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
    The run with the first list of variables did not work either. It looked good except that it too kept climbing when it should have been dropping toward a low point.
    [​IMG]
     
  2. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    I wouldn't trust the simulation - it seems to be getting its drains and sources mixed up (they're probably interchangeable in real life, but the simulation really shouldn't be so confused about it).

    If it's easy to do, I'd try (real-life, not simulated) Rc = 1k5, Rd = 820 and a 1k zero pot (connect its wiper to +15V and each end to an Rd - I hope it's a multi-turn pot).

    I use the word "tail" because over here the differential pair was historically called the "long-tailed pair". The tail is the path from the negative supply up to the two transistors, so the tail current is the current through Rc. Anyway, with these resistor values, you should end up with about 11mA tail current and hopefully 5.5mA in each transistor - in which case the outputs should be sitting at about 8V. I wouldn't expect too much voltage gain from jfets, though.

    When you say "test" do you mean a simulation run?
     
  3. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    The second list of variables didn't work either. Here is how the first (top) and second (bottom) lists started...
    [​IMG]

    I've seen this kind of behavior in the past and I haven't figured out why it happens. I did nothing to cause the jumps shown on the graphs. Something changes in the circuit by itself if conditions are not right. This is the same circuit that gave the other graphs, just with different biasing.
     
    Last edited: Sep 9, 2011
  4. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Hello Z,

    Well the last 2 tests (real life trials with the posted resistor settings) did not work. I just posted the odd jumps in the output data.

    I'll try your latest suggestions now.

    My tests are real life. If a simulation runs I make a record of the variables so I can try them in a real life test at a later time. I get no data from the simulations other than knowing some calculated variables. I set my pots to the resistance in the simulation and trust that the voltages and currents in the circuit are what the simulation has calculated. My graphs are actual plotted data points taken at 1 second intervals.

    The source and drain are interchangeable in real life for the fets I'm using.
    Thanks for the clarification of tail current. I suspected that was what you meant but I wanted to be sure.

    I'll let you know what happens after I collect a few hours of data.

    BTW, I know the fets don't give much amplification which is why I have an instrumentation amp I can plug into if needed. It's an INA121 with variable gain as high as 10,000.
     
    Last edited: Sep 9, 2011
  5. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Z,
    We appear to be calculating things differently and drawing our circuits differently. I just entered your Rd and Rc values into the sim and I did not come close to getting the voltage and current you got. Here is a screenshot of the circuit with all the variables listed. Fet stats are the group in the lower right corner.
    [​IMG]

    You mentioned that each end of the zero pot is connected to one of the Rds. In my circuit, the Rds are above the fets, on the drain side, and the zero pot connects to each fet source. This is exactly how I've found differential amps drawn online.

    I can get the current and drain voltage you mention, but with different Rd and Rc values. I'll post the list of stats when I get all the numbers from a simulation. Then I'll do a real life test.
     
  6. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Here is a screenshot of the sim circuit with all the variables, just as you suggested.

    [​IMG]
     
  7. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Somehow we've got our wires crossed here. What I was suggesting was as follows:

    Starting from the bottom: -15V supply connects to a 1.5k resistor, the other side of which connects to the two fets' sources. The fet drains each connect to an 820 ohm resistor, the other sides of these resistors connect to the 2 non-wiper terminals of the 1k pot. The wiper connects to the +15V supply. (Both supply lines should have local decoupling capacitors to ground.)

    I wouldn't bother with a simulation, it's not to be trusted. The best thing would, I think, be just to try it out on the bench, and to go round checking the voltages at the fets' source and drains. Hopefully these will be sitting at something near the expected values. Next, it would be useful to measure the gain, and you could do this with a dc voltage source (such as a 1.5V battery).

    Assuming everything's satisfactory, then it could be connected up to your antenna and ground and hopefully you would get some sensible results. Ultimately, I would add the diode input "protection" circuitry and a big C at the output!
     
  8. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Z,

    Ok, I have your circuit drawn out now. I'll get to work on it later today.

    The 1N4148 signal diodes will have to go outside at the antenna.

    I can use a large cap on the output as long as I'm not plugging the output into the amplifier. A 4700 uF cap will smooth out some of the small fuzz on the baseline, I've used them before. I'll be relying on the diodes to clip the large spikes from lightning.

    On an earlier version of the circuit I had jacks that allowed me to easily measure current and voltage, but they are not on this version, so it's harder to take measurements during operation. I'll be rewiring part of the circuit now so hopefully I can take measurements off a breadboard.

    Thanks for all your help, I really appreciate it!
     
