with a lower rate than this.
If anyone remembers anything differently, please let me know. Thanks.
Cheers,
Doug
Douglas H. Beck
Professor, Department of Physics
University of Illinois at Urbana-Champaign
1110 West Green Street
Urbana, IL 61801-3003
(217) 244-7994
dhbeck@uiuc.edu
Date: Tue, 5 Oct 1999 23:50:51 -0400
From: Brian Quinn
To: pate@NMSU.EDU, lung@jlab.org, saw@jlab.org, WILLY@UINPLA.NPL.UIUC.EDU,
pitt@vt.edu, CARLINI@jlab.org, RAMS@reg.triumf.ca,
beck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr, Kox@in2p3.fr
Cc: QUINN@UQUARK.PHYS.CMU.EDU
Subject: What we decided in Vancouver
Dear 120 Hz task force,
As requested, here's my summary of what I think we decided/learned from
(at least my part of) our discussion of 60Hz noise and 120 Hz running.
I) We will run with one beam helicity for 1/30 s (a macropulse) there will
then be a dead period (ADDED on to this 1/30 s) of approximately 100 us
duration to allow the spin to flip (of course it will only actually flip about
half the time during this dead period, but we'll mask off for 100 us
regardless). Thus we will have beam of stable helicity for exactly two cycles
of the line power so any 60 Hz-related variation in beam parameters or
detector/electronics performance will have exactly the same chance to 'average
out' for one macropulse as for the next. (Fine print: it's not really
'exactly the same' because different macro-pulses will start at different
phases of the 60 Hz. So, if something has a long memory then it could respond
differently for different phases eg. high-rate;low-rate;high-rate;low-rate
running might not give identical response to what would be seen for
low-rate;high-rate;low-rate;high-rate running.)
The hope is that this will reduce non-statistical errors within the
macropulses which might exist if some macropulses contained the 'good' part of
a 60 Hz cycle while others contained the 'bad' part.
We didn't address the question of how we define 1/30 s. We could define it
in an absolute sense (eg by a quartz oscillator) or in a nominal sense
(defining our macropulse to be two cycles of the line power at the present
average line frequency). In principle line frequency can vary by a Hz or two.
It usually varies by less and the power company tries to keep the average
period at 1/60 s.
II) When running at 120 Hz, we won't pause for 25 us every 1/120 s to allow
the Orsay data to be transferred to the back DSP. Rather, if we use the Orsay
electronics, they will accumulate four separate spectra and transfer all of
them during the 100 us break, discussed above, which occurs after each 1/30 s
of stable beam (which I would still like to call a macropulse).
We WILL need to insert three breaks into the 1/30 sec to allow the scalers
to be latched and cleared for the LTD's (NA and/or Grenoble). These breaks
can be very short, probably under 100 ns, but that should be confirmed for all
THREE types of VME scaler (Grenoble stand-alone, Grenoble on LTD, and Struck
7200). The actual data-taking periods won't be exactly 1/120 s in the '120
Hz' running mode, but roughly (1/30 s - 3*100 ns)/4 ... which is pretty
darned close to 1/120 s.
The actual readout of the data will be done four times per macropulse for
'120 Hz' running of the LTD's and once per macropulse (reading out four sets
of spectra) for the Orsay electronics if they are used.
Best regards,
Brian
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Wed, 6 Oct 1999 15:25:25 -0400
From: Brian Quinn (bquinn@cmu.edu)
To: pate@nmsu.EDU, lung@JLAB.ORG, saw@JLAB.ORG, WILLY@UINPLA.NPL.UIUC.EDU,
pitt@vt.edu, CARLINI@JLAB.ORG, RAMS@reg.triumf.CA,
beck@nialas.NPL.UIUC.EDU, arvieux@ipno.in2p3.fr, Kox@in2p3.fr
Subject: continuing discussion of advantages of 120 Hz running
Hi Roger (and 120 Hz task force),
Perhaps we should just agree to disagree on the question of the Fourier
analysis. We must be talking about different things. I agree there may be
large Fourier components at 30 Hz (and higher!! and at 0 Hz) in the beam
current, if the beam is off or fluctuating for any part of the 100 uS for
reversal. Perhaps that's what you're referring to.
I was thinking of a fourier transform over only the time for which the
experiment is gated on, which would 'censor' any intensity fluctuations during
the beam flip. That F.T. should be a huge spike at zero Hz plus some tiny
spikes at very high frequencies due to heliciy-correlated beam intensity
fluctuation (which aren't periodic, and so should give no strength at
frequencies which are sampled over many periods in the sample being
transformed. What I was picturing was the product of beam intensity times
helicity (over times the experiment is gated on). That should have a very
small 0 Hz component (being positive almost exactly as much as it is
negative). If we sample over many * 1/30 s it should also have a tiny 30 Hz
strength which reflects the helicity-correlated beam intensity fluctuation. If
that effect were absent then the 30 Hz component would be zero (for a long
data sample) because a) random numbers aren't periodic or
b) the amplitude is proportional to integral ( sin(2 pi f t) * r(t) dt) where
r(wt) is +1 or -1 chosen randomly every 1/30 s. For f = 30 Hz, the sin
integrates to zero for each individual r. That's not true if f<30 Hz, but
each 1/30 s interval of sin(2 pi f t) will be weighted by the average of
many choices of random numbers, and surely that should average to zero for
long sampling times. (None of this applies to a small data sample, like one
second, because the numbers just don't average to zero very accurately.)
I suspect we've been agreeing about this, but in such a way that we don't
understand each other.
But I'm not sure the Fourier analysis of the time structure of the beam
pulses is the key point. More to the point, what do we do with the 120 Hz
information? To take a concrete example, suppose that it shows a 1% variation
in counting rate (not related to any change in beam intensity or helicity) for
one half of the 60 Hz cycle compared to the other (for some measured starting
phase). What do we do with that information? Is 1% easily acceptable?
borderline? wildly unacceptable? How do we answer that?
I would guess we average them together in sets of 4 to see whether or not it
averages accurately to zero over a 1/30 s macropulse. If that averages to
zero at the 10**-7 level(i.e. we can't see any width to the distribution beyond
statistics), we're fine. Otherwise we DO have a problem. But
that is exactly the information measured in the 30 Hz running. What
quantitative information is added by knowing the size of the 60 Hz variations
that are cancelling? (We still won't know the size of the 360 Hz variations
that are cancelling. They may be larger. Is that a problem? )
Is there some other way to make use of the 120 Hz measurements that I'm not
seeing?
Best regards,
Brian
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Wed, 6 Oct 1999 11:11:27 -0700
From: "W. Des Ramsay"
To: quinn@uquark.phys.cmu.edu, pate@NMSU.EDU, lung@jlab.org, saw@jlab.org,
willy@uinpla.npl.uiuc.edu, pitt@vt.edu, carlini@jlab.org,
dhbeck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr, kox@in2p3.fr
Subject: how to define 1/30 s.
