Understanding common mode input voltage, can we even use the TL072?

[Fancy Lime]

Hi there,

I realized yesterday that the TL072 and family have common mode input voltage ranges of ((negative supply) + 4V) to (positive supply). Under my understanding of common mode input range that should mean we can feed it +4.5V but only -0.5V if we use it with a single 9V supply and bias it at half supply. That effectively means that we cannot really use that chip in stompboxes the way we often do.

What am I missing? Should we bias TL072's at +6.5V instead to at least get ±2.5V input swing headroom? Using higher voltage supplies would certainly solve the problem. Might explain why many people prefer the OCD with it's 082 at 18V instead of 9. Or am I crazy worrying about safety margins that are irrelevant for our purposes?

Cheers,
Andy

[ElectricDruid]

Does the fact that everyone regularly seems to get away with this suggest that whatever the out-of-spec behaviour that occurs when you do this isn't actually that important to us?

I mean, it could be anything - input bias current no longer remains below Q pA, or input offset goes above 10mV or something - something that we might well not notice.

Interesting though. I don't actually read data sheets for chips I've been using for decades!

[EBK]

"The op amp, of course, flies anyway, because op amps don't care what humans think is impossible."

[Fancy Lime]

Does the fact that everyone regularly seems to get away with this suggest that whatever the out-of-spec behaviour that occurs when you do this isn't actually that important to us?

...

Could be, could be. The thing is, though, that I have always wondered why most circuits in which I used the TL072 always sounded like absolute garbage while "everyone" else seemed to love this chip. Very nasty distortion even when used as a unity gain buffer. Switched to any BJT input opamp, problem gone. I have rarely used the TL072 in recent years, so it is hard to pinpoint what caused that distortion, but it may well have been output reversal due to out-of-negative-range input signals. I tend to mod my bass guitars to include a series mode for both pickups and use the thickest strings I can get my hands on and that combination can produce peak signals well over 1V.

I wonder what commercial pedals use the TL07x family. I know lots of onboard bass pramps use the TL06x series and every single one of them that I ever tried clipped very nastily on hard attacks (apart from creating a noise floor like a jet engine).

Since I cannot possibly be the first person to notice the common mode input range in the datasheet, I wonder what I am misunderstanding. And if I am misunderstanding something, I wonder why my personal experience seems to match what I would expect considering my understanding of the datasheet.

"The op amp, of course, flies anyway, because op amps don't care what humans think is impossible."

Well, as explained above, unlike the bumblebee, the TL072 never really flew for me. I wonder if I finally found out why.


Andy

[antonis]

In case you find the reason, you are free to call it "bumplebee revenge"..

[amptramp]

I noticed this before but did not care for my purposes because I was using ±10 volt supplies with 0 volt ground as the Vref.  In most cases, we don't play with signals greater than a volt which id Vref ±0.5 volts.  Yes, a higher bias voltage like 6 volts should work better.

The worst case is ±11 volts operating from ±15 volts but the typical is +15 / -12 so there is theoretically only one volt of headroom guaranteed but typically 3 volt to 9 volt operating from a 9 volt supply.

[merlinb]

That common mode spec is only a guarantee. As we well know, the TL07x can handle swing much closer to the rail than the datasheet's conservative figure. And it doesn't do the phase invert thing until the input signal exceeds the rail, so it's not something that happens in normal pedal design.

[GGBB]

The thing is, though, that I have always wondered why most circuits in which I used the TL072 always sounded like absolute garbage while "everyone" else seemed to love this chip. Very nasty distortion even when used as a unity gain buffer. Switched to any BJT input opamp, problem gone.

Typical BJT input opamps have the same kind of limitation (unless they are rail-to-rail design). BJTs with P type input devices are limited on the positive rail, BJTs with N type input devices (is there such a bird?) would be limited on the negative rail. Rail-to-rail designs use complementary N and P type input devices. (From https://www.ti.com/lit/pdf/sloa011).

FWIW, I read my TL072 spec sheet (TI) as saying 3V loss not 4V, so +9V would provide for input swing of -1.5V+4.5V. That's triple your 0.5V negative headroom. I think 1.5V vs 0.5V could make a huge difference for our typical 9V circuits.