  9. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    You're very welcome - but I'll only be happy when we've got you up and running!

    Why couldn't you have a capacitor on the output if you're using an amplifier?
     
  10. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Z,

    The datasheet for the INA121 shows that small caps (0.1 to 1 uF) on the inputs can be used for filtering, and they do work, but every time I've tried electrolytics across the inputs (as low as 10 uF) they always upset the balance of the inputs so that the output is driven to one of the limits, about +/-14 V. So it's either the high capacitance or the electrolytic cap itself that it doesn't like.

    I can put a very large cap across the amp output to ground without problems.

    For a while I tried putting 0.1 uF decoupling caps in many places to try to diminish the "noise", but they had no noticable effect. I did seem to get a small benefit from an FM trap.

    For a while I thought some 60 Hz AC might be getting into the circuit from the ground so I made a 60 Hz trap and put it on ground connections. I live in a condo so if a neighbor has something with a bad ground it's possible I could get the AC like that.

    Eventually I removed all the filters because they didn't seem to be any real help. I know it's really better to have them but at least for now, while I'm still doing a lot of rewiring, I wanted to keep the circuit simple to work with.
     
  11. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Hi again,

    Strange that electrolytics pose problems, but never mind - you can get the same effect by putting in a couple of high value resistors between the output of the differential amplifier and the capacitor and instrumentation amp. Use a couple of Mohms.

    I must say, I don't like the arrangement whereby the fets' sources are connected via screened cables. I think a much better idea would be to have Rc at the fets' end, and just use a single cable for the two drains. The cable's screen can be used to provide the -15V supply to the bottom of Rc, as long as you decouple the screen to ground using a capacitor (at the far end).

    How are you getting on with bench testing?
     
  12. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Hi,
    I completed some bench testing and got in 1 full day test with the variables you suggested and it came out fairly good. I found one of my pots had a flakey connection inside that was causing at least some of the noisy results so I removed it.

    I began thinking ahead to connecting the in-amp and I came up with some questions:

    1. Why do you want the zero pot on the drain side of the fets? I thought this was usually done on the source side.

    2. I don't think I can use the dual supply to power this diff amp circuit because I need the power supply 0V ref line to be connected to this circuit like I used to have otherwise the in-amp won't work. The amp needs the inputs to have some connection to the 0V ref line. I don't see how I can connect that ref line to this circuit. There is no point in the circuit that I can say would always be 0V to connect to. Using the 15/0V supply and zero pot on the source side allows me this connection.

    3. The original supply +15/0V would eliminate the potential problem of capacitance between cable shields and source wires which exists with this dual supply because the shields and source wire would be much closer in potential so the capacitance would be much smaller.

    Here is how I connected the diff amp to the in-amp originally. This seems to work ok, though I'm not 100% sure it's wired correctly since it's my first in-amp.
    [​IMG]

    I think we need to look at this before we do any more testing of the diff amp with a dual supply.
     
  13. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    The reason for putting the zero pot in the drain circuit is that putting it in the source circuit will reduce gain - assuming it's working in the linear region, which it will if you're using a split supply and with the correct resistor values.

    The instrumentation amp has sufficient common-mode range to accommodate outputs at around +7V, which is what we're aiming for in the circuit I put forward.

    Please give my suggestions (as I've described) a try.
     
  14. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Here are the results from using your suggested variables. Measurements were taken with -15V rail as reference.
    Rd1 = 821 Ω preset
    Rd2 = 820 Ω preset, adj to 979 Ω (max) to help zero
    Rs = 1.5K Ω preset
    Zero = full CCW (wiper at T1, antenna, side)
    Vd1 = 20.965V
    Vd2 = ~ 20.9V fluctuating (maybe from the bad connection)
    Vs = ~ 20.19V fluctuating (maybe from the bad connection)
    Ids1 = 9.45 mA
    Ids2 = 3.79 mA
    Vds1 = 0.774V
    Vds2 = 0.815V

    I attached a copy of the graphed output.

    In the past I've found I get more amplifiction by working in the saturation region, so I've always aimed for that.

    --------------

    Check this link:
    http://www.allaboutcircuits.com/vol_3/chpt_8/13.html

    About 2/3 of the way down it starts to talk about bias currents and the ground needed with op amps. The INA121 datasheet also says there MUST be a connection between the inputs and reference. I modeled my circuit after the bridge example on the datasheet. See attached.
     
  15. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    Working in the saturation region will guarantee gains of a lot less than 1, especially if the zero pot is in the source circuit. It's something to be avoided, not aimed for.

