Signal Conditioning Problem 1

[lespaul]

I developed a circuit for a module that transforms several types of signals into a digitally compatible signal. I tested this circuit on a breadboard in January and it worked perfectly over the required range of amplitudes and frequencies. I had some PCBs fabricated about a month ago with the signal conditioning circuit and a bunch of digital components. The signal conditioning circuit does not work for a variety of signals now. Low amplitude, low frequency sine waves do not generate a clean signal like I was seeing with the breadboard. So I used the circuit on the breadboard again to make sure everything was the same and that circuit no longer worked! I originally tested the circuit in two locations and it worked perfectly in both locations. I tested it in two locations again and it failed in both locations. I'm completely dumbfounded.

 

[geekEE]

I agree with nbucska. There is no filtering on that circuit. If you have a low amplitude low frequency signal going into that circuit, any noise on the input is going to make the output of the differentiator go nuts.

 

[lespaul]

The instrument amp has a reference pin that is grounded.
When you say I should put a cap across the zeners, do you mean in parallel?

The requirements for the circuit are that it has to operate at 27 Hz and 100 mV for the low amplitude sine waves. The upper limit is not of concern as 20K or so would be fine and the circuit has no trouble at high frequency high amplitude.

 

[itsmoked]

And input leakages.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[Skogsgurra]

Yes, the amplifier has a reference ground. Sure. But what about the signal source? Does it float with respect to the circuit's ground?

Yes. Parallel to the zeners. It will reduce the high-frequency NMR.

Gunnar Englund

http://www.gke.org

 

[lespaul]

Well, this is eventually going to be used in an automotive DA system. So the signals will be coming from sensors that either have no ground or are referenced to vehicle ground. The DA box will be connected to vehicle ground through the lighter plug.

 

[lespaul]

I think the problem is in the back part of the circuit after the differentiator. The signal coming out of the differentiator looks pretty good, but the signal out of the comparator is awful at low frequencies.

 

[Skogsgurra]

Do you want help? Or do you think that this is a conversation piece?

It is now that you have the problem. Isn't it? Or have you verified that the circuit works in a car? Please answer how your signal source is connected. Or do you not understand the question?

Gunnar Englund

http://www.gke.org

 

[lespaul]

The circuit that was on the breadboard works in a car, no idea if the pcb version works because I don't have access to a vehicle for testing at this time.
I don't see how the connection of the signal source affects anything. In my last post I stated that the signal out of the differentiator is fine. This would point to a problem with the comparator circuit would it not?

But to answer your question, the signals are coming from sensors on a vehicle. These sensors are either passive or active. The active sensors are connected with the signal being V+ and the ground being V-. The passive sensors do not have a reference to ground that I know of, that was the point of having the differential amplifier.

 

[Skogsgurra]

Do you have a problem with both the active and the passive sensors? Or does it work with one type and not the other?

The fact that the signal "looks good" after the differentiator is interesting. You have a comparator with only 4 mV hysterisis. And it is referenced to GND. Do you really not see anything on the differentiator output? Using 5 mV/div on the scope?

Gunnar Englund

http://www.gke.org

 

[lespaul]

The problem is mainly with the passive sensor which produces a sine wave that varies in amplitude as the frequency increases.
The active sensors produce square waves that have a steady amplitude as the frequency changes. Those work well for the most part.

The comparator should have 30 mV of hysteresis.
(200K/1K)*15V = 30 mV

It's a recognizable signal on the differentiator output.

 

[Skogsgurra]

No. It is less. You have a 5.1 V zener there. But it is actually around 5/200 = 25 mV. I misread the R4 value.

Gunnar Englund

http://www.gke.org

 

[jimkirk]

This may be a silly question, or I might have missed something in the discussion, but what, exactly, is powering this thing? The only supply in the schematic is for a pull up resistor. It can't be just the ground and 12 volt system of the car, because the comparator reference is at ground, and the AMP02 is not rail to rail.

Also, as shown, the AMP02 is configured for a gain of 1. You need a resistor across the gain set terminals so Gain = (50000/Rgain +1). No resistor equals infinite ohms, giving gain as 1. So as drawn, your differentiator is essentially built around a unity gain buffer.

