Please help me bias this FET bridge

Discussion in 'The Projects Forum' started by Darkstar, Sep 5, 2011.

  1. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Just a thought, but your results would pretty much exactly as you've measured if you had misconnected the fets.

    Looking at the base of the fet, with the leads pointing towards you, and with the flattened side uppermost, the gate is the lead on the right (left to right would be: drain - source - gate).

    If you find you have got them wired up wrongly, it's possible you may have damaged the fets, so it would be best to replace them (correctly connected!).

    Of course, the above hypothesis may be completely wrong!
     
  2. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Hi,

    1. You said you want me to connect the gates to 0V, not ground, but as I've said, I no longer have that line available in this circuit. It's 25' away now. Even if I connect to the cable shields (which are not fully accessable to begin with), they would still be at earth ground, not strictly 0V reference from the power supply unless it was floating.
    On second thought, I could connect the 0V reference to the chassis and earth ground separately, as if the 0V rail was being used instead of -15V. Then connecting to the cable shields would be connecting to 0V and earth.

    2. Here are some thoughts: You say it's doubtful if the circuit would work at all without using -15V, but I've been using it this way from the start. I started with a Wheatstone bridge which had +15/0V and kept the same voltages when I made the bridge into the diff amp. The only difference I saw was a large decrease in sensitivity. Perhaps this was due to using 15/0V instead of +/-15V?

    3. You want the cable shield to supply the -15V to the sources. Then I'll have to disconnect the shield from the chassis ground, especially if I connect the chassis to the 0V reference so I can bring this out to the antenna as mentioned in #1 above. Then it'll be like using unshielded cable.

    4. I attached a diagram comparing the original circuit with the new design showing the main places that have been changed.

    5. I saw a diagram of a diff amp which appears to have a single supply, so the idea doesn't seem totally odd. You can see it if you download the file at:
    http://socrates.berkeley.edu/~phylabs/bsc/PDFFiles/bsc5.pdf
    It's Figure 5.5 on page 7.

    6. I reset your recommended variables and remeasured the circuit and I got numbers similar to the first time:

    Measurements taken with output adjusted to 0V and with -15V rail as reference.
    Rd1 = 820 Ω adj to 770 for zero
    Rd2 = 820 Ω adj to 979 Ω for zero
    Rs = 1.46K Ω
    Zero = 1K, full CCW (wiper at antenna, side)
    Vd1 = 21.54V
    Vd2 = 21.54V
    Vs = 20.71V
    Ids1 = 10.25 mA
    Ids2 = 3.92 mA
    Vds1 = 0.824V
    Vds2 = 0.824V
    Rd1+zero = 779 Ω
    Rd2+zero = 2022 Ω
    Rds1 = 80.4 Ω
    Rds2 = 210.2 Ω

    This time I got a clean graph without the strange looking parts now that the bad pot has been removed.

    7. The fets are installed correctly. I should be able to swap them with a different pair and check the first pair tomorrow.
     
  3. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Here is a look at what the circuit might look like using one cable.
     
  4. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Hi again,

    I was referring to your diagram when I suggested connecting to the cables' screen - it shows the screens connected to 0V. The external ground is shown to connect solely to a fet gate, so I'm confused.

    You're not going to get much of a diff amp if you use 0V as the negative supply - there'll be a reduced common-mode range, and the operating conditions will be very dependent on sample variation and temperature. Unless using the -15V supply presents too many difficulties, you're much better off with the -15V if you want reliable operation.

    When I suggested using the cable's screen as the -15V connection, I suppose I was implying that the chassis would then be at -15V. This shouldn't be a problem, since you have an ac ground connection via the decoupling capacitor. Alternatives would be to use the external ground as chassis ground, or to just use a separate -15V connection. I do think the fets' sources should be kept away from the cables though. But as you say, there may not be any oscillation, in which case you could keep the cables as they are.

    I think you may have forgotten to attach the new circuit diagram - it's certainly a good idea to keep things up to date.

    I'm not saying you can't use jfets in a diff-amp configuration with 0V as the negative supply, it's just that everything is inferior and more difficult to get working decently.

    As I mentioned, a quick test for the fets would be to short the drains to +15V, and check that the sources remain at the same voltage as before.
     
  5. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    This is the schematic comparison I meant to send...
     