Dear Friends of Precision Measurement,
It's probably better to define the 1/30 second in an absolute way, for
example 333,333 ticks of a 10 MHz clock. The gate then opens exactly
on a clock tick and closes exactly on a clock tick, so its length is
known much, much more precisely than one clock tick. This is important
because the number we are trying to measure is the difference between
the data in the + and - gates.
If we try to vary the gate to exactly match the line, then we will
have to normalize to something like a fast clock, but because the gate
and the clock are asynchronous, our knowledge of the gate length will
be uncertain to one clock tick. This shows up as digitization noise.
Because in G-zero we count scattered particles, this digitization or
round-off noise will not be very important compared to the statistics
of a 1/30 second read, but, all else being equal, I would vote for the
fixed gate length.
Des
Date: Tue, 12 Oct 1999 9:35:56 -0700
From: W. Des Ramsay
To: quinn@uquark.phys.cmu.edu, pate@NMSU.EDU, lung@jlab.org, saw@jlab.org,
willy@uinpla.npl.uiuc.edu, pitt@vt.edu, carlini@jlab.org,
dhbeck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr, kox@in2p3.fr
Cc: RAMS@CSV02.TRIUMF.CA
Subject: what to do with the 1/120 s data
Here are some comments on "what we do with the 120Hz measures".
1) If we label the four intervals in a 1/30 second state as A,B,C,D,
then we form the quantity (A-B+C-D) and plot it against the 60Hz phase
at the start of A.
2) If there is a 60 Hz component, then (A-B+C-D) vs phase will show a
cosine-like shape with maximum +ve value at 0 degrees and maximum negative
value at 180 degrees. The amplitude of this component is the integral
of the absolute value of the 60 Hz component.
3) 120 Hz and other even harmonics of 60 Hz will NOT show up.
4) Odd harmonics will show up at a correspondinly shorter "wavelength"
on the plot. Their amplitudes will be 1/n of the integral of the absolute
value of that component.
5) This result will be interesting, and, if the components are
horrendous we might have to fix something, but, like Brian, I don't
know how bad it has to be before we worry. The bottom line is whether
anything can be seen in the plot of the exact 1/30 s integrals vs
phase. This is what has leaked through. I guess it would take a fairly
long run to see small contamination.
6) If there is some residual line noise, I can't see how it can do
anything but make us have to run longer. Because of the phase slip,
it's not helicity correlated and can't bias our result.
Des
Date: Tue, 12 Oct 1999 13:27:11 -0500
From: Roger D. Carlini
To: W. Des Ramsay
Cc: quinn@uquark.phys.cmu.edu, pate@nmsu.edu, lung@jlab.org, saw@jlab.org,
willy@uinpla.npl.uiuc.edu, pitt@vt.edu, carlini@jlab.org,
dhbeck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr, kox@in2p3.fr
Subject: Re: what to do with the 1/120 s data
Comments
If the bandpass of the system at the reversal frequency (say 30Hz) is
sufficiently narrow and there are no components generated from the
beating of lower frequencies (mechanical vibrations, etc.) with the
line noise (which typically manifests itself as beam position motion)
that produce harmonics that can get into the bandpass window of the
reversal, and if all beam carried frequencies stay with the linear
region of our detector+readout system, then I agree that all that can
happen is we may find ourselves running at say ~2x counting
statistics rather than ~1.2x counting statistics.
However, if 60Hz does appears on the beam in the form of position
motion how does our system respond? What is the effective linear
range of our detectors+readout? Say the beam motion is +-0.5mm at
60Hz. Is the effective linearity 10-2, 10-3, 10-4 ? How much 60Hz can
we tolerate before we no longer can operate near counting statistics.
Doubling our ratio to counting statistics with increase the running
time by a factor of 4. We should be able to do studies to
characterize the CMU electronics.
Date: Tue, 12 Oct 1999 15:55:21 -0400 (EDT)
From: Mark Pitt
To: quinn@uquark.phys.cmu.edu, pate@nmsu.edu, lung@jlab.org, saw@jlab.org,
WILLY@UINPLA.NPL.UIUC.EDU, pitt@vt.edu, CARLINI@jlab.org,
RAMS@reg.triumf.ca, beck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr,
Kox@in2p3.fr, vanoers@erich.triumf.ca
Cc: pitt@vt.edu
Subject: Pockels cell flip time
Dear 120 Hz task force,
I spoke to Matt Poelker last week to get the correct current numbers for
the Pockels cell flip time. He says that currently the time required for
the Pockels cell to stabilize after flipping is about 200 microseconds.
He says that the experimental halls currently choose to wait a full
millisecond between flips. He thinks there is room for improvement
in the 200 microseconds, so it will likely be smaller by the time G0 runs.
But, for the moment, we probably shouldn't assume "delta" will be any
smaller than about 200 microseconds.
Also, I had a couple of comments relating to the questions Steve
Pate raised in a previous email:
a) What problems are solved by collecting data at 120 Hz while still
flipping spin at 30 Hz?
I don't think any specific problems are solved by doing this; it is only a
warning flag of something to think about/investigate further. For
example, consider the case where an analysis of the kind Des just
circulated showed a significant noise component in the beam current at
some odd harmonic of 60 Hz. Then one could do a deliberate test by
modulating (at a much slower time scale than 60 Hz) the beam current at
the polarized source in order to measure the response of the overall
system to a variation in the beam current of the observed amplitude. A
similar study can be done for beam position, if that had a large noise
component.
b) What are the specific mechanisms for a false parity signal which could
be corrected by over-sampling?
Once again, I think the over-sampling is merely a warning flag; that data
cannot be used in an explicit sense to do corrections. As a
concrete example, take the beam position data. Here, it is very clear what
to do to make a "false parity signal correction": i) continously monitor
the helicity-correlated beam position differences, ii) occasionally
modulate the beam position deliberately to determine the response of the
system to beam position motion, iii) combine the data from i) and ii) to
make the correction. I don't see a similar straightforward algorithm for
using the 120 Hz over-sampling data for "corrections". So since there is
no way to make continuous use of the 120 Hz over-sampling data, I think it
only makes sense to do it some small fraction of the time (but I think(?)
we all already agree on that).
Finally, I wanted to raise one other point that there has not been much
discussion of. If we do run at 120 Hz for some fraction of the time, is
there any reason NOT to flip the spin at 120 Hz during that period as
well? In Des' notation, this would mean we could separately construct an
asymmetry corresponding to each of the periods A,B,C and D. In principle,
these asymmetries should all be the same within statistics. But if they
were not all the same, then it would be the signal of something bad (like
the phase slip averaging mentioned by Des not completely averaging out the
helicity correlation). If the asymmetries were not the same, then you
could not necessarily make any "correction" based on this information, but
you would have a warning flag of something to investigate further.
As a concrete example of how a line-related helicity correlation
could happen, I mention the following: The helicity-defining Pockels cell
at the polarized source is fed by a switch that switches back and forth
between a positive and negative high voltage supply. Imagine that the
positive high voltage supply has some 60 Hz noise that isn't present on
the negative supply. That would mean that the mean high voltage during
phase A could be different than that during phase B (although this effect
would hopefully be minimized by the phase slip in our data-taking sheme).
