40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  40m Log  Not logged in ELOG logo
Entry  Fri Aug 3 15:53:25 2018, gautam, Update, SUS, Low noise bias path idea LT1055_precOpAmp.pdf
    Reply  Fri Aug 3 16:27:40 2018, rana, Update, SUS, Low noise bias path idea 
       Reply  Sun Aug 5 15:43:50 2018, gautam, Update, SUS, Another low noise bias path idea HV_Bias_schematic.pdfTF.pdfbias.pdfHVbias_currentNoise.pdf
          Reply  Wed Aug 8 23:06:59 2018, gautam, Update, SUS, Another low noise bias path idea 
             Reply  Wed Aug 15 19:18:07 2018, gautam, Update, SUS, Another low noise bias path idea HV_Bias.pdfHVamp_TF.pdfHVamp_noises.pdfcurrentNoises.pdfHVamp.fil.zip
                Reply  Thu Aug 16 23:06:50 2018, gautam, Update, SUS, Another low noise bias path idea HVamp_schem.PDFHvamp.zip
                   Reply  Mon Oct 1 22:20:42 2018, gautam, Update, SUS, Prototyping HV Bias Circuit CoilDriverBias.pdfcurrentNoise.pdfPSRR.pdf
Message ID: 14169     Entry time: Thu Aug 16 23:06:50 2018     In reply to: 14165     Reply to this: 14223
Author: gautam 
Type: Update 
Category: SUS 
Subject: Another low noise bias path idea 

I had a very fruitful discussion with Rich about this circuit today. He agreed with the overall architecture, but made the following suggestions (Attachment #1 shows the circuit with these suggestions incorporated):

  1. Use an Op27 instead of LT1128, as it is a more friendly part especially in these composite amplifier topologies. I confirmed that this doesn't affect the output voltage noise at 100 Hz, we will still limited by Johnson noise of the 15kohm series resistor.
  2. Take care of voltage distribution in the HV feedback path
    • I overlooked the fact that the passive filtering stage means that the DC current we can drive in the configuration I posted earlier is 150V / 25kohm = 6mA, whereas we'd like to be able to drive at least 10 mA, and probably want the ability to do 12 mA to leave some headroom.
    • At the same time, the feedback resistance shouldn't be too small such that the PA91 has to drive a significant current in the feedback path (we'd like to save that for the coil).
    • Changing the supply voltage of the PA91 from 150 V to 320 V, and changing the gain to x30 instead of x15 (by changing the feedback resistor from 14kohm to 29kohm), we can still drive 12 mA through the 25 kohms of series resistance. This will require getting new HV power supplies, as the KEPCO ones we have cannot handle these numbers.
    • The current limiting resistor is chosen to be 25ohms such that the PA91 is limited to ~26 mA. Of this, 300V / 30kohm ~ 10 mA will flow in the feedback path, which means under normal operation, 12 mA can safely flow through the coils.
    • Rich recommended using metal film resistors in the high voltage feedback path. However, these have a power rating, and also a voltage rating. By using 6x 5kohm resistors, the max power dissipated in each resistor is 50^2 / 5000 ~ 0.5 W, so we can get 0.6 W (or 1W?)  rated resistors which should do the job. I think the S102K or S104K series will do the job.
  3. Add a voltage monitoring capability.
    • This is implemented via a resistive voltage divider at the output of the PA91.
    • We can use an amplifier stage with whitening if necessary, but I think simply reading off the voltage across the terminating resistor in the ladder will be sufficient since this circuit will only have DC authority.
  4. Make a Spice model instead of LISO, to simulate transient effects.
    • I've made the model, investigating transients now.
  5. High voltage precautions:
    • When doing PCB layout, ensure the HV points have more than the default clearance. Rich recommends 100 mils.
    • Use a dual-diode (Schottky) as input protection for the Op27 (not yet implemented in Spice model).
    • Use a TVS diode for the moniotring circuit (not yet implemented in Spice model).
    • Make sure resistors and capacitors that see high voltage are rated with some safety margin.
  6. Consider using the PA95 (which Rich has tested and approves of) instead of the PA91. Does anyone have any opinions on this?

If all this sounds okay, I'd like to start making the PCB layout (with 5 such channels) so we can get a couple of trial boards and try this out in a couple of weeks. Per the current threat matrix and noises calculated, coil driver noise is still projected to be the main technical noise contribution in the 40m PonderSqueeze NB (more on this in a separate elog).

Quote:

Looks like this will do the job. I'm going to run this by Rich and get his input on whether this will work (this design has a few differences from Rich's design), and also on how to best protect from HV incidents.

Attachment 1: HVamp_schem.PDF  266 kB  Uploaded Fri Aug 17 00:42:27 2018  | Hide | Hide all
HVamp_schem.PDF
Attachment 2: Hvamp.zip  1 kB  Uploaded Fri Aug 17 00:42:34 2018
ELOG V3.1.3-