Announcement

Collapse
No announcement yet.

Threshold Pot doesn't work

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Carl-NC
    replied
    The only test is to put the output of the demod on an oscope and see what it looks like. You can change the preamp BW, or change the demod sample width, and see how the noise changes.

    Leave a comment:


  • green
    replied
    Originally posted by Carl-NC View Post
    I assume the preamp is the same as before, which had an output noise of 5.7uv/rtHz. The NBW of the demod is 7.6Hz and its effective gain is 1 so the total integrated preamp noise seen at the output of the demod is 15.7uv rms. This does not include noise generated in the demod... that is the so-called kT/C noise which is equal to v = sqrt(kT/C) = 354nv rms. This is about 1/40th of the preamp noise so we consider it "in the mud" and ignore it.

    This seems to imply that the total output noise of the demod is just 12.5uv rms but that only includes one Nyquist zone. The PI is a wideband system and all that wideband noise gets folded into the demod output. How many Nyquist zones get folded in depend on the BW of the preamp; the wider you make it, the more noise gets folded in. Ferinstance, if the preamp has a BW = 3.2MHz (NBW = 5MHz) and the sample rate is 1kHz (Nyquist BW = 500Hz) then you are folding in 10,000 Nyquist zones. One of your lower BWs -- 94kHz -- only folds in 188 Nyquist zones.

    This may seem depressing but Nyquist zones don't fold equally. The first Nyquist zone has a relative weighting of 1, but after that the other Nyquist zones are sinc-weighted and diminish pretty quickly. You'll need an Excel spreadsheet to figure it all out, you can't simulate this stuff without a periodic steady-state simulator which tends to run $50,000. I don't remember the numbers but something like 50 Nyquist zones doubles the noise, 200 zone triples the noise, etc. So you take that 15.7uv rms above and multiply it by a Nyquist gain factor that you have to figure out.

    One last thing... all this was for a single sample window. If you also have an EFE sample window then the noise goes up by another sqrt(2).

    I don't expect anyone to understand all this, it begs for some graphics and equations to make sense of it.
    Thanks for the reply, some things to think about. Any thoughts on the test? Maybe include a 100us GB sample with a 10us target sample.

    Leave a comment:


  • Carl-NC
    replied
    I assume the preamp is the same as before, which had an output noise of 5.7uv/rtHz. The NBW of the demod is 7.6Hz and its effective gain is 1 so the total integrated preamp noise seen at the output of the demod is 15.7uv rms. This does not include noise generated in the demod... that is the so-called kT/C noise which is equal to v = sqrt(kT/C) = 354nv rms. This is about 1/40th of the preamp noise so we consider it "in the mud" and ignore it.

    This seems to imply that the total output noise of the demod is just 15.7uv rms but that only includes one Nyquist zone. The PI is a wideband system and all that wideband noise gets folded into the demod output. How many Nyquist zones get folded in depend on the BW of the preamp; the wider you make it, the more noise gets folded in. Ferinstance, if the preamp has a BW = 3.2MHz (NBW = 5MHz) and the sample rate is 1kHz (Nyquist BW = 500Hz) then you are folding in 10,000 Nyquist zones. One of your lower BWs -- 94kHz -- only folds in 188 Nyquist zones.

    This may seem depressing but Nyquist zones don't fold equally. The first Nyquist zone has a relative weighting of 1, but after that the other Nyquist zones are sinc-weighted and diminish pretty quickly. You'll need an Excel spreadsheet to figure it all out, you can't simulate this stuff without a periodic steady-state simulator which tends to run $50,000. I don't remember the numbers but something like 50 Nyquist zones doubles the noise, 200 zones triples the noise, etc. So you take that 15.7uv rms above and multiply it by a Nyquist gain factor that you have to figure out.

    One last thing... all this was for a single sample window. If you also have an EFE sample window then the noise goes up by another sqrt(2).

    I don't expect anyone to understand all this, it begs for some graphics and equations to make sense of it.
    Last edited by Carl-NC; 10-14-2021, 04:09 PM. Reason: corrected a number

    Leave a comment:


  • green
    replied
    Originally posted by green View Post
    Don't know how to calculate noise at integrator out. Thinking of doing a test with my circuit to try and learn something. Coil not connected, noise caused be amplifier and the two 1k input resistors. Record p-p volts at amplifier out and integrator out at different amplifier cutoff frequencies with two different sample times. Any suggestions for improving the test?

    amplifier cutoff set by the capacitors I have
    A test with amplifier cutoff=3.2MHz, 1000pps(Tx and Rx not connected), 10us sample(both target and EF sample). Schematic reply#88. Integrator gain=1, second stage gain=10(added to get enough scope signal, 2mV/div scope lower limit). Anyone know the formula to calculate integrator out noise from amplifier out noise? Plan on doing the other test points but would like to know if the first test point makes sense or do I need to do something different.
    Attached Files

    Leave a comment:


  • green
    replied
    Don't know how to calculate noise at integrator out. Thinking of doing a test with my circuit to try and learn something. Coil not connected, noise caused be amplifier and the two 1k input resistors. Record p-p volts at amplifier out and integrator out at different amplifier cutoff frequencies with two different sample times. Any suggestions for improving the test?

    amplifier cutoff set by the capacitors I have
    Attached Files

    Leave a comment:


