TL072 Strikes Again / Unity Gain Buffer Hassles

[Paul Marossy]

So I built this CA3080 based variable state filter thing here: http://www.diystompboxes.com/smfforum/index.php?topic=97688.0

I like it a lot and plan to put on my pedal board. There is one thing bugging me about the build though. If you look at the schematic, there is basically three unity gain buffers using a TL072. When I initially built it, it refused to work with anything other than an LM1458 (I tried JRC4558, JRC4580, TL080). I looked at the datasheet for the LM1458 and it specifically says that it can't latch up. I'm OK with using an LM1458 except that is a little noisey in a high gain situation (hiss).

So all that to ask, IS there a dual opamp that is QUIET and WON'T LATCH UP WHEN USED AS A UNITY GAIN BUFFER?!

[GFR]

Strange. You may try some other bipolar input opamp, maybe a 5532.

Or use a discrete follower.

[GFR]

BTW tha data sheet says the TI TL072 is latch free too.

http://www.ti.com/lit/ds/symlink/tl071.pdf

Latch-Up-Free Operation

[Paul Marossy]

Strange. You may try some other bipolar input opamp, maybe a 5532.

Or use a discrete follower.

I was thinking maybe the NE5532 might work but couldn't remember the specifics on it. I don't have any on hand, I'll have to order a few of those.

I can't change anything, it's on a PCB that someone sent to me, so I'll have to focus on finding the right IC chip instead.

EDIT: The 1458 data sheet says that it won't latch up if input common mode range is exceeded. The TL072 definitely does latch up in this case, it's a quirk of the TL072. Been through this before on something else I built that used a unity gain buffer.

[.Mike]

Paul, do you have an LM358 to try?

I have used it in circuits where an LM1458 would work, but a 5532 would not.

Maybe they're similar-- I'm not sure.

Mike

[Paul Marossy]

I used to have some spare LM358s but not anymore.

[Jazznoise]

The latch up/phase reverse thing is something that occurs when the TL07X series are overloaded. I've used TL07X and TL08x with no muss or fuss but you've got to keep at least a volt from the power rails.  Does it latch with the resonance the whole way down? Try removing the O/P buffer!

[Paul Marossy]

The latch up/phase reverse thing is something that occurs when the TL07X series are overloaded. I've used TL07X and TL08x with no muss or fuss but you've got to keep at least a volt from the power rails.  Does it latch with the resonance the whole way down? Try removing the O/P buffer!

It's built on a PCB that someone else designed and supplied for me. It's much easier to just find the right opamp than it is to start removing/adding things.

[R.G.]

Go get a rail-to-rail input and output opamp, or a an opamp which has an input range that includes the minus power supply and goes to the minus supply  on the output. The LM324 quad "single supply" opamp does this, as does the dual version, the LM2904, and the LM35?.

The unity gain buffer thing is not an issue with the TL07x series. I use them that way all the time. The input common mode range *is* a problem.

[Paul Marossy]

The unity gain buffer thing is not an issue with the TL07x series. I use them that way all the time. The input common mode range *is* a problem.

True. It just so happens that both my problem circuits happened to be in a unity gain situation.

[Ronan]

Paul, RG is referring to LM358 (the LM35?). Another IC which has an input range that includes the minus power supply and goes to the minus supply on the output is a TLC2272, it is also very quiet. Small Bear stock them I believe, if you can't find them elsewhere.

I'm wondering why the TL072 are latching up though, would it be possible to measure the voltage at pins 1, 3, 5 and 7 of the TL072's when latched up?

[Jazznoise]

It's built on a PCB that someone else designed and supplied for me. It's much easier to just find the right opamp than it is to start removing/adding things.

True, but you would just be tapping from before the input to the final buffer stage. Not the most extreme PCB acrobatics, just toucha wire from the non inverting input and if it works trim the pin off(!!!) and put the wires in the relevant sockets.

Does this happen with a signal of any amplitude? Have you contacted the guy who did the boards?

[Paul Marossy]

Paul, RG is referring to LM358 (the LM35?). Another IC which has an input range that includes the minus power supply and goes to the minus supply on the output is a TLC2272, it is also very quiet. Small Bear stock them I believe, if you can't find them elsewhere.