The circular polarization distribution and the laser beam direction both
vary with the high voltage. When one folds this together with the
interactions at the surface of the strained GaAs crystal, there are
mechanisms possible for helicity-correlations in the electron beam
intensity, position, and shape. For example, if the normalized count
rate in our detector varied with beam shape, then a helicity correlation
in beam shape (something we don't explicity measure otherwise) such as
that described here would show up as a different detector count rate
asymmetry in period A versus period B.
Although the above is a somewhat contrived example, it does show a case
where we would gain information we would not otherwise have by flipping
the spin at 120 Hz when we do the 120 Hz data-taking. I can't think of
anything we would lose by not flipping at 30 Hz during the 120 Hz
data-taking.
Regards,
Mark
Date: Tue, 12 Oct 1999 16:27:03 -0400
From: Brian Quinn
To: pate@nmsu.EDU, lung@JLAB.ORG, saw@JLAB.ORG, WILLY@UINPLA.NPL.UIUC.EDU,
pitt@vt.edu, CARLINI@JLAB.ORG, RAMS@reg.triumf.ca,
beck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr, Kox@in2p3.fr
Cc: QUINN@UQUARK.PHYS.CMU.EDU
Subject: 120 Hz flipping
Dear 120 Hz task force,
In response to Mark's suggestion...
I guess the reason for not flipping the spin at 120 Hz is the 200 us (or
1ms?) dead time that would be introduced between the '1/120 s'
mini-macropulses. The result would then be that four mini-macro pulses would
not add up to one of our normal 1/30 s macropulses. Presently the problem is
limited to ~100 ns of padding required between the 1/120 s data-taking
periods. The apparent effect of odd harmonics of 60 Hz might then be
artifically enhanced because we would not be integrating over an integral
number of periods.
I'm not sure you gain much from the faster flipping. With 1/30 s
macropulses you can still assemble separate asymmetries for periods A,B,C, and
D. The only difference is that you have to combine data taken over 4*1/30 s
rather than combining data taken over 4*1/120 s. In either case there is some
phase slip between the first and the fourth of the measurements being
combined. But that phase slip is just as large for combing four 1/120 s
mini-macropulses (with 200 us to 1ms between them) as it is for combining four
1/30 s ordinary macropulses (with the same dead time between them).
Best Regards,
Brian
Date: Wed, 20 Oct 1999 10:53:38 -0600 (MDT)
From: pate@lepton.nmsu.edu
Subject: summary of "120 Hz" email discussion
Dear Task Force Members
Here is what I have understood from the email discussion.
(1) Helicity-correlated noise can create a false parity signal. (I thought
I would start with something obvious.)
This is something which seems to be generally understood. We all
agree that we need to monitor and control any helicity-correlations. The
experiment has been carefully planned so that the only system which knows
the helicity (at the moment of the event) is the beam, and it is planned
to monitor some of the beam properties that might change with helicity.
We also plan to measure the response of our system to intentionally
produced changes in beam properties, in the hopes of making a correction
to our parity signal if necessary.
(2) If the average line-voltage frequency is 60 Hz, and if we always
count for exactly 1/30 sec, and if we flip spin randomly, and if we allow
the initial phase of each 1/30 sec data period to slip so that all initial
phases are equally sampled (equal integrals of current*helicity), then all
60 Hz noise and its harmonics will be exactly cancelled.
This is true even if our system responds in a non-linear way to
the 60 Hz noise, as long as we equally sample all phases, and as long as
we do so more quickly than the nature of the 60 Hz noise is changing.
[I recognize this to be a strong statement. I invite clarifications from
others.]
(3) Acquiring data at 120 Hz while we continue to flip spin at 30 Hz can
tell us how much 60 Hz has leaked through, using the ABCD scheme outlined
by Des. It seems that this can be used as a relative diagnostic but not
for any absolute correction.
(4) It is possible for a helicity-correlated change in beam properties to
be caused by line noise, as cleverly outlined by Mark. I think this kind
of problem may need some more discussion.
(5) I didn't understand some of the discussion of "Fourier" components,
despite the fact that I am familiar with Fourier transforms and with the
operation of spectrum analysers (from my previous life at Indiana). But I
agree with Brian that the only thing which matters is the Fourier spectrum
while the DAQ is gated on.
I'll write an agenda for the telephone conference before the week is over.
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Fri, 22 Oct 1999 17:04:25 -0700
From: W. Des Ramsay
To: quinn@uquark.phys.cmu.edu, pate@NMSU.EDU, lung@jlab.org, saw@jlab.org,
willy@uinpla.npl.uiuc.edu, pitt@vt.edu, carlini@jlab.org,
dhbeck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr, furget@in2p3.fr,
kox@in2p3.fr, vanoers@physics.umanitoba.ca
Cc: RAMS@CSV02.TRIUMF.CA
Subject: Fourier Components
On http://www.triumf.ca/E497/g0/ I put frequency spectra of regular
and random spin-flip sequences.
Des
Date: Tue, 26 Oct 1999 13:32:22 -0600 (MDT)
From: pate@lepton.nmsu.edu
Subject: DRAFT recommendations of 120 Hz Task Force
Dear Task Force Members
Yesterday we had our telephone conference. The participants
included:
Mark Des Brian Allison Wim Roger Steve Williamson Steve Pate
I am in the process of writing the report, but I thought I would draft the
recommendations of the task force (as I understand them) in an email to be
sure we agree to them. I added a couple of generic recommendations (1 and
2) to remind our collaboration audience of two important issues.
Since no one from Orsay called into the telephone conference, I need some
feedback from Orsay in order to complete the report -- see recommendation
number 8.
Recommendations:
(1) Everyone working on the experiment must work hard to be sure that
only the beam knows the helicity during the time period immediately
surrounding each 1/30 s data-collection period. Information about the
helicity must only be added to our datastream after the collection process
is complete.
(2) Everyone must also be careful to attend to proper grounding and
isolation, to keep ground currents and noise to a minimum.
(3) The normal 1/30 s data-collection period will be defined by a
standard clock, not by the zero-crossing of the line voltage. The length
of the pause between data-collection periods, during which the polarized
source may change helicity, will be determined by the performance of the
source. Currently, this pause will be approximately 200 microseconds.
The determination of the exact length of the pause must also include a
consideration of the phase-slip between data-collection periods.
(4) The line phase at the beginning of each 1/30 s data-collection period
will be recorded, so that we can be sure that all phases are equally
sampled.
(5) The instantaneous line frequency will also be recorded, so that we
can monitor the size of deviations from 60 Hz.
(6) The method for collecting data at 30 Hz and 120 Hz will be that
described by Brian in his email, which is the same method described at the
Vancouver meeting. [In the report, I will provide all the details again.]
The only change is that the 100 microsecond pause is now a 200 microsecond
pause.
(7) We will normally collect data at 30 Hz. We will collect data at 120
Hz only a few times per day, for about 100 s each time, to check on the
amplitude of the 60 Hz noise component.
