Help! Need Help with TL072 Weirdness!!!

[Paul Marossy]

So I recently built a little box to take the signal from a mono output synth and turn it into two signals to feed the inputs of a stereo effects processor. Essentially, it's simply the front end of the ROG "Splitter Blend" - just a dual opamp unity gain buffer that splits the input into two outputs. Should be a simple bulletproof thing, right? But it has been FAR from being a simple thing.

Here's the problem: it either doesn't work at all, or it works one day and not the next or it starts to distort and/or cut out completely while I'm in the middle of playing. I've been using a TL072, and the first go around I found that the IC chip somehow went bad - it had continuity between the V+ & V- pins. Put a new TL072, but it still didn't/doesn't work properly. Tried different opamps, none of them worked at all - either nothing or the faintest output you can imagine with lots of distortion.

So I was thinking yesterday of why the heck it would be behaving like this. I had a thought that perhap the input had too much current on it and the opamp can't handle it. So I did a little search on the internet and found This Page, which says about the TL072: " It is a quirk of this device that the input common-mode range does not extend all the way between the rails. If the common-mode voltage gets to within a couple of volts of the V- rail, the op-amp suffers phase reversal and the inputs swap their polarities. There may be really horrible clipping, where the output hits the bottom rail and then shoots up to hit the top one, or the stage may simply latch up until the power is turned off."

So after reading this page, from my limited understanding, it seems to be the answer as to why this thing is behaving like it is - the opamp latches up and stops working. From the article I gather that the outputs needed a little bit of a load on them because there is a large capacitance on the outputs, so I added a couple of 100 ohm series resistors on the outputs as a starter. It's been working so far this morning, but I am reluctant to say that this fixed the problem as I might turn it on tomorrow morning and get nothing again. I've been operating this on a 9V battery as it prevents several inconveniences for me.

So all that to say what the heck is going on here?!?!?!? How do I make this simple little circuit bulletproof? Help me understand!  

[merlinb]

Add protection diodes from the input to each rail to stop the input swing from exceeding the rails. The input series resistor can be a small value, say 1k.

[Paul Marossy]

That's an idea. But won't that make the input signal clip, too?

[R.G.]

It's possible that you're seeing exactly what you think - exceeding the common mode input range of opamps is a Bad Idea for most opamps. And yes, the TL07X family can be touchy with highly capacitive loads; a series resistor between the output pin and the capacitance is generally the cure.

It is possible that the death throes of the first opamp damaged one or more electro caps, presuming the large capacitance on the output is electrolytic. And in a balanced +/- power supply setup, output capacitors need to be not polar: either film/ceramic or bipolar electros. Normal polarized electros will eventually go bad from any reverse-polarity signals across them too.

If the above is the case, then the cures could be, in no particular order:
- convert the inputs to inverting; since this is not a guitar, it can be a lower input impedance. Going to inverting means that the + input is tied to ground and the - input is held at ground +/- a few millivolts of offset and error signal. Now the inputs don't experience a common mode offset at all. Among opamp designers, the rule used to be "Always invert unless you simply can't." Absolute polarity doesn't much matter with original sources, so it probably only matters that both outputs have the same phase, not necessarily the same phase as the input.
- If you have to have noninverting, invert once at the input, then re-invert on each of two outputs to get back to noninverted phase
- Try it with +/-15 to +/-18V power supplies. Unless your synth output is really, really hot, it won't then be able to drive it out of the common mode range.
- Divide the input signal down by something like 4:1 with resistors, then amplify by 4:1 in your opamp stage. This gives the opamp inputs a smaller signal, which the amplifiers can make up handily.

@merlin: We were typing at the same time. Exceeding the rails is a problem OK, but for the TL07x family, you have to not exceed about two volts less than the rails to keep the inputs within the specified common mode input range.

[Paul Marossy]

It's possible that you're seeing exactly what you think - exceeding the common mode input range of opamps is a Bad Idea for most opamps. And yes, the TL07X family can be touchy with highly capacitive loads; a series resistor between the output pin and the capacitance is generally the cure.

Ah, in which case I have to move my 100 ohm resistors. I have them after the caps on the outputs.