I'm wondering why the TL072 are latching up though, would it be possible to measure the voltage at pins 1, 3, 5 and 7 of the TL072's when latched up?

I know you built yours with the TL072s, but I think maybe I got a defective batch of TL072s or something because they've given me problems with other things too. When I put any dual opamp other than the LM1458s in there it just flat out doesn't work. I suppose it may have something to do the CA3080s I have, not sure. I'll try to measure some voltages with the TL072s in there when I get a chance.

[gritz]

Paul, RG is referring to LM358 (the LM35?). Another IC which has an input range that includes the minus power supply and goes to the minus supply on the output is a TLC2272, it is also very quiet. Small Bear stock them I believe, if you can't find them elsewhere.

I'm wondering why the TL072 are latching up though, would it be possible to measure the voltage at pins 1, 3, 5 and 7 of the TL072's when latched up?

I know you built yours with the TL072s, but I think maybe I got a defective batch of TL072s or something because they've given me problems with other things too. When I put any dual opamp other than the LM1458s in there it just flat out doesn't work. I suppose it may have something to do the CA3080s I have, not sure. I'll try to measure some voltages with the TL072s in there when I get a chance.

If you're circuit is abusing the common mode input voltage then TL07x from different manufacturers can behave differently - I've seen it myself and the phase inversion thing sounds horrible. However, looking at the schematic it appears that the output of the OTA has no load (the non-inverting input impedance of the TL being essentially infinite) and by the same token the + input of the opamp after the OTA has no vref. A bipolar opamp has a less-than-infinite input impedance and gives the OTA's output something to rotate around. I know that the circuit has global negative feedback, but I think (intuitively rather than by punching numbers) that this is the source of the instability. I also think that the original SVF used a 13700 and it's darlington buffers, rather than discrete opamps.

[R.G.]

CA3080s and other OTAs (LM13700, 3280, 5517) all have their bias current dumped into the most-negative voltage rail. The bias current input is clamped to one (3080 and 3094) or two (the others) diode drops above the minus rail. So whatever drives them has to be able to put out a voltage that's never more than about 0.7 to 1.4V above the minus rail, and source a linearly variable current while doing it, and remain sane and functional.

If the opamp is connected to the same most-negative power supply as the OTA, it will have issues to the degree that its inputs and outputs can go near the minus rail.

There are tricks to getting around this. One can:

- provide a more-negative minus rail for the opamp, so it thinks (and is!) it is working near the middle of its power supply range, not right at the edge of its own most-negative.
- use an opamp that works OK there, as in single-supply or rail-to-rail
- offset both the input and output of the opamp so that it's working within its comfort zone but still providing the right stuff to the OTA. This can be done with tricks like zeners, multiple diodes or LEDs providing a constant shift of the inputs and/or outputs up from the negative supply, but feeding the right stuff into the OTA pin.

Opamps vary in their specializations. The TL0xx series is optimized for AC signal amplification and modest DC accuracy within its power supply range. Other opamps are optimized for different things; gain-bandwidth, low noise, DC accuracy, biggest output swing, high current drive, many things.

TL072s are not a good choice for opamps to do DC rectifiers near their own most negative power supply. LM324 and its single-supply brothers and rail to rail cousins are much better.

Draft horses will not win horse races with thoroughbreds, but the thoroughbred is a very poor work animal. Some thoroughbreds run better on sand, some on turf, some better in wet/muddy conditions. Horses for courses. And be sure to read the horse's datasheet before betting.

[gritz]

Valid points R.G. but looking at the schemo in question the Iabc to the OTAs isn't being fed by opamps - the unity gain buffers in question are connected to the OTA outputs with no direct Vref connection.

I have a filterboard with  LM13700s and TL072s happily coexisting, but in that case the filter cells consist of opamp integrators fed by the OTAs.

[Ronan]

Wow, you guys are incredible! Thanks RG and Gritz! As you probably worked out, I was the one who pieced that schematic together. Despite a lot of research, and checking other designs, app notes, and breadboarding, I still didn't get it right!