---
(8) Due to the 4 Mbyte/s data-transfer limit, the Orsay group should
consider ways to reduce the size of their datastream during 120 Hz
operation. Various methods have been discussed:
(a) 30 Hz: fine time bins for both normal and buddy spectra
120 Hz: fine time bins for normal spectrum, but buddy
spectrum is reduced to a single number
(b) 30 Hz: fine time bins for both normal and buddy spectra
120 Hz: coarser time bins for both normal and buddy spectra
(c) 30 Hz: fine time bins for both normal and buddy spectra
120 Hz: non-uniform time bins for both normal and buddy
spectra, using fine bins at peak locations and
coarse bins away from peaks
(d) the SAME non-uniform time bins for BOTH kinds of spectra,
for BOTH 30 Hz and 120 Hz operation; this makes the
analysis simpler, as the spectra to be analysed are the
same for both operations
The sense of the telephone participants was that the buddy spectra are
important, as they contain information about the background not found in
the normal spectra, therefore option (a) is least favored. The Orsay
group should comment on the other three options, or variations of them.
Also, since we want to be able to switch between 30 Hz and 120 Hz
operation quickly, the Orsay group should consider if the programs for 30
Hz and 120 Hz could be combined into a _single_ program. The option to
run at 30 Hz or 120 Hz would be determined by a single command or a status
register bit.
---
I look forward to everyone's constructive comments.
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Wed, 27 Oct 1999 12:43:40 -0400
From: Stephen A. Wood
Subject: Re: summary of "120 Hz" email discussion
I have not been paying too close attention to these discussions, but it
seems to me that we are assuming/proposing that the 30 hz reversal rate not
be phase locked to the AC power 60 hz.
I just talked to Bob Michaels in Hall A. Apparantly, for HAPPEX, the 30 hz
reversal IS phase locked to the 60 hz. The period of time over which they
acquire/integrate during a given helicity is fixed, so variations in line
frequency are taken up by letting the blanking period vary slightly.
Steve
--
---------------------------------------------------------
Stephen A. Wood Jefferson Laboratory (formerly CEBAF)
Internet: saw@jlab.org MS 12H
12000 Jefferson Avenue
Phone: (757)269-7367 Newport News, VA 23606
FAX: (757)269-5235
Pager: (757)849-7367 Office: CEBAF Center C121
---------------------------------------------------------
Date: Wed, 27 Oct 1999 11:22:03 -0600 (MDT)
From: pate@lepton.nmsu.edu
Subject: Re: summary of "120 Hz" email discussion
Dear Steve
> I have not been paying too close attention to these discussions, but it
> seems to me that we are assuming/proposing that the 30 hz reversal rate not
> be phase locked to the AC power 60 hz.
That is correct.
>
> I just talked to Bob Michaels in Hall A. Apparantly, for HAPPEX, the 30 hz
> reversal IS phase locked to the 60 hz. The period of time over which they
> acquire/integrate during a given helicity is fixed, so variations in line
> frequency are taken up by letting the blanking period vary slightly.
Mark Pitt discussed with the source group how the reversal is done, and
they said we could use a clock instead of the line-voltage-zero-crossing
if we wanted. I don't recall anyone mentioning the use of a variable
blanking period length by Hall A. We wanted to have an exact amount of
phase slip during each blanking period (so that we would understand the
precession through these starting phases), so we agreed to the use of a
reference clock instead of the line-frequency. But it sounds like Hall A
achieves the same goal in a different way.
I am not really sure which method is better. If the deviations from 60 Hz
frequency are small, then it doesn't matter.
Does anyone know if Hall A has been monitoring the deviations from 60 Hz
and also measuring the initial phase of each data-collection period, as we
propose?
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Thu, 28 Oct 1999 12:08:17 -0400
From: Brian Quinn
To: PATE@lepton.nmsu.edu
Cc: QUINN@UQUARK.PHYS.CMU.EDU
Subject: stop gap answer....
Hi Steve,
I'm still meaning to update the info on the web about the buddy system to
answer your questions. Meanwhile, Val Zeps just asked for an introduction to
what it's all about so I fired off a quick reply. I've included that
buddy primer below to get you started until I have something better written
up.
Take care,
Brian
Yeah, I'm trying to update some info about them on the web. But the
minimal idea is that they monitor the deadtime to account for the possibility
that the beam intensity is non-uniform. The big problem would be if the
beam intensity had a different structure for one helicity than for the other,
but had the same average intensity. Then a naive dead time correction
based on average intensity would under-correct the helicity with the
bigger intensity fluctuations, introducing a false asymmetry.
The idea is that we would like to monitor the usual dead-time numbers
(how often did an event happen when we weren't ready to accept it) but the
deadtime is intrinsic in the discriminators and meantimers so we can't
see the events we miss (well, duh!). So instead we measure how often an
equivalent detector (the buddy) saw a hit when the LTD channel of interest
was busy and wouldn't have been able to record it. If rates were identical
it would be semi-straightforward to unfold pile-up correction. In fact the
buddy and detector of interest will probably have different average
rates because of different thresholds. In principle it should be possible to
unfold the corrections (with some model such as gaussian for intensity
fluctuations) numerically. But the more important thing is to see that
the buddy scalers see the same pile-up for both helicities. Then we may never
calculate the deadtime correction, but we will at least know it's the
same for both helicities.
An extra wrinkle is that the deadtime loss effects would change
character in the (unlikely?) event of an even/odd pulse-to-pulse structure
in the intensity. (ie alternating intensity every 32ns) Because the
next pulse is locked out by the deadtime logic (to give the meantimer/
discriminator time to recover) the deadtime losses would actually be
LESS in the presence of that beam structure. So we monitor same pulse
pile up separately from next pulse pile up.
Hope that helps!! ... I'll let you know when the longer writeup is
updated.
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Fri, 29 Oct 1999 17:14:33 +0200 (MET DST)
From: GRAVE Xavier
Subject: Orsay DAQ & 120 Hz
Dear Betsy and members of the electronics and 120Hz clubs,
Sorry but we could not get the connection to the teleconference (we have
found a solution but too late, we'll do better next time).
Hardware DAQ requirements for Orsay Electronics:
===============================================
DAQ material :
For the Orsay DAQ test one needs the following equipment :
-a vxi crate (Betsy had proposed one)
-a vxi slot 0 : MVME2301 + STR8032/ADA (from bastian technologie)
perhaps we can buy a spare one and send it to JLab for the
tests
-a unix station with a cross compiler to vxWorks (could be the
linux computer from the Steve Wood's list)
120 Hz data running:
===================
1.The time to change all DSP Programs (40) is about a few seconds (I would
say 5 as a first hint). Changing from 30 Hz to 120 Hz mode can be done
automatically without operator intervention.
2.We prefer case (b) as proposed by Stephen Pate:
-We propose a 250 ps binning for the 30 Hz regime for the normal
and buddy spectra and a 1ns binning at 120Hz. The front DSP will produce
the same kind of spectra but the concentrator one will sum the bins 4 by 4
at 120Hz. Since the rate allows it, the channel depth will be 16 bits.
