Please help me bias this FET bridge

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

  1. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    House gnd = earth gnd = external gnd = chassis. This is what is used by a plug with a ground prong. "0V" = power supply 0V reference line (+15/0/-15). I made the power supply with the circuit isolated to prevent the possibility of AC induced in the chassis by the transformers getting into the power supply output. However, the chassis can be connected to 0V as necessary. I looked this up once online and found that it didn't really matter if they were connected because it usually made little difference in the operation. I believed it because I tried my circuit both ways and saw little difference (at the time I was only seeing something like a 3V difference between 0V and the chassis. I haven't yet traced down the source of the 14V I'm seeing now.)

    I understand how the fets work and the diff amp. It's ICs like op amps and the in-amp which come premade with complicated circuitry inside and seemingly extra connections than I don't know where to connect. I'll admit though that I don't know all the details. For example, I noticed that when a fet was used in an audio amplifier, it was run in the linear region and I thought this was to avoid distortion. I thought there was more amplification if it was pushed harder, and because I'm not worried about distortion, I aimed for the saturation region.

    I didn't know any of the stuff you described up to and including my Rds values. I was simply using Ohm's law on the fet.

    How do you decide the tail current to use? Is this tail current = standing current? I've seen examples of calculations for audio amps and they would go for about 5 mA use by the fet. I've also been told that fets get quieter as Ids goes up and I have had good results last year in the 10-12mA range. I was also told to keep Vds "low, but not too low." I've found a few formulas online for Vd and Rd but they give very different values and I never knew which to use. At least one was for an audio amp, which this isn't, so I tended to discount that one.

    It didn't seem like the gate and drain connections were reversed in the simulation. It would draw the fet symbol so you could see which was which and it also reacted by stopping immediately if the gate got reverse biased. I do find it entirely plausable that the math could have been done with the values swapped, thus resulting in an opposite sign on the results. I can't change this. It's an online site and I can only draw a circuit. If nothing is grossly wrong, it'll operate and do all the calculations for me (which I like, because I don't know how to do many of them). You can see this at: http://www.falstad.com/circuit/e-index.html
    It's possible to make changes to the circuits shown, or blank the screen and draw a circuit from scratch.
    As for my own preferences, being a chemist, I learned that current is electrons that leave atoms and it runs from neg to pos. In electronics class they said the opposite, but I know for a fact it's the electrons that move, not the positively charged wire ions they leave behind. It would be nice if everyone used the same standard nomenclature.

    The noise and spikes are not due to any of my present pots. I have tapped them and get no reaction. I strongly feel they are some kind of artifact of either the meter or the software when faced with a data point they can't measure for some reason. This is a Radio Shack meter that came with data collection and graphing software called MeterView. It's got a few quirks, but it's all I have for visualizing my data. I've looked for a forum for owners of this software so I could ask about the spikes, but there isn't any. Radio Shack does not provide any support either.

    I would get a cheap scope if I could find one. For the little bit I would use it, spending several hundred dollars for even a low end scope would be mostly a waste of money and just add one more thing to my menagerie of stuff I rarely use. I have many hobbies and I shift from one to another depending on what's interesting at the time so all the stuff I accumulate for each just sits in storage until I come back to it.

    So, I keep trying to go as far as I can with this circuit. Which, BTW, is the first I ever designed (with lots of help from anything similar I could find online as a guide.)

    What do you think about connecting the antenna and ground to the in-amp inputs? I don't know how it's sensitivity compares to the diff amp alone, and there would be problems with bias return resistors and noise and oscillation in the long antenna line that won't have a chance to be eliminated by a prior circuit, but it may be an option.
     
  2. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    You mustn't have a floating power supply. For the fets to operate properly, the gates must be at a clearly defined potential relative to +/-15V and 0V. - and preferably as close as possible to 0V. So, do make a connection - otherwise your results will be all over the place!

    I think a feature of op-amps is that they are like black boxes - you really don't need to know anything about them internally (though for someone who understands the circuitry, it's actually quite helpful to know what's inside). However, you do need to operate them sensibly, so you need to pay attention to at least some of the parameters associated with the device you're using. My biggest objection to them in this application is related to slew-rate limiting - they're not very tolerant of large, fast, transients.