  6. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    The cable screens are connected to the chassis originally, which is also 0V ref and earth grounded indoors. The second fet gate is earth grounded outdoors, but it is the same ground.
    I think I was pointing out what it would take to add the 0V ground to the circuit with the +/-15V supply. It's 4:35 am now, I haven't slept. Same as yesterday when I wrote my response. I was tired then, I'm tired now and I'm getting confused too.

    I don't have a problem with using +/-15V supply, except that I haven't yet been able to understand the main difference - the loss of the 0V ref connection to the diff amp outputs / in-amp inputs. I've tried using 10M resistors on each in-amp input connecting each to 0V, but I didn't like the operation. Using the diff amp as the 0V connection means having only 1 connection, not 2 that need to be balanced. The diff amp itself can be balanced easily unlike the in-amp which needs more circuitry to do it right and I'm barely understanding the circuit as it is.

    I don't understand why the decoupling cap makes it ok to connect the -15V rail to earth ground.
    I don't understand the next sentence either about the alternatives.

    I'll check the fets tomorrow (later today).

    Sorry to be slow on the uptake. Thanks for being patient.
     
  7. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I thought you might find these numbers more helpful.

    Here are the measurements with the output NOT zeroed:


    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, set to center (537 & 539 ohms)
    Vd1 = 20.12V
    Vd2 = 20.42V
    Vs = 19.59V
    Ids1 = 6.80mA
    Ids2 = 6.56mA
    Vds1 = 0.561V
    Vds2 = 0.857V
    Rd1+zero = 820+537=1.357KΩ
    Rd2+zero = 822+539=1.361KΩ
    Rds1 = 82.5Ω
    Rds2 = 130.6Ω
    pwr1 = 3.8mW
    pwr2 = 5.6mW
     
  8. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I set the DVM to measure the source voltage then shorted each fet drain to +15V in turn.
    Sources without drains shorted= 19.57V
    With the antenna fet shorted to +15V, the source voltage increased to 28.25V.
    With the other fet shorted to +15V, the source went up to 28.14V.
     
  9. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Ok, I goofed up and didn't know it. I swapped out the fets with different ones and found that I did have the existing fets turned around as you suspected.
    Here are the new measurements:
    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, set to center
    Vd1 = 18.5V rising - started about 16.5V
    Vd2 = 28.3V dropping
    Vs = 14.6V dropping
    Ids1 = 9.4mA dropping
    Ids2 = 0.53mA rising
    Vds1 = 2.2V rising
    Vds2 = 14.2V
    Rd1+zero = 1.357KΩ
    Rd2+zero = 1.361KΩ
    Rds1 = 234 Ω
    Rds2 = 26,792 Ω
    pwr1 = 20.7mW
    pwr2 =7.5mW

    The voltages were changing steadily at first, then they flattened out. I attached a graph of the first 10 minutes of a new test.
     
    Last edited: Sep 14, 2011
  10. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Ah, we've got it sorted at last! I thought it seemed the most likely source of the problems.

    But I'm worried about your health - you must get some decent sleep!

    Anyway, we're getting somewhere now. And it doesn't seem to be oscillating (though you'd need to get a 'scope with a reasonable bandwidth to prove that). That probably means we don't need to bother with the -15V supplied via the cable screen, and we can proceed with the existing arrangement (you'll be very pleased to hear!).

    What we should do now is get an idea of how well matched the two fets are; the easiest way is just to short the two gates together (the external ground appears to be pretty close to 0V for our purposes) and let the external ground define the gate voltages. A perfectly matched pair would result in equal drain voltages (assuming the zero pot is centred). Hopefully, it won't be necessary to adjust the zero pot too much in order to get the drain voltages equalised.

    If your test results are with the gates both connected to ground (or 0V), then it looks like the fets are NOT well matched - I'm hoping you had the antenna fet's gate just connected to the antenna, and that it was sitting at some potential above/below ground, thereby giving rise to the appreciable differential output voltage. Can you let me know if the inputs were shorted or not?

    Then we ought to try and measure gain by connecting a 1.5V battery across the gates.

    The reason for the drift would, I'm sure, be due to thermal effects. The fet carrying most of the current seems to be heating up owing to the fact that it's dissipating about 40mW. If this is going to be a problem, we can reduce this by reducing the operating current (I'll estimate some new resistor values).