The bin width will be 4 times larger at 120Hz than at 30Hz but the amount
of data will be 4 times less , so the data stream will be exactly the same
at 30 and 120 Hz : 1MByte/s.
The data format will be the same with a bit at one in the DSP information
word to make the difference between 30 and 120 Hz. We can always sum up
off-line the bin 4X4 in the 30 Hz running mode (uniform bin width 1 ns) if
we want to compare exactly the spectra with the 120 Hz mode.
-In case (a) we loose the buddy spectra information.
-The (d) case would have a worse binning in the background region,
and we think that it can be interesting to have a good binning for the
noise.
-For the (c) case the spectra are more complicated to produce in
the DSP and the fine binning region would have to be well positioned with
respect to the elastic peak. Since we have the whole 32 ns coded, the
peak position isn't relevant in our case.
Cordially,
Xavier, Jacques, Dominique
grave@ipno.in2p3.fr
33 (0) 1 69 15 79 59
Date: Mon, 1 Nov 1999 10:36:09 -0800
From: W. Des Ramsay
To: quinn@uquark.phys.cmu.edu, pate@NMSU.EDU, lung@jlab.org, saw@jlab.org,
willy@uinpla.npl.uiuc.edu, pitt@vt.edu, carlini@jlab.org,
dhbeck@nialas.npl.uiuc.edu, arvieux@ipno.in2p3.fr, furget@in2p3.fr,
kox@in2p3.fr, vanoers@physics.umanitoba.ca
Cc: RAMS@CSV02.TRIUMF.CA
Subject: different runtimes in freq domain
120Hz people,
On http://www.triumf.ca/E497/g0/, I have added some spectra to show
the effect in the frequency domain of varying the runtime.
Des
Date: Fri, 12 Nov 1999 13:20:55 -0800
From: W. Des Ramsay
Subject: DRAFT report of 120 Hz Task Force
I think the report is clear and very nicely written. One comment I
have is that I notice there is no mention of the possibility of using
the spin sequence +--+-++- or -++-+--+, with the starting state chosen
at random. This method cancels linear and quadratic drifts each 8
state cycle. I joined this group later on, so I assume this has been
discussed, but maybe the report should say something about it.
Des
Date: Sat, 13 Nov 1999 1:40:38 -0500
From: Brian Quinn
Subject: 120 Hz report
Hi Steve, et al,
I have a couple of suggestions for modifications.
In 'why do we flip spin using a random sequence of helicity states?', the
report mentions the problem of monotonic changes in experimental conditions.
A much worse (deadly) problem would result from ANYTHING which oscillated with
*** the same frequency*** as our spin flip. (If we flipped the spin regularly
at 50 Hz, the G0 experiment could probably detect Europe just by it's line
power radiation.) Flipping pseudo-randomly ensures that NOTHING can change in
synch with the helicity (as long as nobody cheats and gets undelayed helicity
info from the source).
You might also address Des' question by mentioning that we will randomly
choose a +--+ or -++- pattern at the beginning of each quartet of macropulses.
(It's not envisioned that we'll use +--+-++- or -++-+--+, so the cancellation
is only linear, not quadratic.)
At the end of 'What about the phase slip between data-collection periods?',
you asy we will precess through all phases and helicites in 166 consecutive
data-collection periods. I agree about the phases, but I don't understand
what is meant about the helicities, which are random.
In 'What are the "buddy" spectra for?' there are a couple of points.
One clear mis-statement is that 'the input to the meantimer is blocked'.
Actually it is the input latch of the LTD that is 'blocked' in the NA and
Grenoble boards at least (I don't remember exactly how Orsay works, but I
think it is the TDC input that is disabled. The meantimer still sees the hits
and may suffer deadtime or pileup.
Although it's true that this latching mechanism would be needed if there
were no oneshot in the meantimer, the most important role of the enforced
deadtime is to ensure a controlled (fixed) deadtime, not one which might
depend on beam parameters.
The wording of your description of buddies gives the impression that
a coincidence between one FP detector and a different backup detector is
what is being monitored. That is not correct. A buddy scaler is incremented
if BOTH the focal plane detector and backup detector fire (ie a perfectly
good hit) while the buddy LTD channel is busy (presumably because it also saw
a perfectly good hit (although the busy can be set if either the FP or backup
detector fires.) We would like to monitor how often a channel fires (all four
tubes in perfectly good meantimed coincidence) while the electronics is
unable to see it. Since we can't do that, we measure how often a channel
fires (still all four tubes with proper timing) while a similar channel would
have been unable to see it. ...that similar channel is the buddy.
A couple of small points:
In "What can we learn from collecting spectra at 120 Hz instead of 30 Hz?"
You say the A-B+C-D will be maximum at zero degrees. That is true only if
we DEFINE zero to be the phase which maximizes A-B+C-D.
I'm not sure what it means to measure the instantaneous line frequency
beyond measuring the time of zero-crossing relative to the clock that
regulates the beam pulses. Howdo items 4 and 5 of the recommendations
differ?
Other than that, it looks good.
Take care,
Brian
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Wed, 1 Dec 1999 15:32:04 -0700 (MST)
From: pate@lepton.nmsu.edu
Subject: Re: 120 Hz report
Dear Brian (et al.)
Thanks to Brian (and Des) for your comments. I am preparing a new
version of the draft, which I hope will be the final version. I thought I
would comment on some of your remarks.
> In 'why do we flip spin using a random sequence of helicity
> states?', the report mentions the problem of monotonic changes in
> experimental conditions. A much worse (deadly) problem would result
> from ANYTHING which oscillated with *** the same frequency*** as our
> spin flip. (If we flipped the spin regularly at 50 Hz, the G0
> experiment could probably detect Europe just by it's line power
> radiation.) Flipping pseudo-randomly ensures that NOTHING can change
> in synch with the helicity (as long as nobody cheats and gets
> undelayed helicity info from the source).
> You might also address Des' question by mentioning that we will
> randomly choose a +--+ or -++- pattern at the beginning of each
> quartet of macropulses. (It's not envisioned that we'll use +--+-++-
> or -++-+--+, so the cancellation is only linear, not quadratic.)
I made changes to reflect these comments, and this covers Des' comment
too of course.
>
> At the end of 'What about the phase slip between data-collection periods?',
> you asy we will precess through all phases and helicites in 166 consecutive
> data-collection periods. I agree about the phases, but I don't understand
> what is meant about the helicities, which are random.
On average, all combinations of helicities and phases will get sampled in
that time. The statement is exactly true for the phases, but true only in
average for the helicities.
>
> In 'What are the "buddy" spectra for?' there are a couple of points.
> One clear mis-statement is that 'the input to the meantimer is
> blocked'. Actually it is the input latch of the LTD that is 'blocked'
> in the NA and Grenoble boards at least (I don't remember exactly how
> Orsay works, but I think it is the TDC input that is disabled. The
> meantimer still sees the hits and may suffer deadtime or pileup.