It is possible that the death throes of the first opamp damaged one or more electro caps, presuming the large capacitance on the output is electrolytic. And in a balanced +/- power supply setup, output capacitors need to be not polar: either film/ceramic or bipolar electros. Normal polarized electros will eventually go bad from any reverse-polarity signals across them too.

Yes, 10uF electro caps on the outputs. Do they need to be that big in my case? This might explain why the opamp went bad. Never ran across this one before!

I suppose going with the inverted opamp setup might be something to try if I can't find a simpler fix.

[R.G.]

10uF electro caps on the outputs. Do they need to be that big in my case?

Depends on the (unknown) load you will connect them to.
C = 1/(2*pi*F*R) where the R is whatever gets connected. With 10uF, you're good down to R = 1/(2*pi*20*10E-6) = 796 ohms for a 20Hz lowpass. 1uF gets you to 20Hz with an 8K or greater load. I would use either 1uF film or 1uF/50V NP electro.

I suppose going with the inverted opamp setup might be something to try if I can't find a simpler fix.

Changing to inverted is pretty simple if you already have it built. Same number of opamps, just a rearrangement of the resistors on the front end. Ditto cutting the input signal then amplifying.

Note that rail-rail opamps give you the same problem, just another two volts bigger signal. They at least can be clamped to the power supplies as Merlin suggested.

[merlinb]

That's an idea. But won't that make the input signal clip, too?

Well, yes, but the opamp will have clipped it before the diodes do.

@merlin:  Exceeding the rails is a problem OK, but for the TL07x family, you have to not exceed about two volts less than the rails to keep the inputs within the specified common mode input range.

In my experience the TL07x doesn't actually invert when you exceed the common-mode range- that's more of a conservative rule of thumb- you actually have to exceed the rails. However, the 0.6V drop needed across the diodes is a problem, I suppose.

[Paul Marossy]

10uF electro caps on the outputs. Do they need to be that big in my case?

Depends on the (unknown) load you will connect them to.
C = 1/(2*pi*F*R) where the R is whatever gets connected. With 10uF, you're good down to R = 1/(2*pi*20*10E-6) = 796 ohms for a 20Hz lowpass. 1uF gets you to 20Hz with an 8K or greater load. I would use either 1uF film or 1uF/50V NP electro.

Cool, I was hoping that some 1uF film types would be good enough. I'll try those tweaks first and see what the results are.

According to the effects unit manual, the input impedance is 21K balanced/13K unbalanced.

[R.G.]

In my experience the TL07x doesn't actually invert when you exceed the common-mode range- that's more of a conservative rule of thumb- you actually have to exceed the rails.

My experience matches that.

However, I have seen some opamps that do invert if you get a few millivolts outside the ICMR. I take datasheets to be advertising literature which is barely constrained by legal issues. So when a datasheet tells me that Bad Things will happen beyond X limit, I take into account the idea that the sales guys back at the semiconductor company stated the limits as optimistically as they thought they could get away with. If they thought it was a fraction of a volt/ampere/watt better, they'd put that better value on the datasheet. 

I have seen and participated in using parts beyond their stated limits. But at the time I also had an incoming parts inspection department to test them to be sure each batch did meet the excessive limits that we needed. Mostly we got away with it. Using a part outside the specified limits amounts to betting that you can get better parts than the maker is willing to guarantee. Sometimes, maybe even most times, it works out fine. And it's an OK way to bet if you have only the cost of a handful of parts and some hours scratching your head riding on it.

[PRR]

> In my experience the TL07x doesn't actually invert when you exceed the common-mode range

Mine sure did. (Maybe there's different detail designs?)

I was driving three TL072 in cascade. The wave inverted (did NOT "latch"). My fix was to break the power rails and insert diodes so the stage before the unhappy one ran +/-14.4V instead of +/-15V.

This is different from Paul's case where unknown levels may arrive.

Since this is synth-level, I'd be inclined to take 2:1 attenuation in front, then set up for gain of 1:2 in the opamp.

But my hard-knocks gut says there is an undiscovered mistake in the build. Bad joint, bad cap, etc etc etc.

[R.G.]