    I can't stress how important it is to do a proper bench characterisation of the circuit. If you're saying that you've connected the cicuit as I've advised, are operating with a +/- 15V supply, and have both inputs connected to 0V, then it's surprising if the sources are at about +5V.

    If this is really the case, then the drain resistors will have to be reduced or Rc increased. But we can't proceed until the circuit is properly evaluated on the bench (this would include crude gain tests). There is even the possibility that the fets are faulty.

    I'm going to assume that you haven't got both inputs connected to 0V, and have them connected to the antenna/ground system. You can't measure the voltage at the antenna input (because you don't have a high enough impedance voltmeter, I'm assuming), but you could measure the voltage at the other (i.e. ground) input. If this has floated up to plus a few volts relative to 0V, then this would explain what's going on: in other words, you've run out of common-mode range. We really must consider how much the ground input is going to move relative to the local 0V, and take the necessary steps. If we need a large conmmon-mode range then we may need a total rethink.

    Another thing to consider is the possibility of gate (i.e. input bias) currents causing problems - after all, measuring the voltage of a physically large capacitor in the sky is something out of the ordinary. We may be better off with mosfets, for instance.

    Regarding the voltage presented to the instrumentation amp: it can be anything in the amp's common-mode range. As I've mentioned, we may be having problems with the fet circuit's common-mode range, but the INA121 can make sense of anything on its inputs which is within its own supply voltage range (i.e. +/- 15V). All your link is saying is that the inputs must bear some relation to the local 0V (i.e you can't just connect them to something which is floating).
     
  16. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Ok, I can accept being wrong about being in the saturation region. This is the first circuit of its kind I've ever made and I'm learning as I go.

    When I make modifications to the circuit I rely on a simulation to know where I stand. I usually don't breadboard because a bad connection can have a big effect on the operation, so I solder all connections. Also, the circuit is sensitive to my presence near it which changes the operation. I work with the circuit in the metal grounded chassis and all cables are shielded.

    The supply is +15.00V and -15.00V. The circuit is currently floating because I don't have a ground any more. I'm not sure what you mean by "both inputs connected to 0V". It is wired just as my diagram shows. I made the measurements by connecting the negative lead of the DVM to the -15V rail and moving the positive DVM lead around the circuit. Vds was measured directly between the source and drain leads.

    To make many measurements on the working circuit, I would put the jacks back into the circuit that let me easily plug into the circuit without having to unsolder connections and make them without having to handle the circuit.
    As far as making adjustments to the pots in the operating circuit in order to set them optimally in real time, it's just not possible because each influences the output and where the others should be set. With the simulation, I can move a pot a certain number of "clicks" and move another the exact amount, or back them off, but in real life they would immediately get out of sync. (I hope I explained that clearly.)

    You're correct, I can't measure the voltage at the antenna input.
    How would the ground input float up to a few volts? The "non-antenna" fet has its gate directly connected to a good, nearby ground. The ground potential can only vary locally when a thundercloud passes overhead. This is what I want to detect. If the cloud is negatively charged, the antenna fet would detect the negative atmospheric field and the second fet would be detecting the change in local ground potential creating a double input and thus a larger signal. I haven't had any thunderstorms here during these recent tests.

    I don't know anything about using MOSFETS. I recall seeing other static detecting circuits made with JFETs, but I don't recall MOSFETs. I mostly see MOSFETs being used in bridge arrangements for motor controls.

    Because of the way my data collecting software works, I like to operate with the bridge zeroed. At larger signal levels, the software changes the resolution so I don't see the small features I'm used to seeing. I can only blow up the data after I end the data collection. I want to use this as a real time sensor of thundercloud charge, not after the fact.

    The bridge is a nice way to keep the in amp inputs balanced. Without it I would have to add 10M bias return resistors to the inputs and if they weren't matched the bias currents would differ between inputs causing an imbalance.

    I would not be connecting my inputs to a floating, virtual ground. Notice in my schematic that the 0V reference would be connected to earth ground. The same grounded reference is used for both circuits since they are operated from the same power supply. The supply and its connections are not shown on the schematic.

    I have no problem with changing to operation in the linear range, or moving the zero pot to the drain side of the fets. However, I do think I had the rest of the circuit wired correctly and I was able to get fairly good signals from it at times, though they are not as big as I got from my original wheatstone bridge circuit with 1 fet. I don't know why this is.

    I hope all this is clear.
     