    As for the fet operation, if we could start calling the two operating regions the "resistive" and "constant-current" regions (because this is less open to misinterpretation), then there's no advantage in operating in the resistive area, unless you're specifically wanting a voltage-controlled resistor. All amplification applications will make every effort to stay in the constant-current region, and that means maintaining a decent Vds. That means some knowledge of what potentials the gates will experience. It also helps define the drain load resistors - they need to be low enough to keep Vds high enough, though not so low as to lose too much gain, and to raise the common-mode output mode voltage too close to the positive supply.

    As I say, if you're operating in the constant-current region (shown as the "saturation" region in the graph I linked to), you can't use Ohm's law to give a resistance value because that part of the curve isn't a straight line through the origin. You could give the fet an incremental resistance value at that Vds and Vgs, which would be 1/slope, and this would be maybe in the tens of kohms.

    In terms of choosing standing current, you'd normally choose a reasonably high value (but not so high as to threaten maximum current/power specs), in order to get a reasonably linear relationship between Vgs and Ids. You will experience some degree of self-heating at these higher current levels, and this will cause a degree of thermal drift until warmed up.

    As for current flow and electron flow, I wouldn't worry too much - it's just a convention really. And for a cheap 'scope, I was thinking seriously sub-$100 - probably too much to ask for I suppose (but you know what to do if one comes along!).

    As for dispensing with the fets and using just the instrumantation amp, I suspect everything would be straightforward - except for two things: the problems with fast transients I mentioned and the fact that the antenna would "see" the capacitance of the screened cable, which would probably rule this out completely. The better solution would be to buffer the signal from the screened cale using a picoamp-input op-amp as a unity gain buffer (as the article you gave me) and bootstrap the cable screen. But the potential for slew-rate problems with fast transients would remain.

    Once you've got everything working OK (get that 0V-ground connection!), we should then be considering constraining the inputs using 1N4148 diodes, and filtering the output, using 1-10 Mohm resistors to connect the diff-amp outputs to the in-amp's input filter capacitors.
     
  3. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    The tiny 3 component circuit I originally started with had no special gate circuitry and the gate just floated. It was very sensitive, so when I began adding to the circuit I just let the gate float as it had originally. I still get best results with a floating gate for the antenna.

    Back when I was taking electronics class they were still teaching about tubes. They switched to solid state the next year. I like knowing what each part of a circuit does and then ICs came along, doing things I was never taught in ways I was never taught. That's why I'm uncomfortable with them.

    Ok, no calculating Rds. I only did it out of curiosity anyway.

    It looks like more things are going wrong with my power supply, yet I can't locate a bad component. It looks like I'm going to have to rebuild it a bit at a time till I find the bad component.
    Fortunately I have a second power supply which I'm using now.

    I connected 2 back to back 1N4148 diodes today. They are shielded from light also so they don't conduct excessively during the day.

    These are the current circuit stats using my second 1.25-17V power supply.
    1N4148 diodes are installed between the gates as protection for the antenna fet.
    The power supply 0V ref line is grounded to power supply chassis.

    29.93V total
    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, wiper centered (572, 572), output not zeroed.
    Vd1 = 17.70V <
    Vd2 = 26.88V <
    Vs = 15.20 V <
    Ids1 = 8.22 mA <
    Ids2 = 2.21 mA <
    Vds1 = 2.550V <
    Vds2 = 11.65V <
    Rd1+zero = 1.392K Ω
    Rd2+zero = 1.394K Ω
    pwr1 = 21.0 mW
    pwr2 = 25.7 mW
    Output = -9.15V
    I did not have to wait for anything eo equilibrate.
     
  4. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    There don't appear to be any problems with my second power supply. (Originally, there weren't any with the first either.) I measure the same proper voltage between the pos and neg outputs and the 0V ref as I do with respect to the chassis ground. Measuring between the 0V ref and chassis ground I get -2.007V. Chassis is positive.

    I'm running a test now on my circuit. I'll have results Wed. afternoon, but so far it looks like there is only noise and little to no signal. If I don't get any signal, I'll disconnect the diodes and see how it differs.
     