Mine sure did. (Maybe there's different detail designs?)

It may depend on how hard it's driven and what the undocumented substrate isolation wells do. Which in turn means which semiconductor fab process was used and therefore who made it and when, as well as design details.

One thing's for sure - they really don't like being driven out of the Icmr. 

I like the invert solution and the divide-down-and-then-amplify solutions both.

[PRR]

Going back to the Real Problem:

> take the signal from a mono output synth and turn it into two signals to feed the inputs of a stereo effects processor.

Gee, I just Y-split the one output to several loads.

In my lab, the synth was 1K output impedance and most loads were 50K or higher. Four loads was negligible load.

In live-performance you may be splitting to ugly places. When I was mastering live concerts for the department and also feeding whatever recorders the students begged me to tap-in, I had isolation resistors. 150 ohm output driving 470 ohm series to "my" machines, 2K2 to "strange" machines. There could be up to 3 "my" shorts or a dozen "strange" shorts and still put good signal to the remaining machines.

That's probably unimportant if only going to one stereo effects processor.

[Paul Marossy]

The thing is that I also want to be able to use it for guitar too

[R.G.]

The thing is that I also want to be able to use it for guitar too

Ahah!

Any good engineer will tell you that the road to H*ll is paved with creeping featurism. 

[Paul Marossy]

Ha ha, well at least I have to try. 

[DavenPaget]

Oh well , unless you would have to deal with switching resistors in and out of circuits , i don't really think you can use it for guitar too .
THAT'S what i know .

[earthtonesaudio]

Mine sure did. (Maybe there's different detail designs?)

I have a big tube of TL074 that I've been using for a long time.  So far they all have this "feature".

[Paul Marossy]

Oh well , unless you would have to deal with switching resistors in and out of circuits , i don't really think you can use it for guitar too .
THAT'S what i know .

It also works OK with guitar, but it distorts a little bit with my volume knob up at 100%. Kind of like the distortion people used to get years ago by plugging straight into the console. My Parker Fly has a preamp in it, so it's probably just too hot at max. volume for the effects unit input.

Anyway, I ended up staying with the non-inverting dual buffer schematic which is basically the front end of the ROG "Splitter Blend", but I changed the 10uF caps to 1uF films and added a 100 ohm series resistor between the caps and the outputs of the opamp. It's been working reliably so far. I think the original build as it was must have been exceeding the common mode input range just like I thought.

The synth doesn't care about feeding this dual buffer circuit, it's an old 1982 analog pre-midi Korg that could have been plugged into a guitar amp or a keyboard amp. So I've set out what I originally intended to do and learned something unexpected in the process. It's a win-win situation AFAIAC.  

[m4j0rbumm3r]

Very interesting read..

I don't mean to hijack this thread, but do you guys reckon TL072 quirks might have anything to do with a loud hissy noise from my Morley wah pedal that is triggered by peaks in the amp-mains (while the pedal runs off battery)?
I described the problem more clearly in this thread: http://www.diystompboxes.com/smfforum/index.php?topic=97542.0

I'm considering a replacement of the TL072s as an attempt at fixing this erratic noise problem. Admittedly, it would be a stab in the dark given my very limited knowledge of opamp types and specs. But then, I'm pretty much at a loss here.

Thanks in advance.

[Paul Marossy]

Very interesting read..

I don't mean to hijack this thread, but do you guys reckon TL072 quirks might have anything to do with a loud hissy noise from my Morley wah pedal that is triggered by peaks in the amp-mains (while the pedal runs off battery)?
I described the problem more clearly in this thread: http://www.diystompboxes.com/smfforum/index.php?topic=97542.0

I'm considering a replacement of the TL072s as an attempt at fixing this erratic noise problem. Admittedly, it would be a stab in the dark given my very limited knowledge of opamp types and specs. But then, I'm pretty much at a loss here.

I don't know that replacing the opamps would help with that, but if you've got nothing better to do you could try it just for kicks. Put some sockets in and you can try several different types.

Regarding your other post, I would also suggest checking C8. Maybe it's going bad and it won't filter out those little spikes in the power anymore. Or possibly it's a grounding issue of some sort.