    (The reason the cable screen as -15V supply would be OK is that the screen is there to shield the inners from stray ac fields, and the decoupling capacitor would achieve this because at ac it looks like a low impedance to ground. But we don't need to consider this any more, so we can forget all about it.)

    Once we've got all the above sorted, then we can think a bit more about protection and filtering...

    I hope you got a decent night's sleep last night!
     
  11. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    At least the graph looks good to you, it looks very wrong to me. I thought for sure one of the fets had gotten fried or something but they seem ok. I've run the fets at high power levels before and never saw that steep, steady change that just seemed to hit a ceiling and suddenly go horizontal. I let the unit run for several hours. Normally I should see the output varying a bit in that time depending on the time of day, but I still had much the same output with only a few bumps on the line. It looks very much like the backwards fets were working much better than the correct fets are!

    The antenna fet is connected only to the antenna. I checked out the circuit like this first to see what the results looked like compared to what I've seen in the past. Next I'll connect the gates and make new measurements and a graph. Then I'll connect a battery and do the same.
     
  12. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    It may be that your results in the past have given you a false impression of what's actually happening. I've looked at the graph again, and it does remind me of a capacitor being charged slowly at a constant current - so there seem to me to be 2 possibilities: the thermal drift and the slowly charging capacitor. If it's the latter, I don't think it should have any adverse effect when observing over 24 hours. If it's the former, we may consider putting the zero pot in the source circuit instead (I'm guessing that we've more than enough gain, so taking a hit here wouldn't be a problem).

    It'll be very interesting to see the results for the gates strapped together and for an attempt to demonstrate the gain. It may be that 1.5V saturates the diff amp - be sure to try both polarities of battery connection (because we may have a large offset due to fet mismatch, and this would help in one direction and hinder in the other).

    It would also be interesting to see how much displacement of the zero pot from centre is required to equalise the outputs when the gates are strapped.
     
  13. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    If past results were wrong, it's some coincidence that the sinusoidal graphs look like other people's published graphs of atmospheric potential, but stranger things have happened.

    I've seen charge/discharge shaped curves in the data, and they have the right curve. This is ruler straight, then it hits the ceiling. I can't explain this. This is not a normal result. The single fet wheatstone bridge circuit did not require an equilibration period. Waiting a long time for true results makes it hard to get results quickly when needed.

    This first result is with both gates grounded (Test 39). Graphs are attached.

    Measured variables taken with output NOT adjusted to 0V.
    (Measurements are taken with -15V rail as reference)
    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, wiper centered
    Vd1 = 21.55V climbing
    Vd2 = 23.75V climbing
    Vs = 13.69V
    Ids1 = 6.25 mA dropping
    Ids2 = 5.04 mA dropping
    Vds1 = 8.73V
    Vds2 = 10.43V
    Rd1+zero = 1.357K Ω
    Rd2+zero = 1.361K Ω
    Rds1 = 1396.8 Ω calc; 134.3 Ω measured, not operating
    Rds2 = 2069.4 Ω calc; 146.1 Ω measured, not operating
    pwr1 = 54.6mW
    pwr2 = 52.6mW

    The second is Test 40 with +0.898 V on the antenna relative to ground (-). I thought it'd be better to go with less than 1.5 V. I lucked out making the antenna pos because the positive output offset the original output of -1.783 V. I don't know what caused the irregularities in the graph. At right I was able to bring the output very close to zero with the 1K zero pot and I have some to spare.

    Measured with output NOT adjusted to 0V........Measured w output zeroed
    (Measurements taken w -15V rail as reference)
    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, wiper centered
    Vd1 = 23.88V climbing....................................22.85V
    Vd2 = 21.44V climbing....................................22.98V
    Vs = 14.43V climbing......................................14.87V
    Ids1 = 4.24 mA dropping.................................0.5 microamps
    Ids2 = 6.43 mA dropping.................................6.6 mA dropping, steadied at 5.09 mA
    Vds1 = 9.80V dropping....................................12.53V
    Vds2 = 7.04V dropping....................................0.0001V
    Rd1+zero = 1.357K Ω
    Rd2+zero = 1.361K Ω
    Rds1 = 2311.3 Ω calc, 175.9 Ω measured .........25.06 Megohms calc
    Rds2 = 1094.9 Ω calc, 133.5 Ω measured .........0.0196 ohms calc
    pwr1 = 41.6mW............................................6.3 microwatts
    pwr2 = 45.3mW............................................0.51 microwatts

    To zero the output, the pot was moved from center (537 ohms for antenna side, and 539 ohms on ground side) to 75% of the way CW (toward ground side = 801ohms on antenna side & 275 ohms on ground side.)