  5. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    If you have floating power supplies, then the 0V line potential with respect to ground is undefined. That means it can take any arbitrary value, according to minute leakage currents, and it doesn't mean you have a fault if you find yourself with large voltage offsets. The fet circuit can't operate with its power supplies at any arbitrary value with respect to ground, it needs to be in a fairly tight range - so you must connect 0V to ground.

    If you like tubes, then jfets are just the thing for you: of all the semiconductor devices they're the closest thing you can get to tubes. And you don't have to worry about high voltages or heaters either!

    On the subject of protection diodes, I'm not so sure you have them connected properly: they're not like protection zeners, you can't connect them back-to-back in series with each other - each has to be connected across the inputs. I'm guessing you've misconnected them because you have a substantial output voltage, implying an input voltage greater than a diode drop (though the diff amp's offset voltage may be making a significant contribution to the output voltage).

    You'll probably need to cater for more than a +/- 600mV differential input voltage, but you can get round this by putting several diodes in series.
     
  6. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    The first graph here is the diff amp output with the new power supply and the diode protection.
    These measurements taken during the noisy part at right to see what has changed from the previous measurements:

    29.74V total, under load
    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, wiper centered (572, 572), output not zeroed.
    Vd1 = 17.35V <
    Vd2 = 27.10V <
    Vs = 15.10 V <
    Ids1 = 8.50 mA <
    Ids2 = 1.85 mA <
    Vds1 = 2.340V < fluctuating a lot
    Vds2 = 11.95V <
    Rd1+zero = 1.392K Ω
    Rd2+zero = 1.394K Ω
    pwr1 = 21.0 mW
    pwr2 = 25.7 mW
    Output = about -9.80V

    The second graph is the output with new power supply and the protection diodes disconnected. The blips are from me turning the room light on and off.

    29.74V total, under load
    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, wiper centered (572, 572), output not zeroed.
    Vd1 = 19.90V <
    Vd2 = 24.02V <
    Vs = 15.10 V <
    Ids1 = 6.80 mA < down from ~9.2
    Ids2 = 4.06 mA < up from ~0.7
    Vds1 = 3.95V <
    Vds2 = 8.33V <
    Rd1+zero = 1.392K Ω
    Rd2+zero = 1.394K Ω
    pwr1 = 26.9 mW
    pwr2 = 33.8 mW
    Output = about > -2V and climbing

    These took 3-5 min each to equilibrate.
    ----------------------------------------------------

    Thanks once again for explaining how things operate. I do have 0V connected to earth ground now. I'll check the power supply again.

    Right again, I had the diodes connected like I had done the Zeners. I'll re-connect them. I will need several in series unless I zero out the diff amp output beforehand.

    Look at the schematic here and tell me if I'm connecting the diodes correctly. Remember, the antenna/gate can go either positive or negative by itself. Also if someone touches it.
    Until I hear back from you on the diodes, I'm going to connect a 10M ohm bleeder resistor between the antenna and ground to see how well that operates now.
     
  7. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I disconnected the diodes and connected a 10M ohm bleeder resistor between the antenna gate and grounded gate so that the antenna wouldn't float. I got these measurements:

    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, wiper centered (575, 577), output not zeroed.
    Vd1 = 22.21V
    Vd2 = 22.08V
    Vs = 16.62 V
    Ids1 = 5.67 mA
    Ids2 = 5.69 mA
    Vds1 = 5.62V
    Vds2 = 5.51V
    Rd1+zero = 1.395K Ω
    Rd2+zero = 1.399K Ω
    pwr1 = 31.9 mW
    pwr2 = 31.4 mW
    Output = about 100 mV

    There was no waiting for any of these numbers to equilibrate.
    The diff amp output was very different. It did not start at -14V and very slowly climb to about -2V, nor is the circuit very sensitive to my movements. The output started off at about 100 mV.
    It looks like there is some oscillation going on.
     
  8. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    This is the graph I got with the bleeder resistor on the antenna. I do not see evidence of the usual signal. The jumps were from turning a TV on and off in another room.
     
  9. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I tested 4 FETs and they all checked out OK, then I put them into the circuit at the antenna, and when I measure the resistance between the source or drain cable wires and earth ground (from indoors), I keep getting about 45K Ω even though the FETs were >40M Ω when measured out of the circuit.
    There is no connection from source or drain cable wires to ground when the FETs are removed from the circuit, so it has to be the FETs that are making the connection. I guess I'm burning them out when I handle them despite my precautions.
     