  • green
    replied
    Originally posted by Carl-NC View Post
    That sounds a bit high. For TP5 you would take the noise at TP4 but leave it in nv/rtHz, then add that to the input-referred noise created within the demod, multiplied by the demod's noise gain. But this is a wideband system and the noise from multiple Nyquist zones get folded in, but with a sinc (sin(x)/x) roll-off effect which roughly doubles the noise. After the demod remaining circuit noise is added in the normal way.
    Minimum noise is with Rx input shorted. Thinking main noise sources are 1k input resistor and amplifier noise. Think I know how to calculate RMS noise at TP4, not TP5. Guessing amplifier BW and sample time effect noise level. Maybe EF sample spacing between target samples, centered or not. Maybe others? Using 1k input resistor and NE5534 amplifier with Rx input shorted and 1000pps could someone calculate RMS noise at TP4 and TP5. Should I be considering some other parameters?

    amplifier R feedback gain BW sample time TP4 TP5
    single stage 1 meg 1000 10k 10us
    single stage 1meg 1000 10k 40us
    two stage 31.62k 1000 316k 10us
    two stage 31.62k 1000 316k 40us
    Attached Files

    Leave a comment:


  • Carl-NC
    replied
    You are correct, I was thinking in terms of RMS noise. So, yes, if you put an oscope on TP4 you should see about 3.4mv pp of noise (0.57mv rms). But the BW of the preamp isn't the BW of the system so it's misleading. The BW of the system is set by the demod (and often other downstream amps) to be in the range of 15-50 Hz. So you leave the noise in nV/rtHz until you apply the final system BW. But, again, if you actually probe TP4 then you need to apply the preamp BW to match what shows up on the oscope.

    Leave a comment:


  • green
    replied
    Originally posted by Carl-NC View Post
    That sounds a bit high. For TP5 you would take the noise at TP4 but leave it in nv/rtHz, then add that to the input-referred noise created within the demod, multiplied by the demod's noise gain. But this is a wideband system and the noise from multiple Nyquist zones get folded in, but with a sinc (sin(x)/x) roll-off effect which roughly doubles the noise. After the demod remaining circuit noise is added in the normal way.
    How I calculated it, not saying it's correct. Uncorrelated noise voltages add in a "root sum of squares" manner; i.e., noise voltages V1, V2 & V3 add to give a result of . Noise powers, of course, add normally. Thus, any noise voltage which is more than 3 to 5 times any of the others is dominant, and the others may generally be ignored, which simplifies noise assessment. from Analog Devices : Rarely Asked Questions (RAQs) : Operational Amplifier Noise (wisc.edu). Input resistor 1K=4nv/rtHz, amplifier=4nv/rtHz. (16+16)^.5=5.7nv/rtHz. p-p=6*noise voltage. p-p=5.7*6*1000(gain)*100(BW^.5)=3.42.

    Another problem is how to determine p-p voltage. If I had an analog scope maybe I could use a low sweep rate. With my digital scope I use persistence to display many samples at a higher sweep rate to be sure I sampled all the peaks. A scope picture of my amplifier out, not the same gain and BW as HH. The last sample shows in the many stored samples. Maybe there is a better way?

    Maybe someone could measure noise at HH amplifier out with Tx disabled and Rx input shorted.
    Attached Files
    Last edited by green; 09-23-2021, 03:05 PM. Reason: added sentence

    Leave a comment:


  • Carl-NC
    replied
    That sounds a bit high. For TP5 you would take the noise at TP4 but leave it in nv/rtHz, then add that to the input-referred noise created within the demod, multiplied by the demod's noise gain. But this is a wideband system and the noise from multiple Nyquist zones get folded in, but with a sinc (sin(x)/x) roll-off effect which roughly doubles the noise. After the demod remaining circuit noise is added in the normal way.

    Leave a comment:


  • green
    replied
    Originally posted by green View Post
    Thinking there should be a way to calculate(estimate)peak to peak noise with Rx input shorted. Anyone have a formula or a place to look? All PI's have noise, would be nice to know if noise is higher than expected before trying to lower it.
    With Tx disabled and Rx input shorted I calculate about 3.5mV p-p noise at amplifier out(TP4). No idea how to calculate p-p noise at TP5 or TP9 with Rx input shorted and Tx disabled.

    Leave a comment:


  • green
    replied
    Originally posted by green View Post
    What is the peak to peak noise at TP5 and TP9 with R15 removed(using mini O scope)? Coil connected and coil shorted.
    Thinking there should be a way to calculate(estimate)peak to peak noise with Rx input shorted. Anyone have a formula or a place to look? All PI's have noise, would be nice to know if noise is higher than expected before trying to lower it.

    Leave a comment:


  • green
    replied
    Originally posted by h9361 View Post
    Is your means figure 8 in your HH article(Third Gain Stage) ? I have also noise in this point (Tp9).
    What is the peak to peak noise at TP5 and TP9 with R15 removed(using mini O scope)? Coil connected and coil shorted.

    Leave a comment:


  • h9361
    replied
    Originally posted by Carl-NC View Post
    If the coil is connected with no TX and you still have noise, then either you have a lot of EMI or there is a gain set way too high. The next step I would recommend is to make a Figure-8 noise canceling coil.
    Is your means figure 8 in your HH article(Third Gain Stage) ? I have also noise in this point (Tp9).

    Leave a comment:


  • Carl-NC
    replied
    Originally posted by h9361 View Post
    I removed R15 and i have stil noise!
    It seems that i have to build a new.
    If the coil is connected with no TX and you still have noise, then either you have a lot of EMI or there is a gain set way too high. The next step I would recommend is to make a Figure-8 noise canceling coil.

    Leave a comment:


  • h9361
    replied
    Originally posted by green View Post
    Thinking disabling Tx might be a good place to start trouble shooting. EMI and EF only. (schematic)https://www.geotech1.com/pages/metde...rhead/HHd1.pdf Could connect Q3 gate to -VB. Where does the jumper go to enable Tx?
    Thanks dear green.
    I am using rev. C2.

    Leave a comment:

Working...
X