From: 40m-admin@ligo.caltech.edu on behalf of Alan Weinstein
[ajw@hep.caltech.edu]
Sent: Tuesday, June 15, 2004 3:02 PM
To: 40m@ligo.caltech.edu; aligo_systems@ligo.caltech.edu
Cc: 'David B. Tanner'; 'David Reitze'; varvella@lal.in2p3.fr; 'Volker
Quetschke'; desalvo@ligo.caltech.edu; 'Alan J Weinstein'
Subject: [40m] Minutes of telecon on Mach-Zehnder

Minutes of telecon on
Mach-Zehnder interferometer to eliminate sidebands of sidebands
for 40 meter and Advanced LIGO

Tues June 15, 11 am PDT (2 pm EDT)

Attending:
Osamu Miyakawa, Seiji Kawamura, Steve Vass, Riccardo DeSalvo, 
Shihori Sakata, Bob Taylor, Rob Ward, Alan Weinstein, 
Ben Abbott, Dennis Coyne, Bill Kells, Hiro Yamamoto, Jay Heefner, 
Monica Varvella, Dave Reitze, Dave Tanner, Volker Quetschke, 
Peter Fritschel, Nergis Mavalvala, ... who else?

Minutes: Alan Weinstein.
The telecon was recorded; see separate email from Dennis Coyne
on playback instructions.

Osamu went through the transparencies in
http://www.ligo.caltech.edu/~omiyaka/20040615_MZ.ppt
http://www.ligo.caltech.edu/~omiyaka/20040615_MZ.pdf .

There was some discussion of the sense in which the sidebands
of sidebands is "amplitude modulated", given that it arises
from pure phase modulation of the input beam.
I enclose a couple of plots that illustrate the effect;
although there is no explicit amplitude modulation,
the net effect is to produce sideband pairs
at f1+f2, and also at f1-f2, whose relative phases
are characteristic of amplitude modulation.

Can this be observed and measured with an optical spectrum
analyzer and/or an RF photodiode? Yes, these are real
sidebands that should be present in a measuring device.

Bryan's note
http://www.ligo.caltech.edu/~cit40m/Docs/SbOfSb.pdf
which distinguishes  "real" sidebands 
("present before detection of the light on a photodetector")
from "hidden" sidebands ("amplitude modulation ... only present 
once the light is detected on a photodetector")
may be an artifact of not consistently taking the expansion
to 2nd order in all terms (I don't think I understand why
that would be so, but, Bryan, can you look at it?)

The proposed modulation depth would be an "effective" 0.1;
and twice that (0.2) is needed at each EOM to achieve the
effective modulation depth, because of the loss of 3/4 of 
the sideband power in the M-Z as compared with series EOMs.
We're not sure whether this will be sufficient to extract
robust signals for lock acquisition and control, but we 
have some headroom, since modulation depths of 0.4 or so
are achievable with these New Focus EOMs.

Bryan explored the possibility of using the series EOM 
RF structure (ie, with sidebands on sidebands) to control
the interferometer. He found that with a careful choice of
demod phases, the signal matrix could become closer to diagonal,
but with significant DC offsets, and with extremely tight
tolerances on the demod phase, which would be very difficult
to establish. Not a robust solution.

We could imagine coping with the very non-diagonal 
signal matrix to control the IFO with a gain heirarchy,
but it would make lock acquisition much more difficult,
and it's already difficult enough!

Is angular stabilization required? We don't think so,
based on experience with the current setup through the 
mode cleaner: angular instability with one beam would produce
noticable fluctuations in the mode cleaner transmitted beam,
which are not observed (and which could be compensated for
with intensity stabilization). Combining 2 beams with a M-Z 
shouldn't make it much worse, unless the PZT mirror causes 
considerable angular instability; no reason to think that it will.

There are many alternative approaches to generating separate 
sidebands: making use of separate beams, with phase-locked
sidebands applied to them separately; 
using separate polarizations; etc. 
In addition, it may not be so difficult to kill the sidebands
of sidebands using EOMs or AOMs...
These can be looked at for more carefully in the future.

Osamu estimated the amplitude noise caused by length fluctuations
in the M-Z, assuming a servo UGF of 100 Hz, no lock point offset,
typical motion of lambda/10 at 1 Hz, falling in frequency
in some way (Osamu will clarify), and then peaking again at 1 kHz 
due to the PZT mirror resonance. 
The amplitude noise is everywhere less than the noise measured
before the M-Z. And, a servo UGF of 100 Hz is quite conservative;
we could probably increase it to 1 kHz, before we run into
PZT mirror/mount resonances.
If this limits us, we can consider employing more stable mounts.

Frequency noise introduced by additional mirrors should not make
things much worse than they are now.

We encourage all interested parties to think of other noise sources
which may be important to consider!

The M-Z is relatively simple, inexpensive, and easy to construct. 
All optics have been procured and have now been installed 
on the 40m PSL table, and rough-aligned. 
They're all jammed together in a very limited space, 
like a "jigsaw puzzle", but it fits. We have a 33 MHz RFPD, 
a PZT mirror actuator (for specs, please see 
http://www.ligo.caltech.edu/~omiyaka/mach1.pdf ),
a slightly modified PMC board for demodulation/servo/pzt driver.
The RFPD and PMC board will be installed, and first commissioning
will be done, this week. Osamu leaves for SPIE in Glasgow next week,
and will finish the commissioning the following week.

We will measure the frequency noise with the mode cleaner
and the intensity noise with a thorlabs PD, and also check
the rf structure with an optical spectrum analyzer
(all of these things are in place, and we have "before" measurements).
We hope to have first results in the next few weeks.

I welcome corrections and additions to these minutes.
Many thanks to all parties for their participation and input!

Alan J Weinstein
California Institute of Technology
256-48 Caltech
Pasadena, CA 91125
ajw@caltech.edu
626-395-6682
fax: 626-795-3951