  10. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    When you're measuring 45k to ground, are you saying that there's nothing connected to the drain or source, and that the gate is connected to the antenna, with no 10M resistor or diodes? Don't forget that the jfet looks like a diode between gate and drain or gate and source, so if you make the gate sufficiently positive with respect to drain/source, you'll get conduction through the gate.

    I think you've proved that the 10M resistor is bleeding charge far too quickly from your antenna. Hopefully the diodes will be much better (you've now got them connected correctly, by the way). If you were to zero the fets with inputs shorted, and then remove the short (leaving just the antenna connected), I would expect the differential output to now be limited to about 7V. It'll be interesting to see what happens in practice.
     
  11. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I just rechecked the drain/source line to ground connections and I got different results. I guess this is good because I don't see the leakage any more. Previously, I did try reversing the polarity of the ohmmeter connections and I got the -same- leakage both ways. Now, when ground is pos and source/drain is neg, I get about 2.6M ohms through the gnd fet and an open (>40M) connection through the antenna fet (this has the 10M bleeder still between the gates.)
    With ground being neg and source/drain being pos, I get opens for both. It has been several hours since I last worked on the circuit. The only thing I can think of to account for the diffrerent results is that there may have been charges on components after handling which subsequently leaked away. It's getting cooler here and the relative humidity is down to 35%.
    I'll do some quick tests, then connect a hopefully appropriate number of diodes into the correct series/parallel arrangement as diagrammed and test that arrangement.

    Question:
    Do you have an explanation for the jumps I sometimes see such as when the TV is turned on/off, but especially similar jumps which I have no explanation for? It seems like something in the circuit is changing on its own from the initial settings to some other settings which it finds itself and may be at least as stable as the first settings so that the change is more permanent than temporary.

    Question:
    Do you have an explanation for why I detect no AC anywhere in my power supply output (still using the second, new supply), and despite with no ac connections to the diff amp circuit, I regularly measure several volts of 60 Hz AC in the output? For example now, I have 0.660VDC in the output but 2.995 VAC also. I find it hard to believe that this much could be picked up
    somehow by the shielded circuit and amplified?
     
  12. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    The first graph here is with the 10M bleeder resistor still in place. It started out fairly quiet without spikes, then something kicked it into a more noisy mode. Later, turning on a TV in the next room made it get very noisy. Turning the TV off did not make the noise go away.

    Right after this, I removed the resistor and installed 1N4148 diodes wired correctly this time:

    ......|---->|---->|-----|
    ant.|----|<----|<-----|..gnd


    The second graph shows the results with the diodes.
    Here are circuit stats:

    Rd1 = 820 Ω
    Rd2 = 822 Ω
    Rs = 1.46K Ω
    Zero = 1K, wiper not centered (637 antenna side, 506), output IS zeroed.
    Vd1 = 22.31V
    Vd2 = 22.24V
    Vs = 16.32V
    Ids1 = 5.28mA
    Ids2 = 5.87 mA
    Vds1 = 5.99V
    Vds2 = 5.91V
    Rd1+zero = 1.457K Ω
    Rd2+zero = 1.328K Ω
    pwr1 = 31.6 mW
    pwr2 = 34.7 mW
    Output = about 75 mV

     
    Last edited: Sep 22, 2011
  13. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Are you saying that the DVM indicates several volts of ac when it is switched to "AC Volts" and less than a volt dc when it's switched to "DC Volts"? If so, that puts a different complexion on things - for instance, is the circuit oscillating, or is it really just 60Hz that it's picking up? I have to say, I feel lost without knowledge of the info a 'scope could provide.

    As for the "jumps", I think we need to get the above sorted before we can draw any conclusions.
     
  14. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    That is correct, at this time I got <1VDC and several VAC in the diff amp output. Sometimes I can't get any DC reading because it jumps around so much. Other times, I think when there is more DC, I will get a stable DC reading but I will still get an AC reading also. At other times I get no AC readings. 60Hz is the only frequency measurement I get these days. I'll try and find a record of the conditions used the last time I had no AC measured.
     