    I think I found another flakey pot.
     
  14. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I had an idea for connecting the 0V ref ground to the dual supply circuit. Just connect to the center of a voltage divider between the pos and neg rails. It would waste some current, but it would work.

    When bias return resistors are used on the in-amp inputs alone, they can be 10M and even 100M. However, I already found that when using the bridge as the ground connection, the amp does not like excessive resistance in its ground connection. I compromised with the 10K resistors. I hope they are not too high, yet I'm trying not to waste too much current, but I will if I have to.

    I think this is a good compromise so I can use this connection and you can use the dual supply.
     
  15. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    I'm not saying that the previous stuff was wrong in general terms, but I think you were getting extremely low sensitivity. So you were able to see a sinusoidal trend but not necessarily at a useful or consistent level.

    I'm pretty sure, looking at your latest results, that the initial slope, following power-up is due to the devices warming up. Is it unacceptable to have to wait for a minute or so after switching on? There are all sorts of steps that can be taken to prevent this, including the use of monolithic dual fets. If you were to use these, you could really minimise differential thermal drifts, but obviously the two fets would be co-located; I don't know if this would be unacceptable, but I do know that co-location would make the diode input "protection" easier.

    Anyway, according to the results, it seems the fet mismatch is causing about 2V offset at the differential output. As you've noticed, this could turned to your advantage if you were expecting a given polarity of input signal. Is this the case (e.g. is the cloud always positive with respect to ground?).

    As for gain, you've got a bit more than I expected - it looks like it's about 6.5. I think it would be a reasonable idea to consider switching the zero pot to the source circuitry - I suspect we don't want that much gain anyway. I'll have to estimate some new resistor values for this.

    It does look like you've got a flaky pot - there are some strange anomalies in the figures.

    I think I've begun to understand what you're trying to accomplish with your two 10k resistors. Are you saying that there's no 0V connection to your power supplies? Because it's very confusing (i.e. wrong) if you have "0V" actually marked on the circuit diagram. I think it's pretty undesirable to supply +/- 15V without the 0V connection, and I wouldn't advise it (if only for safety reasons, because, presumably you've got 0V connected to ground at the power supply end).
     
  16. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    My best sensitivity was with the wheatstone bridge using 1 fet. I got 200 mV p-p. 170 mV was more common. The amplitude changed day to day depending on the weather, time of year, and no doubt other unknown things (sunspots?) The 2 fet circuit has been significantly less, a few dozen mV at most.

    I don't mind a few minutes warm up. I want to be able to turn this on if I hear thunder so I can monitor if anything is getting too close (I hope I can do it with this.)

    By "co-location" do you mean located in the same place? The 2 fets are currently outside, under the antenna, inside a sealed jar, plugged into opposite sides of an 8 pin IC socket, just mm away from each other. I put them together so they'd be close to the antenna like the 1 fet used with the wheatstone bridge and so they would be the same temperature, summer and winter. They used to be indoors but I thought being close to the antenna would be better. Maybe it was just the wheatstone bridge that was better?

    Typically, the bottom of a thundercloud is negative and the top is positive. The earth underneath would be positive. Lightning brings the negative charge to the ground. As the negative charge is reduced, the positive charge at the top becomes more dominant. About 10% of lightning discharges are from the positive cloudtops to the ground. They are about 10x more powerful than the negative discharges from the cloud bottoms.

    Why wouldn't we want as much gain as we can get? True, the amp can boost the signal, but it seems to me that with more gain I can presumeably detect smaller variations in potential so when amplified, I'm amplifying a varying signal. But if gain (sensitivity?) is low, I wouldn't see the small features and I'd end up with more of a flat baseline. Amplifying this isn't going to add the info I didn't detect. Having the antenna on my roof would give me more sensitivity, but that's not possible.

