Lightwave communicator: amplifier questions stumping me

Discussion in 'The Projects Forum' started by chisquare05, Mar 22, 2010.

  1. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    I've been playing around with designs for IR voice communicators, I thought it would be fun and informative to start with something basic, like the AM transmitter/receiver pair given by Forrest Mims, which I've attached. I built both circuits, using the same components listed (in particular, the 741), even though better options now exist. As I've built and tested the circuits, I have found myself stumped by three questions about the receiver circuit that I'm very much hoping you can help me with.

    The first is a feature of the design. The 741 can only swing within a couple volts of the negative rail, so by tying the non-inverting input to ground, the output will be stuck at about (say) 2V unless the inverting input swings *below* ground. And this seems to be what happens when I measure it. Why not bias the inverting input to half the supply to allow symmetric swing? Am I missing something here?

    My second question is about the role of the capacitor C1 at the inverting input. Here's how I'm seeing it, but please tell me if and how I go wrong. When phototransmitter Q1 is out of conduction, C1 charges exponentially toward 9V through R1. If Q1 begins to conduct, the collector voltage drops suddenly, as the current flows through R1. This happens faster than C1 can discharge, so the voltage at the 741's inverting input drops by the same amount. (This voltage rises as C1 discharges.) If the drop is sufficient, then the inverting input actually drops below ground allowing the 741's output to rise. And this change is transmitted through the potentiometer R3 into the 386's input. Is this the correct way of seeing it? If not, why not?

    My third question deals with the LM386. In the version I built, the input to 386 look good on the scope but the output does not vary. Perhaps I fried the chip, but what may be the problem is that I don't exactly understand the relationship between the 386's output and its input. I've gone over the data sheet, and I've bypassed the power supply pins properly (I'm pretty sure). I understand that the 386 output is self-centering at half the supply voltage, but how does the output voltage vary with the input when the inverting input is grounded? For instance, does the average voltage map to Vs/2 with deviations from that average amplified? If not, what determines the output's deviation from Vs/2? If so, over what time scale is that average taken? And I know that the 386 does not act like a standard op-amp, but then what is the difference between the inverting and noninverting input?

    I'd appreciate any help you can offer on these questions. Thanks so much.
     
  2. awright

    Well-Known Member

    Jul 5, 2006
    84
    7
    Although I'm extremely reluctant to challenge Forrest Mims' circuit design, that does look like a misapplication of a 741. Where did you find this circuit?

    Notice that the transmitter circuit has a voltage divider (R3 & R4) to bias the non-inverting input of the 741 at half the supply voltage allowing that 741 to operate in its linear range. They then provide a blocking capacitor (C2 in the transmitter) to block the DC output voltage of the 741 while passing the audio on to the bias network for the transistor Q1.

    I'd do the same for biasing the 741in the receiver. In fact, it looks like you could just duplicate the transmitter 741 circuit up to and including C2 for that shown for the receiver. The 10uF blocking capacitor would then be the input for the 10K pot. It's OK to have the 386 inputs ground-referenced since it is designed for that. The input to the transplanted 741 circuit would be just as shown for the receiver circuit.

    There is nothing tricky about the R1, C1, Q1 circuit and no charging/discharging action under normal light and transmitter conditions. R1 simply provides bias current and a load resistor for Q1. A phototransistor is not self-generating and requires a bias current to operate. As the light level on the phototransistor varies, the current through it varies thus changing the voltage drop across R1.

    Under most conditions the audio signal due to the modulated transmitter light output will be very small compared to the DC voltage at R1/Q1/C1. That DC voltage will be determined by the average light level on the photodiode and the resulting voltage drop across R1. This DC voltage could vary quite a bit, depending upon ambient and transmitter average light level. C1 blocks this DC voltage from reaching the 741 thus preventing the DC voltage from saturating the output of the 741. Only the audio signal generated across R1 is passed on to the 741 input to be amplified.

