Weird input voltage problem

[PBE6]

I finally finished building a little specialized preamp for a piezo sensor attached to a wooden box, and everything seemed to be working fine until I plugged it in (!). After testing it with the scope, it seems that with input voltages greater than about 0.6 V everything behaves like it should, but less than 0.6 V causes some very strange behavior in the waveform. The output voltage takes a big nose dive, and the negative swings seem to be heavily distorted (but the positive ones seem ok, just too small).  Here's the schematic:



One other weird thing, with my (slightly used) battery reading 8.9 V, Vref is reading as 3.7 V. Also, I'm getting that voltage at one end of the 3.3M resistor in the first stage, but 0 V (DC) at the input to the first opamp. Huh?? I checked to see if I had accidentally grounded the input, but it's fine.

Any thoughts?

[samhay]

Don't have your answer, but a couple of thoughts:

You have your Vref loaded heavily (R6 and R11 are essentially in parallel), so I am not surprised it is sagging. A large capacitor to ground from here won't raise it to half supply, but might help with transients.

Can a TL081 cope with a 1k3 load without distorting?

Any chance you are using 1u electrolytic caps and have 1/more the wrong way round?

[PBE6]

Thanks samhay. I was a little sloppy on my diagram, I do have a 47uF cap on Vref and the opamps are two TL072's (program always makes them TL081's as a default, forgot to change them). The caps are electrolytic, but the first cap has the negative pointing towards the input (away from Vref) and the second has its negative facing ground. Very puzzling...

[amptramp]

The divider set up by R5 and R6 should sit at half of V2.  OA1 is set for unity gain so its DC output should be the same as its input.  That should force the junction of R10 and R11 to the same voltage, so R11 cannot load anything because there is equal voltage on both ends of the resistor and therefore no current from a DC bias perspective.  (Of course, once you add signal, the resistor becomes active.)  The connection to OA3 and OA4 should not draw any current.  Something is happening that is not on the schematic.

[duck_arse]

I'd say you need some blocking caps between stages. maybe. if you lift OA2, 3 and 4 (tough with those duals ...), does Vref come good?

[anotherjim]

I've never had success with electrolytic coupling caps on a high (Meg ohms) impedance input. The leakage current has always caused a massive DC offset on the amp. You have 1uF which I assume is electro? Change it -  100nF film cap is plenty big enough working down to 1Hz!

[PBE6]

I ended up rebuilding this circuit for scratch and I could find the source of a few bad connections, could have been a sketchy veroboard. I ended up making a few alterations and it seems to be working ok now:

- changed C1 to 1.2nF to cut frequencies below 40 Hz
- changed R4 to 10k to provide more upfront gain
- changed R10 and R11 to 1.3k, added a 3.3uF to R11's tail and connected the cap to ground - this was a major issue!! Still not 100% why, but it was causing the inverting buffer to not invert?!? Attaching R11 to ground through a cap fixed that
- changed R15 and R18 to 10k, found out I was inadvertently creating a voltage divider and losing too much juice

[Digital Larry]

I'm not convinced R2 is needed.

[PBE6]

Ok new problem, I just noticed a strange high-pitched crunchy sound in the background when the input exceeds a certain (relatively low) threshold. I checked the circuit with an audio probe, and it appears at the output of the first opamp gain stage (it does not appear at the input to that stage). I thought a transient might have been getting clipped so I changed the 10k feedback resistor to a 2.2k, but it's still there. I also tried adding a 1k resistor to the circuit input (thought it might be static or something), adding a 100nF cap across the feedback resistor (to kill high freq content) and adding a large cap (22uF and 1000uF) across both Vref and V+, again no luck. Any guesses? Is this a peculiarity of TL072s or opamps in general?

[PBE6]

Digital Larry you may be right. I just wanted to prevent pops when plugging in, but that's easily done by plugging in the input and then the output afterwards. I'll snip it.

[samhay]

Try putting caps parallel to the resistors in the feedback loops of the op-amps. You can calculate their value based on the low-pass frequency requirements you have, but if the feedback resistors are 10k, then 1n is a good place to start.

[PBE6]

I tried putting a 100nF cap across the 2.2k feedback resistor, but it didn't help. Looking at in on Logic's multimeter, it appears to be white noise  like a static discharge or DC leak.

[PBE6]

Is it possible there's not enough feedback being applied? Or that the ground resistor is too small?

[PBE6]

I'm really stumped on this. The only thing I can think of is that I'm asking the opamp to put out more current than it can handle?

The breakup seems to happen when the input goes above about -24 dB, which is 0.06 V. I I'll assume there's a loud transient at -10 dB, or 0.3 V. The total voltage at v- should then be equal to v+ = 0.3 + 4.5(ref) = 4.8 V. The current through the ground resistor is iG = 4.8 / 1.3k = 3.7 mA. Checking the data sheet, it says that the supply current Icc for each amplifier is 2.5 mA max with no load (Vo = 0). Does this mean that I am asking too much from the opamp? Will increasing the feedback, ground and load resistors help? I just ran across a thread on diyaudio.com that says the rule of thumb for a TL072 is that the load should be at least 2k.

[PRR]

> it says that the supply current Icc for each amplifier is 2.5 mA max

Idle supply current is NOT output current.

All the general-purpose op-amps will deliver over 10mA. 20mA is very common. TL07x is coy about this spec because it is pretty saggy, not a hard limit like some other amps, but you can be sure of >10mA.

> rule of thumb for a TL072 is that the load should be at least 2k.

At full 30V supply, for low-low distortion, that is a fair guide. The internal impedance is around 300 Ohms, so when the external load gets around 10X that or 3K the opamp starts to strain. Actually it is pretty energetic and is perfectly fine at 2K. When loaded in 600 Ohms, the output is a bit down and the THD rises enough to see on a meter; however working at less than 30V you can drive proportionally lower impedance with similar results.

1.3K should not be horrible.

If easy, change all to 10K and see.

I have not seen any real DC voltages. The plan in your Post #1 shows all biased from one-half of 1 Volt, which sure would give "input voltages ...less than 0.6 V causes some very strange behavior". But I doubt that plan is as-built. What ARE the DC voltages at each op-amp Output, and at the supply pins?

[PBE6]

Bizarre. I replaced the 1.3k ground resistor with a 3k, no change. On a lark, I decided to stack the opamp - voila! It actually fixed the problem! Not sure how that worked, just speculating that both opamps are now sharing the current load, but honestly I have no idea.

[PBE6]

Sorry, did not see your post PRR. At this point it's all boxed up, plus the sight of it is beginning to sicken me, but it is a prototype so when I get it back from my friend for changes I'll check the voltages.

Thanks for the reply though!! This one was a doozy, made me want through my soldering iron out the window more than once.

[PRR]

> the sight of it is beginning to sicken me

Know what you mean. My Honda has had a surging idle going on 3 months now. By expensive elimination it "has to be" an air-leak somewhere..... but finding it is an unending ordeal. (Many of the classic tricks, like blowing propane at suspect areas, do not work on this too-smart car.)