> Although it's true that this latching mechanism would be needed if
> there were no oneshot in the meantimer, the most important role of the
> enforced deadtime is to ensure a controlled (fixed) deadtime, not one
> which might depend on beam parameters.
> The wording of your description of buddies gives the impression that
> a coincidence between one FP detector and a different backup detector
> is what is being monitored. That is not correct. A buddy scaler is
> incremented if BOTH the focal plane detector and backup detector fire
> (ie a perfectly good hit) while the buddy LTD channel is busy
> (presumably because it also saw a perfectly good hit (although the
> busy can be set if either the FP or backup detector fires.) We would
> like to monitor how often a channel fires (all four tubes in perfectly
> good meantimed coincidence) while the electronics is unable to see it.
> Since we can't do that, we measure how often a channel fires (still
> all four tubes with proper timing) while a similar channel would have
> been unable to see it. ...that similar channel is the buddy.
This has been fixed up based on our private communications.
>
> A couple of small points:
> In "What can we learn from collecting spectra at 120 Hz instead of 30 Hz?"
> You say the A-B+C-D will be maximum at zero degrees. That is true only if
> we DEFINE zero to be the phase which maximizes A-B+C-D.
Yes, that's right. I have changed the wording slightly.
> I'm not sure what it means to measure the instantaneous line frequency
> beyond measuring the time of zero-crossing relative to the clock that
> regulates the beam pulses. Howdo items 4 and 5 of the recommendations
> differ?
They don't differ in regards to what you might actually do in some
electronics. They do differ in what we recommend to be recorded.
The new version can be found at
http://www.jlab.org/~pate/g0run/g0_120hz.ps
Everyone: Please comment within one week.
yours,
Steve
Date: Fri, 3 Dec 1999 17:05:51 +0100 (MET)
From: BIMBOT Louis
To: 120HzTaskForce
Subject: About the draft report from the 120Hz task force
==============================================================
Version of 12-03-1999
(Louis BIMBOT Dominique MARCHAND)
General issues addressed to management:
About 120 Hz task force report and other considerations
First we want to thank the task force people and especially Steve PATE
for trying to make things clear and written about the 120 Hz aquisition
periods in the G0 experiment.
Before making comments and asking questions related to the draft report
we would like to take the opportunity to propose to build similar task
forces on the following subjects:
1)- The NPN mode
2)- The "Buddy" method
3)- The compensation time in mean timers
4)- The use of Laser light for calibrations
(as you must know with the Orsay electronics it is possible to run with
laser or generator permanently for crosschecks during the data taking
using the 200 microseconds of helicity change. Is it worth being used ?
We say yes but what about you?)
We plan to make a first contribution for the end of January on the
first 3 subjects to initiate exchanges of points of view and
to go if possible to conclusions.
Comments on the 120Hz report:
What we think that we have clearly understood:
1) The DAQ will be synchronised on an independent clock running at exactly
60 Hz whatever the electrical power is!
2) The DAQ in normal mode will be done after every exact 1/30th of a
second of data taking plus some 200 microseconds used for data transfer
and helicity change. So we will introduce a regular shift between the
DAQ-clock and the Electrical Power so that after a while all current
phases will be covered for the same "kind" (helicity state) of data
acquisition.
3) The normal DAQ is done by 1/30th of second because we need frequent
helicity changes to minimize the effect of slow changes of any kind
(temperature, HV,LV, etc)
4) The 120 Hz DAQ mode is supposed to look at independent phases
of the alternating current.
What is not clear
1) Why the DAQ rate cannot be 1/60th of second (Final volume of data ?
Time lost by helicity change?)
2) The paragraph on BUDDY stuff is introduced because it gives the
estimation and possible corrections for variable counting rates and
so is relevant also to do 120 Hz measurements. But the description
given seems erroneous and inexact (see B.Quinn email) . The whole
chapter need to be rewritten in a clearer manner saying what signals
are to be used, what corrections are possible and planned to be done
and how will they be taken into account.?
3) If the beam is very stable what deadtime corrections are planned?
And what if it is not stable? The hope that everything will count
the same and that no correction will be necessary seems optimistic.
We hope that these comments will arrive in time to be taken into account
for the next version of the reports.
Best regards from the ORSAY ELECTRONICS WORKING GROUP.
=============================================================================
Louis BIMBOT
IPN Orsay B.P. n°1 F-91406 Orsay FRANCE
Tel: 33 1 69 15 51 99
Fax: 33 1 69 15 64 70
Date: Fri, 3 Dec 1999 10:22:50 -0800
From: W. Des Ramsay
Subject: Stability of 60 Hz.
Dear Data taking people,
This does not affect the 120Hz Task Force Report, but I got some
information from BC hydro about the system frequency, which I think you
will find very interesting. I talked to Brent Hughes, a BC hydro
engineer, and he sent me two days of system frequency measurements and
some information on how and why they vary the frequency. Please look at
http://www.triumf.ca/ramsay/g0 and see what you think.
I think the main conclusion is that the accuracy of our integration gate
only has to be better than the part per thousand or so short term
stability of the line frequency. I assume all North American utilities
are similar.
Des
P.S. An unrelated question for Brian (and, as he says, all the others
caught on this list): What is the disadvantage of using an 8-state
(+--+-++-) cycle instead of the proposed 4-state (+--+) cycle?
Date: Fri, 3 Dec 1999 18:20:15 -0500
From: Brian Quinn
Subject: a few thoughts
Dear 60Hz'ers,
To answer Des's question: I don't know of any particular advantage of
4-macropulse sequences over 8. That plan predates me in the collaboration.
I guess arguments can be made for sequences of 8, +--+-++- and inverse,
or 16 +--+-++--++-+--+ and inverse or 32 etc. Longer seqences, in principle,
kill off higher order polynomial drifts but result in longer sequences between
random changes. Where's the point of diminishing returns? I don't know, and
I have no strong preference for any one sequence length over another. (You
notice the subject line says 'a few thoughts', not 'a few answers'.)
Des' info (text, that is, I got error 404 trying to look at the plot) about
the stability of the 60Hz fits with my own experience from plugging a
multimeter into line power and looking at it from time to time. I saw numbers
ranging from about 59.97 to 60.03 Hz varying changing over time periods of
minutes. I didn't mention it earlier because I'm not very sure of the
stability of the clock in the fluke mutlimeter so I wasn't convinced that I
was making meaningful measurements.
As for the question of why we run at 1/30 s rather than 1/60 s, I'm not sure
I can make a strong argument for 1/30 s. It is probably safer. Certainly we
would not want our acquisition efficiency to have a large 60 Hz component.
So 1/120 s is a bad period to run at except for intentional investigation of
60 Hz. I'm not sure I can make a strong argument that a 1/60 Hz gate must
intoduce a big 60 Hz component. It seems like potentially asking for trouble
since the basic period of the data acquition would then be 60 Hz, so 1/30 s
'feels' safer. Of course there is also the argument that the data volume
would double. Is there an agument in favor of switching at 1/60 s?
I suspect that there may be a stronger argument against 1/60 s than I can
think of at the moment... Doug, do you want to field that?