    Your 386 is probably still OK. See how it acts with a blocking capacitor in series with its input. As shown, the 386 has a gain of 20 so any voltage above about +0.2 volts on its input (pin 2) will drive its output to ground. The 741 output may not be able to get to within 0.2 volts of ground so the 386 might be saturated. What is the DC voltage at the 386 inoput and output?

    Don't try to think of the 386 in terms of an op-amp. As you observe, its output is biased at half the supply voltage with no differential voltage between the inputs. A small differential voltage between inputs produces a 20 times larger deviation of the output from its 50% bias point (with the 386 configured for a gain of 20).

    Have fun.

    awright
     
  3. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    Thanks so much for the very helpful reply. I'd like to
    follow up on the questions you asked and raise a point
    that is still less than completely clear to me.

    > Although I'm extremely reluctant to challenge Forrest
    > Mims' circuit design, that does look like a misapplication
    > of a 741. Where did you find this circuit?

    The images I posted are from Forest Mims's Circuit Scrapbook
    Vol 1, and a virtually identical circuit appears in Getting
    Starting in Electronics (p. 123 bottom). I share your
    reluctance (and then some), but what you describe does
    appear to be happening. In particular, the 386 output holds
    more or less constant at about 7.5-8V, with one input
    substantially above ground and the other tied to ground.


    > I'd do the same for biasing the 741in the receiver. ...

    That makes a lot of sense, thanks. I'll definitely try that.

    > What is the DC voltage at the 386 input and output?

    The typical voltage at input is 1.92V scaled down by the
    potentiometer. The output holds around 7.5-8V which is
    essentially saturated with an 8 Ohm load according to the
    386 data sheet. When I probe the 386 input with a scope, I
    see audio-like fluctuations from the transmitter, with a
    little distortion. (I can also see the IR signal from a
    remote control and a nearly perfect 60Hz signal from some
    nearby incandescent lights if I expose the phototransistor
    to them.) But this all rides on that DC bias which seems to
    be saturating the 386.

    > Don't try to think of the 386 in terms of an op-amp. As you
    > observe, its output is biased at half the supply voltage
    > with no differential voltage between the inputs. A small
    > differential voltage between inputs produces a 20 times
    > larger deviation of the output from its 50% bias point (with
    > the 386 configured for a gain of 20).

    That clears it up nicely, thanks, and fits perfectly with
    what I am seeing at the output. It also makes your
    suggestion of a blocking capacitor before the 386 input
    especially compelling. Doesn't that imply immediately
    that the circuit has an error?

    One last question about the R1,C1,Q1 circuit:

    > There is nothing tricky about the R1, C1, Q1 circuit and
    > no charging/discharging action under normal light and
    > transmitter conditions. R1 simply provides bias current
    > and a load resistor for Q1. A phototransistor is not
    > self-generating and requires a bias current to operate. As
    > the light level on the phototransistor varies, the current
    > through it varies thus changing the voltage drop across R1.

    Thanks for the nice explanation. But to understand the state
    at the 741's inverting input, don't we need to see what is
    happening at a finer level of detail. For example, if the
    741s output is to drop above its minimum value (which it
    does), the inverting input needs to drop *below* ground.
    Perhaps I'm just overthinking the problem, but isn't my
    story consistent with your explanation? How would you find
    the voltage at the inverting input for any given level of
    current through the phototransistor? Doesn't that depend on
    a voltage drop across the capacitor being transmitted as the
    voltage at R1/C1/Q1 changes?


    I really appreciate your help on this. I will try your
    suggestions and report back.
     
  4. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    Here are the results of my tests based on awright's suggestions.
    They led me to three new questions which I've appended below.

    1. The blocking capacitor on the receiver's 741 output fixed the
    main problem. The circuit now works in that a recognizable
    voice is coming out. (But see below.)

    2. The biasing of the receiver's 741 input to make a virtual ground
    was decidedly less successful. I tried both reproducing the
    amplifier from the transmitter and biasing the input for
    the differentiator in the original circuit. Both reduced noise,
    but both showed a lot of instability, especially the former.
    And worse, only a sequence of clicks comes through from the received
    signal, which the scope showed this to be a series of spikes.
    No voice, in either configuration.

