J. Kissel, [emotional support from L. Dartez remotely]
Executive Summary
Contrary to the grand plan, I instead unplugged and re-plugged in some cables, and lost 4 hours of my life. The OMC DCPD electronics chain remains as it did before today. I'm confused yet again.
Full Summary
After we last left the saga of "the in-vacuum OMC DCPD Transimpedance Amplifier (TIA) frequency response has changed a little since the OMC Swap; let's re-measure it in analog" (LHO:77735), the 2024-May-13 plan was to swap the apparently broken S2300003, D2200215-style OMC DCPD Whitening Chassis, for a spare (I chose S2300002), in hopes that we'd be able to make a new analog measurement of the TIA without the non-linearity in the whitening chassis' measurement setup.
Today, the installed-in-the-racks measurement setup using S2300002 got even more weird :: very different from the exact same setup of the same chassis in the EE shop. "Weird" is
:: 2 [V/V] rather than 1 [V/V], and
:: a very obvious, 10 [%] / 3 [deg] wiggle, frequency response above 1 kHz, rather than "subtle, slightly different nonlinear roll off."
Indeed, to confirm my insanity, I quickly switched over to CH B of S2300003 while there with everything else the same, and it's similarly weird -- and different from the previous in-rack measurements taken only a few weeks ago where I'd probed its subtle non-linearity around 1 [V/V].
Just maddening.
That being said, after dividing the maddening response of the measurement setup out of the data we really want -- the analog measurement of the in-vacuum TIA -- we see a good clean measurement, with finally with the same DC gain as the good old 2023-03-10 measurements. BUT -- DCPD A's response shows change from the best 2023-03-10 measurement data set -- but at *high* frequency rather than "below 25 Hz." I don't think this is real. More on this in a follow-up aLOG.
In the end, after four hours in front of the SR785, Louis and I decided to revert the signal chain back to using S2300003 as had and has been the case for all of O4.
I'll hang my head, eat my hat, put foot in mouth, and head back to the EE shop.
Attached is a collection of plots the shows today's results.
:: Pages 1 & 2, for DCPDA and DCPDB, respectively, the TIA response, with the maddening measurement setup divided out. One can see that the for DCPDB (page 2) that today's 2024-05-28 measurement divided by the currently installed compensation (from 2023-03-10) agrees very well with the 2023-03-10 measurement over compensation. However, for DCPDA (page 1), we see a diverence from the model and old measurement above 1 kHz.
:: Pages 3 & 4, for DCPDA and DCPDB, again show the obscene difference between today's measurement setup's response -- the 2 [V/V], with 10 % / 3 deg wiggle above 1 kHz -- and the previous "bad" result from 2024-04-23, and the originally "I'm happy with that" 2023-03-10 result, both of which are 1 [V/V] with "a little bit of phase loss, but that's it."
:: Page 5 shows the ratio of all of today's measurement setup responses, showing that even from channel to channel and chassis to chassis, I get this exact same response. I even swapped out the SR785 accessory box for another to see if this was the issue, and I got the same response.
The only piece of electronics left as suspect is the SR785. But --
(1) Out of the utmost caution, I run a factory reset on any SR785 I touch before I begin configuring it for measurement, and
(2) This is a *differential* transfer function measurement with the SR785. That means the ratio CH2 (A-B) / CH1 (A-B) is changing. I find it quite hard to believe that this could be happening.
And in case you're suspicious that "Oh, you must have blown the input electronics by driving too hard into it," We're careful to keep the input voltage for each A and B spigot of CH1 and CH2 below the specified 5 [V_pk], and we're watching for its ADC saturations at all times.
So, ya, maddening.
More details in the comments for posterity.