This is interesting, as I've never had any latch-up on my circuit. After reading the last few posts of this thread, I checked all the pin voltages, all good at 4.7V with a power supply of 9.4V. So I tried powering up the wah pedal with the pedal set for full treble, on my unit this is with the pedal in the up position, as I use wah's "backwards". It latched up, absolutely no sound at all, and came good with a slight "pop" when I put toe down. I tried this several times, same result. Now I understand why there's so few build reports on that thread...on my unit, it always powers up with the pedal down (bass) position, so I never had any problems.

At switch on of power, with the pedal up, latching the TL072, the TL072 output pins sat at around 8V, with one input pin (pin 3) around 0.6V and the other input pin (pin 5) around 8V. I tried an MC1458 and TLC2262 for IC1 and no latch up.

Well that's totally my fault. Thanks guys for figuring this one out! I owe you a beer or five or twenty. Now to work out how to fix this mess I created. But the lesson is a good one!

To start with, Paul, IC1 is the one that needs to be swapped out, the one closest to the trimpot for mixing LP and BP. The other TL072 isn't affected (IC4). Until or unless a better workaround evolves...need to think about it.

[gritz]

If you want to use a TL07x then (if you can find room on your board!) try a 1M resistor in parallel with each of the the 4n7 caps on the OTAs outputs. As the caps' reactance equals 1M at about ~34Hz it shouldn't affect goodness in the audio band too much. I would be interested to hear how it goes.

Edit: Be aware that the schematic is a little schonky - a better method in general is to feed the output of an OTA into a virtual earth (like an opamp's inverting input via no input resistor!) as the OTA's output is a current. Check where the outputs of your OTAs sit with no signal - a 1M load is pretty weak, so the -ve inputs of the OTAs may need trimming to get the outputs to sit at 4V5, but we can cross that bridge if needed.

[Ronan]

I tacked a 1M resistor across each 4n7 cap under the board, but it still latches up the TL072 buffers. There is a side effect, in that the DC voltage at the output of the OTA's (pin 6) varies by approx 0.6V as the pedal moves through its sweep. I then tried 1M from the output of the OTA's (pin 6) to Vref, to give the buffers a Vref on the + input at power up. Still no dice. I don't understand how the output of the OTA can bias an emitter follower, but not an opamp, unless as you said the FET input is zero load. I measured the voltage at the output pins of the OTA's with no buffer IC fitted in place and the voltages were around 0.6V for one and 8V for the other. Maybe it gets back to RG's horses for courses, the TL072 is just too fast at switch on with DC conditions that simply don't suit it?

Edit: And I guess that does indeed make it a "schonky" circuit, very true, but it sounds good (when not latched up)
BUT - all help and guidance is appreciated very much, please remember that.
And I am so sorry I ever offered the pcb for sale, instead of sharing enjoyment, I have created some stress, not good, lesson learnt.

[gritz]

I tacked a 1M resistor across each 4n7 cap under the board, but it still latches up the TL072 buffers. There is a side effect, in that the DC voltage at the output of the OTA's (pin 6) varies by approx 0.6V as the pedal moves through its sweep. I then tried 1M from the output of the OTA's (pin 6) to Vref, to give the buffers a Vref on the + input at power up. Still no dice. I don't understand how the output of the OTA can bias an emitter follower, but not an opamp, unless as you said the FET input is zero load. I measured the voltage at the output pins of the OTA's with no buffer IC fitted in place and the voltages were around 0.6V for one and 8V for the other. Maybe it gets back to RG's horses for courses, the TL072 is just too fast at switch on with DC conditions that simply don't suit it?

Edit: And I guess that does indeed make it a "schonky" circuit, very true, but it sounds good (when not latched up)
BUT - all help and guidance is appreciated very much, please remember that.
And I am so sorry I ever offered the pcb for sale, instead of sharing enjoyment, I have created some stress, not good, lesson learnt.

Doh! I said across the caps without thinking that the cold ends of the caps go to -V. Still, I see you were smart enough to try the 1M from the OTA output to Vref (which is what I meant). Apologies, I'm an idiot. IIRC the LM13700 has internal biassing to the darlington output buffer (according to the specsheet anyway, but it's not shown on the schematic), so that may be why it works in that application. Still, if it works with a bipolar opamp, then go for it!