 

[benta]

Decoupling, decoupling, decoupling of the supply.
Especially at the pullup resistor at the comparator output. If that 15 V is not decoupled close to the circuit, your comparator will freak.

Benta.

 

[lespaul]

By decoupling the power supply are you referring to placing a cap between the powers and ground?
If thats what you mean, the PCB version of the circuit has those caps.

 

[logbook]

Skogs, it's a bit off-topic now but "superstition", I don’t think so.

The day before yesterday I had an oscillator producing spurious tones, but only when run from one particular bench power supply. When I changed to a different model of power supply the tones disappeared. I changed them back and forth a couple of times to make sure.

I was going to investigate what was wrong with the "noisy" power supply yesterday but the effect no longer occurs! I tried for an hour to reproduce the noise by moving wires around, turning on adjacent equipment and so forth, but to no avail.

Now this is a relatively complicated test in that I am having to mix two oscillators and amplify the difference frequency of a few gigahertz in order to see anything on the spectrum analyser. Nevertheless the bench setup had not been changed overnight. I don’t have a week to try to reproduce the faulty condition. It might have been some external interference to which only one power supply was susceptible and may never repeat whilst I am watching the display. In these cases you just have to move on. …

 

[logbook]

As for the original posting, beware of the LM311 inputs/outputs. The + and - input polarity gets effectively swapped depending on how you wire the output. Looks like you are using the output emitter grounded and using the open collector drive. Have you verifed you are actually getting positive feedback?

 

[Skogsgurra]

Having problems with parasitic effects in a GHz design is one thing. Trying to explain problems in a marginal low frequency design with other phenomena than design flaws is another thing. That's why I used the word superstition.

I have not examined the design in detail and cannot say what is failing. One thing that I suspect is that lespaul has a poor understanding of reference levels, operating points, circuits swinging close to rails and such things. It has also been questioned if he at all is using the components as intended.

There is a (fuzzy) limit to what we can and shall do when helping a poster. A poster that starts with a declaration that a circuit was thoroughly tested in January and that it doesn't work in May shall not be told that there are lots of subtle influences that may or may not make his circuit work. That is why I call it superstition.

Instead, we shall make it clear to him that he has to design with the right topology and components, enough margins, filtering, shielding and whatever is needed to make the design reliable and reproducible.

Giving him excuses not to improve his design is not to help him.

Gunnar Englund

http://www.gke.org

 

[logbook]

Lespaul,
thanks for posting the circuit. On closer inspection I am worried about the power rails. Do you actually have a negative power rail for U1?

If U1 were an ordinary opamp (as drawn), and if it had a negative power rail, then the output would naturally want to swing symmetrically up and down around ground (0V).

However U1 is not an ordinary opamp, it is in fact an AMP02 instrumentation amplifier. This also needs a negative rail; in fact it needs at least -5V on the negative rail according to the data sheet. Without a negative rail this circuit cannot function correctly, even with a "rail-to-rail" opamp.

I am suspicious of your requirement for a differentiator. Why do you want to have a gain that increase continuously with increasing frequency? In actual fact R1 is already in series with C1 so the high frequency gain is only around 7 anyway.

 

[VEBill]

For the bandwidth limitation as already suggested (a good idea), how about some nF caps in parallel with the negative feedback resistors (to knock off the unneeded high frequency gain). Might be much better there than at the zeners (input).

For trouble-shooting, don't forget to twist the time scale knob on your o'scope to check for RF oscillations. I've seen cases where the only clue was a slight fuzziness of the traces in the audio range. Twist the knob to find out that the fuzziness is a low level 40MHz RF oscillation on the desired audio. Added a cap in the negative feedback loop to nuke it. You can even choose the cap value to series resonate with its own lead lengths when installed.

Of course, the problems with your circuit might be something else entirely...

 

[lespaul]

Skogs, if you are just going to attack me than please do not post anymore in this thread. I've let it go long enough. I don't need your help that bad that I need to take flak from you.

Log, The AMP is powered by +/- 15 V.

Like I said before I think the problem is with the comparator circuit. I wired up a few sample circuits with a comparator to see how it functioned without the front part of the circuit. I tried using several different LM311s, but none of them worked at all. They were all brand new from the bin. There must be some sort of interference somewhere. I think I'm going to scrap the second half of the circuit and try something else.