    3. The transmitter is working in the sense that V_BE on the transistor
    shows up on the scope as a beautiful representation of the sound.
    But I had trouble achieving the right base bias. I think the
    goal is to bias the transistor into conduction so that the current
    through the IR LED is about half max. Then, set the gain so the
    V_BE variations produce current variations through LED at the
    full dynamic range. BUT: when the bias trimmer is set above a
    certain level, it produces a strong oscillation that increases in frequency
    with the size of the bias. Unfortunately, this happens at a lower
    level than it should. The LED has a forward drop of 1.4, so I
    was shooting for about a 2V bias into the base. But that's right
    around where the instability begins. I'm guessing that this is
    not a coincidence, but I don't understand what's causing it.
    If I could control that problem, I could improve sound quality
    markedly.

    My questions:

    A. Why doesn't the biasing of the receiver pre-amp work?

    B. What could be causing the instability in the transmitter as the
    base bias is increased?

    C. A differentiator seems like an odd choice, but that's how the
    receiver 741 is configured. A it's working, certainly
    better than a standard preamp. From a design standpoint,
    why would this be a good choice, especially given the instability
    and noise susceptibility that differentiators tend to exhibit?

    Thanks for your help!
     
  5. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    If you make the circuits with all the missing parts included then they should work fine.
    I don't know where you will find a crystal mic today.
     
  6. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    Thanks very much!

    I'm using an electret as Mims suggested in his later published version. Do you think that's a problem?

    I had already put in the Zobel network on the speaker (from the 386 data sheet) and the power supply 100uF caps. But I'd appreciate if you could briefly explain the motivation for the other differences (e.g., the 2.2uF's with the 100k dividers, the 100k on the tx 741 input). I'd like to learn from this for future designs.

    I also bypassed the power supplies for the three chips with
    0.1uF ceramic || 100uF electrolytic. Would you suggest a different approach?

    Thanks again.
     
  7. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    You did not post the later version so I don't know without seeing it.
    Please post the circuit you used.

    But you posted a circuit without them. Why?

    It filters the reference voltage so the opamp does not oscillate due to feedback from the fluctuating supply voltage.

    The 100k resistors create a "half the supply voltage" as a reference voltage for the opamp so the input and output can swing up and down the most.

    Good. Why didn't you show them?

    Please post the entire circuit you used.
     
  8. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    A fair point.

    The circuit I made originally appeared in two of Mims books, differing only in the microphone. The images I originally posted were from the older source but were the ones I could find on line. I've now attached the newer images, which show the microphone. This is what I originally implemented exactly, using the electret and phototransistor options. But when it did not work and when my testing raised questions, I started making modifications as I discussed and as suggested by awright above. These were just a 10uF bypass capacitor at the rx 741 output, decoupling capacitors on the power supply (100uF) and at the chips' power supplies (0.1uF||100uF), and the 0.047uF and 10ohm resistor at the speaker as suggested by the 386 data sheet. (I have experimented with other alternatives as well but this is the base circuit, which "works".)

    But I do understand why you want to see the whole thing. Sorry about that. I will make up my own images of the schematic I am using now and post them.

    And I will try your suggestions as well. Thanks for the explanations and help.
     
  9. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    Yes, your original schematics are full of errors.
    I think the resistor at the collector of the transistor that is driving the IR LED should be at the emitter instead so that it is in series with the LED. Then the current in the LED will be smoothly modulated by the variable voltage at the emitter of the transistor.
     
  10. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    Sorry it took so long, audioguru, but I've attached the schematics below. Version A gives the version of the circuits I had before your suggestions, modified from Mims's original. Version B gives the circuits I have now based on your suggestions.

    I look forward to your comments. Thanks.
     
  11. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    R4 in my Lightwave receiver can be replaced by a piece of wire.
     