  15. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    During the measurements questioned above, the power supply 0V ref was connected to the earth grounded chassis as it's supposed to be.

    Here is an excerpt from some of my notes when I was trying to fight the noise in my outputs:

    "In-amp output: AC V = 3.1 mV NO 60Hz interference

    I verified that this result is due to the two 4700 µF caps connected pos to pos across the in-amp output.
    When caps are removed, I get a difference of ~1.25 V AC at 60 Hz.
    I still don't know where the 60 Hz is coming from, but at least I can now eliminate it from the output.
    Putting 2200 µF caps across the +15 and -15 V lines to earth gnd does not remove the 60 Hz or reduce the 1.2 V AC potential difference between amp out and ref.
    So far, it hasn't mattered much where I put the two 60 Hz LC tank traps. Except that when the trap is put on the detector ground, the output has less noise when the chassis gnd and 0V ref gnd lines are connected at the power supply.
    Having the two 4700 µF caps in series (positive-to-positive) across the in-amp output does the best job of eliminating AC on the output line."

    See the attached for what that graph looked like.
     
  16. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    The best I can do for measuring frequency is using my DVM which can measure from 10Hz to 4MHz with sensitivities from 50mV to 250mV RMS depending on the scale used.

    I have gotten frequency readings which jump around all over the place in the past, but lately it's always 60Hz.

    The big limiting factor with the graphing is that the software has a top sampling speed of 1 sample per second.
     
  17. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    All the following graphs have diodes between the antenna and ground.
    Graph 1 below shows the results without any extra filtering.
    Graph 2 shows the results of sending the diff amp output through a 60Hz trap at the start. At 1400 sec an FM radio trap was added. (No effect.)
    Graph 3 is the same as graph 2 but now also has two 4700 uF caps across the diff amp outputs. They're wired back to back to form a non-polarized electrolytic of 2350 uF. (Caps help with some noise.)
    Graph 4 shows the results of measuring frequency.
    I also tried putting the 60Hz trap on the ground line but that had no effect.

    Note that the x and y scales are not the same.
     
  18. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    You can get rid of 60Hz at the output easily enough - I'd just connect the outputs to the in-amp filter capacitors via a couple of Mohms, and this'll get rid of any interfering signal. The trouble is, any non-linearity in the diff amp will result in a dc offset due to incoming ac, and this changes according to the amplitude of the 60Hz. In fact, this may account for the abrupt changes in dc level that you're seeing at the output.

    We can tackle this in 2 ways: try to minimise 60Hz pickup at source; make the diff amp (let's call it "input amp", because we might change it) as linear as possible.

    Firstly, the antenna: is this similar to the plate used in the article you linked to? Or is it more like a normal antenna? I think, for detecting charge, you need to be as close as possible to being like the plate of a capacitor, positioned above and parallel to ground. Also, the connection to the input amp should be as short as possible.

    I know you don't want to mess about with op-amps, so I'll try to think of the best possible configuration for the jfet circuit (bearing in mind we don't really have the ability to check whether or not it's oscillating). I'll have to get back to you. Meanwhile, keep plugging away!
     
  19. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    I don't understand when you say "connect the outputs to the in-amp filter capacitors via a couple of Mohms". The in-amp has 7 decoupling caps on its various inputs.

    My antenna is similar to the plate in the paper except that it is a large inverted steel bowl about 17'' in diameter. Connections are inside the downturned bowl. It is parallel to the ground, but I'm limited to placing it on my balcony. It is around 20' above the ground but about 3' from the building. I can't change this.

    If I put any more of the input amp circuit outside I won't be able to work on it much, not to mention that I'd need a bigger weatherproof housing. There would still be long wires coming indoors. I compromised by putting the fets outside so their gate connections would be short and only their power lines would be long.

    I don't know why I haven't had better luck with eliminating high freq oscillations. I've tried low pass filters (like 1 Hz and even less) and signal averaging caps as high as 22,000 uF but something always remains.
     
    Last edited: Sep 24, 2011
  20. oldtech33709

    New Member

    Sep 24, 2011
    26
    0
    Try decoupling Rc with a few Picofarads, This should greatly improve DC stability.
     
Loading...