    I bought new 1K pots. 18 mile round trip to the only store in the area with simple 1K pots.

    I attached a full schematic here so you can see the power supply. It's a standard dual supply circuit with a center 0V reference. The reference is not connected to the chassis/earth ground, but it can be connected easily. The schematic shows separate earth grounds for power supply and other units, but I would only use one so there wouldn't be ground loops. I can make the connection in different places, wherever convenient. So, the power supply has a 0V line, but there has been no place to connect it to the fet circuit when +/- 15V is used for power. I used to use +15/0V as you know. So, the 0V on the schematic is accurate. I know the 0V line should be grounded as it is just a virtual ground and not a real ground.
     
    Last edited: Sep 16, 2011
  17. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    From your first test following the swapping of gate and drain connections, I think I'm seeing that the antenna was at around -1V relative to ground. I don't know how much sinusoidal variation we can expect in a 24 hour period, but with a gain of 6.5, the variation would have to be about 30mV p-p for a 200mV response at the output. I don't know if this is what you'd be expecting (we don't know the gain of the single transistor circuit), but 30mV is pretty small, and we'd want to be as free from drift as possible. I think a monolithic dual jfet would be a good idea for the future.

    Am I right in thinking that the 2-fet circuit has always had the fets incorrectly connected? If so, then we can't draw any conclusions from the results.

    One thing that srikes me is that the single-transistor circuit will have been quite sensitive to changes in ambient temperature. Could this be the cause of the output variation over 24 hours? It sounds distinctly possible to me, so I think you'd be much better off with the balanced (i.e. differential) circuit (as I say, I think the monolithic dual-fet would be the best solution, long-term).

    Regarding gain, the more we have the more we run the risk of overloading the amplifier. In order to cope with differential inputs of more than a couple of volts, we're going to have to reduce the gain, and put the zero pot in the source circuit (luckily, both of these go hand in hand). So, as I said, I'll try to estimate some new resistor values.

    Yes, I think it would be about a 30 mile round-trip for me to acquire electronic components! Just as well I don't indulge any more.

    I don't think you can go far wrong by having chassis and cable sceens connected to the power supply's 0V. I don't think the two 10k resistors are giving you anything.
     
  18. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    I think for the low gain version, you would make the Rd values 1k2 or 1k3, and connect them straight up to +15V; the 1k zero pot would have its wiper connected to the top of Rs, and each of its ends would be connected to a fet's source. Rs would now be a 1k resistor.

    Actually, this will have a fairly drastic effect on the gain (a 500 ohm pot would have been perhaps a better value), but I think the overall results would be worthwhile (stability and dynamic range capabilities). I'd certainly recommend giving it a try, assuming it's not too difficult to make the changes.

    PS: if you do have a 500 ohm pot, you could use that, with Rs now 1.25k.

    PPS: this diff-amp application is more about providing predictable power gain, rather than voltage gain. You've got the admirable instrumentation amp for the voltage gain bit!
     
    Last edited: Sep 16, 2011
  19. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Of course, you may like to leave the drain circuitry as it is (i.e. with 820 ohm resistors and the 1k pot, set to centre). Probably easier to do this.
     
  20. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I don't know the absolute voltage change during a 24 hr period. The best I could do is look up papers online and see what others have reported. I don't know if I could accurately calibrate my system.

    I'm afraid the incorrectly connected fets have been in place for the past 2 months. I was sure they had been installed correctly.

    The 1 fet design did respond to a combination of potential and temperature changes, but I felt sure the potential part was there. I moved the fet indoors to avoid the temperature effects, then soon after added the second fet. I agree, the differential circuit is better, or supposed to be anyway.

    I'll take your advice about the gain limits though I don't fully understand.

    I attached a group of 3 schematics I've shown previously but here I highlighted the path between the 0V ref line and the in-amp inputs that I keep talking about. The in-amp needs this connection. The schematic with the dual supply and no voltage divider has no connection whereas the others do.
    ------------
    If I chose a good in-amp, luck had a big part in it. I went for large amplification capability, and I got the amp for free. Voltage gain I understand but I don't think I've heard of power gain.
    -----------
    I'll leave the circuit as-is for now rather than make too many changes all at once so we can evaluate the effect of the fet proper connection.
     
Loading...