  17. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    It's going to be very tricky to know what's going on without making some measurements on the differential amplifier in isolation, i.e. with the gates (which is what I mean by inputs) set to defined voltages. You could take the gates to 0V by connecting them to the cable's screen (this would allow you to zero the outputs). The trouble here is that you have the capacitance of the sceened cable hanging on the fets' sources, which is likely to cause oscillation. That's why I was in favour of getting the circuit on the "bench".

    My advice is to ditch the simulations entirely - they're of absolutely no value and will serve only to mislead. My suggestions assumed fixed value resistors, with just one pot, the zero pot. This is included to take out offsets caused by fet mismatch, and should be adjusted so that the outputs exactly match (i.e. zero differential output voltage) when both gates are at 0V. However, you seem to be saying that you'd like to use it to zero the outputs to counteract a "background" offset due to cloud potential - is this correct? If so, then it effectively serves two purposes, which is fine. But we need to use it initially to evaluate the circuit when both gates are tied to 0V.

    If the external ground is pretty much at the same potential as your 0V, then that's helpful, because we won't need a large common-mode range.

    If jfets are commonly used in static detection circuits, then that bodes well - but we need to get the basic circuit running properly first. (Huge mosfets can be used in power applications, but there are many more tiny ones used in very small-signal applications - your computer alone uses a few billion of them).

    I wasn't suggesting you were attempting to use floating inputs - all I was trying to say that it's perfectly valid for the instrumentation amp to be supplied with inputs which are both at about +7V referred to the local 0V.

    I think that previously you were always operating with the fets saturated, so you'd be getting low gain, low common-mode range and therefore inconsistent behaviour.

    By the way, if your measured voltages were true for when both gates are connected to 0V, then I think I'd be suggesting that the fets are dead - and this may actually turn out to be the case. But we need to do some controlled circuit evaluation (and don't mention the simulator!).
     
  18. Darkstar

    Thread Starter Senior Member

    Sep 3, 2010
    117
    1
    Ok. I can connect the antenna to ground so both fet gates are at an identical potential. I can also connect a battery between the fet gates to put known potentials on them.
    For now lets leave the cables as they are. If I see oscillations then we'll look into them, but the last test, the graph I posted, was extremely quiet, except for a section at the start which may have been from the bad connection. To rewire the cables would be very difficult and time consuming.

    I can build a breadboard circuit that won't have the cables. I can't bring the antenna assembly indoors to be together with the rest of the circuit. It is outside on my balcony. I keep all adjustable parts of the circuit indoors where I can reach them. Furthermore, any changes I make outside have to be able to fit into the container holding the circuit board with the 2 fets which protect them from the weather. This is where the gates would be connected to each other or to a battery.

    I can use fixed resistors if I know what to use. I used the pots instead so I could make changes more easily.

    You are correct that I want to use the zero pot to counteract the cloud potential, whatever that may be, whenever I begin running the unit. This gives me a nice starting point (0V). The atmospheric potential varies in a sinusoidal pattern over a 24 hr period. It also changes from day to day (it's not exactly the same pattern) so it's nice to start at a known spot.
    Using the zero initially to evaluate the circuit when gates are 0V is fine.

    I'll check the fets for shorts and swap them with different fets which check out ok to see if there is any significant differences.

    So, can I go ahead and change the pwr supply back to +15V/0V and ground my circuit before I start on any other measurements??
     
  19. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    When you ground the antenna, make sure you connect both gates to 0V, rather than the "external" ground - you should be able to do this easily using the cable's screen. This should establish quiescent working conditions, but do make use of the zero pot to get the drains at an identical voltage. Using pots for the "fixed" resistors is fine, but make sure you end up with the correct values (I'm not sure if measuring them in-situ will be the best method here). And a 1.5V battery would demonstrate how much gain we're getting.

    Regarding oscillation, I have my doubts as to whether the circuit has ever operated out of saturation, so that might be the reason for lack of oscillation.

    Regarding the -15V supply: this is crucial to correct operation, and indeed Rc is chosen specifically on the basis that -15V will be used. There is no disadvantage in using -15V, only substantial advantages. It's doubtful as to whether the circuit will work at all if you don't use a negative supply. By far the best way of implementing this is to rewire your cables as I suggested (one cable's screen provides -15V, inners carry the drain outputs, decouple screen to 0V at the receiving end, both fets sited together, external ground connection extended if necessary to facilitate this, discard the other cable).
     
  20. Hi-Z

    Active Member

    Jul 31, 2011
    158
    17
    By the way, a quick test for a dead fet would be to short a drain to +15V: if the source voltage goes up to nearly +15V, then you have a bad fet.

    Hopefully your fets are OK and the "problem" lies with the gate voltages being a few volts above 0V.
     
Loading...