  12. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    Oh, OK, thanks. I've removed it in the attached schematic and will take it out of the circuit.

    Otherwise, what do you think?
     
  13. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    I would replace the noisy low bandwidth 42 years old 741 opamp with a low noise wideband TL071 opamp.

    The circuit resonds to the amplitude of the IR beam so will pickup hum from AC lightbulbs.
    IR communicators use FM to avoid the hum.
     
  14. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    I agree and am planning both steps (better op-amp and FM). I just wanted to try the old style circuit as a baseline for comparison. I was surprised by some of the behavior and what appeared to be mistakes in the circuit, so I wanted to get that working before moving on. I've sketched out a design for an FM circuit and have started building that already.

    Thanks for your help.
     
  15. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    Update: The circuit works very well. The better opamp helps
    stability and sound quality, but even with the 741, it
    sounds pretty darn good and gets surprisingly good range.
    I've started an FM circuit basd on the 4046 phase-locked
    loop, and though it's early in testing, so far so good.


    But if you all aren't too tired of this thread yet, I'd
    appreciate your thoughts on three design/analysis-related
    questions relating to the original transmitter circuit
    (posted above). I want to learn from this experience for
    future designs.

    1. The first attached image shows my analysis of the
    transmmitter output stage, inspired by Horowitz&Hill, which
    leads to different resistor values than in the current
    transmitter circuit (see e.g., second attached image). In
    particular, it lowers the divider resistance looking into it
    from the base. Is this analysis sound, and if so, is it
    sound enough to suggest using the new values (as in the
    second attached image)?


    2. In the transmitter circuit (see e.g., second attached image),
    how does one figure out the voltage at the negative
    lead of C_2 and at the inverting input of the op-amp?

    I would expect the voltage at the inverting input
    to fluctuate around 4.5 (and the op-amp output likewise).
    (Treating the op amp as ideal, the quiescent state should
    have 4.5V at the inverting input, at the output, and at
    the negative end of C_2.) But my measurements of the circuit
    show that the voltage at the capacitor is off by more than
    a volt, which the feedback current can't really make up.


    3. Suppose that before the output stage, the design required
    chips with a smaller supply voltage, say 5V, but that the
    output stage requires the higher voltage/higher capacity
    supply, say to drive multiple LEDs. So I run the low voltage
    part through a 7805 or other regulator and the output
    stage from the full supply, with both from a common ground.
    What kind of decoupling should I use between the stages?
    Or is this approach wrong?


    Thanks for all your help on this!
     
  16. Audioguru

    New Member

    Dec 20, 2007
    9,411
    896
    I think the non-linear IR LED and non-linear photo-transistor would cause plenty of distortion.

    Both dividers are almost the same. The trimpot is not needed.

    The DC gain of an opamp is 200,000 so its inputs have the virtually same voltage.

    Your meter must have a low input resistance to load down the small current from the 1M feedback resistor.

    This circuit uses coupling capacitors between its stages.
     
  17. chisquare05

    Thread Starter Member

    Jun 4, 2009
    10
    0
    Yes, of course, plenty. I should have been clearer that I meant "pretty darn good" relative to my expectations. I was pleasantly surprised.


    I wasn't comparing the dividers in the two figures of that post but the divider in the modified circuit (1k, 1k trim; 1k) with the divider in the original circuit (1k; 50k trim; 1k). My analysis led me to the former, and suggested the latter had too high a resistance. I wanted to know basically if that conclusion was reliable.

    As to the trimpot, what you say makes sense. I did notice in the circuit that weird instabilities would arise for some biasing levels, so I thought the trimmer would help avoid that.


    I understand that the op amp "wants" to make the inputs equal but was taken aback by the measurements. I hadn't thought that the scope could be the problem. (It's a small pocket scope.) Thanks for the tip.


    Not sure what you mean here. Are you saying that the decoupling in this circuit would be sufficient for the hypothetical case I raised?


    Thanks, as always.
     
Loading...