Best regards,
Brian
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Sat, 4 Dec 1999 22:10:21 -0500
From: Roger D. Carlini
Subject: Practical Issues
Just to pass on a comment and some info. Reality check! First, by now
everyone must realize that you can't do a high precision parity
experiment by reversing the helicity at the AC line frequency (60Hz).
Secondly, after discussions with both Larry and Charlie I believe it
is fairly safe to say that G0 will most likely be running (other than
for short diagnostics runs) with a pseudo-random reversal rate of
30Hz locked to the line frequency with delayed state reporting.
Remember, there are other users of the machine and to most if not all
of them any arguments against syncing the reversal to the line will
be viewed (to be polite) as bogus. This is because of the obvious
reason that all line induced beam properties are by definition locked
to the line and best suppression is achieved by a reversal rate of
30Hz locked to the line. The gating of the experiment is of course
entirely up to us.
Best Regards
Roger Carlini
Date: Sun, 5 Dec 1999 15:05:33 -0700 (MST)
From: pate@lepton.nmsu.edu
Subject: Re: About the draft report from the 120Hz task force
Dear Louis and Dominique
Thanks for your thoughtful comments.
>
> What is not clear
>
> 1) Why the DAQ rate cannot be 1/60th of second (Final volume of data ?
> Time lost by helicity change?)
This is a good question. Most of the answers I have heard consist of
statements like "It is a bad idea to collect data at 60 Hz due to the 60
Hz noise from the mains supply." Certainly it seems like a very good idea
to stay away from such a large noise frequency. But I recall that Des'
experiment (E497) is running at 60Hz. I suppose that Des is not trying to
measure an asymmetry as small as in G0, however. Maybe Des could comment
on that.
I think that "60Hz is noisy" is a good reason not to run the DAQ at 60Hz
myself. If anyone knows of an additional reason, it would be good to hear
it.
> 2) The paragraph on BUDDY stuff is introduced because it gives the
> estimation and possible corrections for variable counting rates and
> so is relevant also to do 120 Hz measurements. But the description
> given seems erroneous and inexact (see B.Quinn email) . The whole
> chapter need to be rewritten in a clearer manner saying what signals
> are to be used, what corrections are possible and planned to be done
> and how will they be taken into account.?
Well, this paragraph (in the latest version) was written by Brian and I.
I guess it still is not clear enough, so Brian and I will try to
improve it.
> 3) If the beam is very stable what deadtime corrections are planned?
> And what if it is not stable? The hope that everything will count
> the same and that no correction will be necessary seems optimistic.
I think the identity of the signals that will be available for calculating
a deadtime correction was made clear in the report. But perhaps Brian and
I need to work to make it even more clear. I don't think we meant to
imply that we expect everything will count the same and that no correction
will be necessary -- we simply meant that this would be the best
situation.
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Sun, 5 Dec 1999 15:50:52 -0700 (MST)
From: pate@lepton.nmsu.edu
Subject: Re: Practical Issues
Dear Roger et al.
I don't understand how "the gating of the experiment is of course
entirely up to us" if we are told that the reversal will definitely be
30Hz locked to the line frequency. Our gating will of course have to
follow along with the prescribed reversal schedule.
Regarding suppression schemes, there is nothing "bogus" about the
method proposed in recommendation #3 of our report -- it is simply a
different (but equally valid) method. Since both are valid, there is no
reason to make a fight over it, so I propose that we change the report to
recommend reversing the helicity at 30Hz locked to the line. Any
objections?
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Mon, 6 Dec 1999 11:36:55 -0500
From: Roger D. Carlini
Subject: Re: Practical Issues
Greetings:
Sounds good lets ask for 30Hz line locked. Certainly, if the resersal
is done at 30Hz and locked to the line, then the experiment can gate
off the detector both slightly before and after the actual reversal
takes place. Therefore, the experiment data acquisition (live) gate
can be as precisely timed as we want. This is what I mean by the
"gate".
I still believe that a free running reversal is at best harmless and
probably is poorer than a line locked reveral since it has much less
common mode rejection. That is the degree to which the beam
properties (integrated current for example) of the two helicity
states are the same. This goes back to the issue of effective
linearity of our electronics.
The experiment will have to defend it requirements with hard numbers
and the weakest case we have for a free running reversal is that a
line locked reversal works great for HAPPEX.
Regards
Roger
Date: Mon, 6 Dec 1999 07:45:17 -0500 (EST)
From: Mark Pitt
Subject: Re: Practical Issues
Steve,
> different (but equally valid) method. Since both are valid, there is no
> reason to make a fight over it, so I propose that we change the report to
> recommend reversing the helicity at 30Hz locked to the line. Any
> objections?
>
I just want to be sure I understand your language. I think we all agreed
on the following: the data collection period should be equal to twice
the average line period (about 1/30 sec.), so that we get the most
complete cancellation of 60 Hz noise possible. Since we know we want the
pause during helicity flipping, the actual helicity flip rate will be
closer to about 29.8 Hz, and its phase will slip continuously relative to
the line. The polarized source is currently configured so that the data
collection period PLUS the helicity flip pause is equal to twice the
average line period. Thus our recommendation would require a change to
the way the helicity flipping is currently done. We should probably
mention that explicity in the report, since it then becomes an action
item.
I have discussed all this informally with Matt Poelker and John
Hanskenecht, but I haven't mentioned it to Charlie. If we plan to pursue
it, I assume it would require the agreement of Charlie and the hall
leaders since it is a change to the way things are currently implemented.
Mark
Date: Mon, 6 Dec 1999 05:58:56 -0700 (MST)
From: pate@lepton.nmsu.edu
Subject: Re: Practical Issues
Dear Mark
Thanks for the message -- you are right, the current running mode
for the source doesn't obtain the best noise cancellation, since it
doesn't integrate over exactly two cycles of 60 Hz. So we do need to make
a proposal to Charlie about how we want to run.
That throws open again the question of which method (running for
twice the line period or for a fixed time period) we will use for the
experiment. Maybe the best idea is to propose both methods to Charlie et
al. and together work out which one is best, all things considered.
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Mon, 6 Dec 1999 12:20:54 -0500
From: Brian Quinn
Subject: Re: Practical Issues
Dear Steve, Roger, Mark, et al.
Right. I agree with Mark (and now Steve). The important part of the
decision is a 1/30 s data collection time. That is definitely not
consistant with a 30 Hz period, whether locked to the line or set by an
external clock. I certainly don't think we should cave and recommend what
we don't want. I have no idea why Roger is now describing this as bogus... we
want to minimize noise at 60 Hz, it makes sense to acquire data in multiples
of the period of the noise.
The arguments for reversing at a fixed rate, rather than at a period
related to the instantaneous line frequency, are not directly related to noise
minimization. It's much simpler and cleaner. But, to a small extent, the
noise reduction should be better if the 'gated-on' period (or 'stable beam
helicity period', if that's a better way of emphasizing that it is an
accelerator-related quantity, not just an experimentor's decision) is twice
the period of the line, rather than 1/30 s. It's important to note that this
still wouldn't lock us to the line frequency! It just adds the complication
of building a circuit which measures the instantaneous period (integrated over
some 'reasonable' number of periods to minimize inaccuracy without integrating
away the instantaneous character) and modifies the 'stable beam helicity
period' accordingly. We would then need to keep track of the duration of each
macropulse in order to monitor intensity fluctuations. But I don't see any
fundemental objection to having the length of a macropulse fluttering around,
even within a quartet of macropulses, as long as we monitor it.
Regards,
Brian
Email: bquinn@cmu.edu Physics Department
Phone: (412)-268-3523 Carnegie Mellon University
Fax: (412)-681-0648 Pittsburgh, PA, 15213 USA
Date: Mon, 6 Dec 1999 10:12:28 -0800
From: W. Des Ramsay
Subject: reality and phase slip
Here are some remarks on the data taking:
1) If Roger is right that we have to lock to the line, then it's too
bad we didn't know this earlier. We could have dispensed with much of
the discussion.
2) Why does it matter to Jlab or to other experimenters whether the
flip is locked to the line or has a controlled slip?
3) The use of the term "Hz" for flip rate is confusing. I don't think
we should use it. For example, a regular spin sequence of 1/60 s per
state is really 30 "Hz" and will be sensitive to 30 Hz noise, not 60
Hz noise. If you integrate for 1/60 second you are blind to 60 Hz and
all its harmonics.
3) In E497 we integrate for 1/60 second, but flip every 1/40 second
because we have to measure transverse polarization profiles every
spin state. Our controlled phase slip is very slow, 360 degrees in 20
minutes. It never showed any sign of 60Hz contamination, so in
retrospect I guess you could say we didn't need it -- but it
reassured us.
4) The Swiss (Simonius, Haeberli,...) also integrated for exactly one
line cycle (1/50 second) and had controlled phase slip. Like us, they
were going for 2 X 10^-8 precision on Az.
5) If you are worried about what frequencies are the most noisy,
remember that flipping every 1/30 second makes you most sensitive to
noise around 15Hz.
6) My guess is that flipping locked to the line or flipping with a
phase slip will both give equally good data. The phase slip just
makes us more sure of it.
Des
Date: Tue, 7 Dec 1999 14:32:35 -0800
From: W. Des Ramsay
Subject: Line locked acquisition
Data acquisition people,
Here are a couple more remarks regarding my impression that both the
line-locked and non-line-locked systems should give good data:
1) If we use a 1/30 second absolute gate, we cancel effects of line
induced beam properties for each spin state -- to the extent that 1/30
second is really 2 cycles. Typically, the gate will be within about a
part in 2000 of its ideal length. Small residual contamination due to
imperfect gate length will be further greatly reduced after a +- or -+
pair.
2) If we flip locked to the line our gate will be about 1% too short
for perfect cancellation on each spin state, but after a +- or -+
pair, line noise will cancel -- to the extent that the noise is
constant for two states.
If we must run line locked, would it be possible to choose our phase?
This might prove to be a useful diagnostic.
Des
Date: Wed, 8 Dec 1999 07:26:44 -0700 (MST)
From: pate@lepton.nmsu.edu
Subject: Re: Line locked acquisition
Dear Des
> 2) If we flip locked to the line our gate will be about 1% too short
> for perfect cancellation on each spin state, but after a +- or -+
> pair, line noise will cancel -- to the extent that the noise is
> constant for two states.
If there is a coupling between helicity-correlated beam motion and 60Hz
beam motion, then the noise in the two states won't necessarily be
constant.
If we fix the reversal to be always at the same phase point (I assume
that's what we mean by "locked to the line"), and don't collect data for a
full two cycles, then it seems to me that we are _introducing_ a coupling
between the helicity and 60Hz. The helicity-correlated beam motions will
always occur in phase with 60Hz, so the "1%" left out won't necessarily be
the same in the two states.
The nice thing about the phase slip (in the "integrate for two full
cycles" plan) is that it destroys any correlation between helicity-related
effects and 60 Hz effects, and they can then be treated separately.
Integrating for two full cycles takes care of the 60Hz effects, and then
we are left only with making sure the helicity-correlated beam motions are
small. We have already seen that the JLab beam has excellent properties
in this regard.
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Wed, 8 Dec 1999 13:04:03 -0700 (MST)
From: pate@lepton.nmsu.edu
Subject: latest (final?) version of the 120Hz report
Dear Task Force Members
I changed recommendation 3 to reflect the two competing (but
equivalent) methods of running for "1/30 s" -- running for a fixed 1/30 s
period, or running for two full periods of the instantaneous line voltage
cycle.
I did not change the "buddy" section since that discussion is
somewhat peripheral to the report. Our Electronics Subsystem Manager
should provide a more complete description of the buddy system in a
separate document.
I hope that this can be the final version of the report -- I will
entertain comments for one more week. You can find it at:
http://www.jlab.org/~pate/g0run/g0_120hz.ps (dated 8-December-1999)
yours,
Steve
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
Stephen Pate Department of Physics office: 505-646-2135
Box 30001, Dept. 3D dept: 505-646-3831
New Mexico State University fax: 505-646-1934
pate@nmsu.edu Las Cruces NM 88003 home: 505-382-6880
Date: Wed, 8 Dec 1999 16:11:57 -0500
From: Roger D. Carlini
Subject: Re: Line locked acquisition
Greeting:
I am not convinced that your conclusions below will in practice prove
correct. To make such conclusions you need to either do some careful
measurements with a spectrum analyzer and/or a rigorous numerical
analysis. Again, the dominant issue here may be the effective
linearity of the system (which we don't yet know) - so good common
mode rejection at 30Hz may be critical. Loosing 1% due to gating will
effect us in both reversal schemes.
>Dear Des
>
> > 2) If we flip locked to the line our gate will be about 1% too short
> > for perfect cancellation on each spin state, but after a +- or -+
> > pair, line noise will cancel -- to the extent that the noise is
> > constant for two states.
>
>If there is a coupling between helicity-correlated beam motion and 60Hz
>beam motion, then the noise in the two states won't necessarily be
>constant.
>
>If we fix the reversal to be always at the same phase point (I assume
>that's what we mean by "locked to the line"), and don't collect data for a
>full two cycles, then it seems to me that we are _introducing_ a coupling
>between the helicity and 60Hz. The helicity-correlated beam motions will
>always occur in phase with 60Hz, so the "1%" left out won't necessarily be
>the same in the two states.
>
>The nice thing about the phase slip (in the "integrate for two full
>cycles" plan) is that it destroys any correlation between helicity-related
>effects and 60 Hz effects, and they can then be treated separately.
>Integrating for two full cycles takes care of the 60Hz effects, and then
>we are left only with making sure the helicity-correlated beam motions are
>small. We have already seen that the JLab beam has excellent properties
>in this regard.
>
> yours,
> Steve
Comments and suggestions to Stephen Pate,
pate@nmsu.edu
Last modified: Wed Dec 8 16:17:44 1999