[garcho]

Very nasty distortion even when used as a unity gain buffer.
I have rarely used the TL072 in recent years, so it is hard to pinpoint what caused that distortion

There's zero chance the rest of us don't hear that, and probably small to zero chance many of us are using similarly modded basses with thickest strings. Cut out the variable. Have you tried something recently with a more standard instrument?

The datasheet I'm looking at says typical is -12 to 15 (±15VDC @ 25°C). How did you arrive at negative +4? Wouldn't it be +3? I certainly could be misunderstanding, just curious. Now I'm the one missing something NE553x is ±13 @ ±15. Is that one of the BJTs you tried?

More datasheet threads! I dig it.

[Fancy Lime]

That common mode spec is only a guarantee. As we well know, the TL07x can handle swing much closer to the rail than the datasheet's conservative figure. And it doesn't do the phase invert thing until the input signal exceeds the rail, so it's not something that happens in normal pedal design.

Hi Merlin, that is some very relevant information, thanks! What other bad things happen before exceeding the rail then? I only read a few design guidelines some while ago (I think from Analog Devices,  but I'm not sure EDIT: I think I found it again: https://www.analog.com/en/analog-dialogue/articles/robust-amplifiers-provide-integrated-overvoltage-protection.html#), which stated that the recommended common mode range should not be exceeded to avoid the inversion problem but not if there are other reasons as well.

@GGBB, garcho
Yes, the datasheet says -12 to +15 at ±15V supply *typical*, which, although no specified directly, implies -1.5 to +4.5 in a single supply 9V circuit. But *minimum* is ±11 and the Recommended Operating Conditions say "Vcc- +4V to Vcc+" as common mode range. I find it useful to design by minimum or maximum (whichever is the relevant one for the parameter in question) and disregard "typical". The typical value may be interesting to statisticians (if they get an explanation out of the manufacturer, what "typical" even means) but it is pretty meaningless for an individual device. The average shoe size for men is about 10.5 (in the US) but that is not exactly a useful number for me when I go shoe shopping for my personal pair of feet, is it. That being said, yes, if we take the "typical" values, the problem goes away. That one volt makes all the difference between unusable and no problem.

EDIT: yes I did try the 5532, it's my go-to opamp for most things. Input currents / input impedance make it a less than perfect substitute for a jfet input opamp in certain situations, though.


Thanks for the input,
Andy

[anotherjim]

One that we should bias away from 1/2V+ is the LM358/324. In order to work down to the negative supply, these BJT are skewed away from V+. You can expect the output to miss the +rail by 1.5v and not drive an input closer than 1v below V+. This actually is better than many normal BJT amps in terms of swing range. So 9v projects ought to bias closer to 3.75v. And don't forget to hang a 10k load to 0v from the outputs to counter the crossover distortion these parts are accused of.

[Fancy Lime]

I looked over the datasheets of the usual suspects for JFET input opamps. Most have identical or very similar specs to the TL07x series, including the TLE207x and LF353. The only ones doing significantly better among the sort of reasonably priced are the OPA2134 and AD8626 and their respective series brethren. These are also protected from phase reversal. However, they are quite a bit too expensive to replace a TL072 for most applications, especially if we can avoid the problem just as well with careful biasing. Not sure if any of this is very useful information to anyone, I just thought I share it since I looked it up anyway.

Andy

[Fancy Lime]

So it turns out once again that I should read the entire datasheet before asking stupid questions. On page 29 of 69 of the TI datasheet for the TL07x series it says quite clearly:

8.3.2 Detailed Design Procedure
•  Avoid input voltage values below 1 V to prevent phase reversal where output goes high.
•  Avoid input values below 4 V to prevent degraded V_IO that results in an apparent gain greater than 1. This may cause instability in some second-order filter designs.

So phase reversal should indeed not be a problem with any half reasonable stompbox design. I wonder about the second point. Should not be a huge problem for unity gain buffers apart from some potential non-linearity. So really not an issue, right?

Sorry to waste everyone's time,
Andy

[Nitefly182]

I wonder what commercial pedals use the TL07x family. I know lots of onboard bass pramps use the TL06x series and every single one of them that I ever tried clipped very nastily on hard attacks (apart from creating a noise floor like a jet engine).

TL071/2/4 are in everything. They are the go-to low noise opamp for audio almost everywhere in the music industry where designers need an inexpensive opamp. Onboard preamps use the 62 because it draws much less current. The 82 is higher spec but in guitar/bass applications there isn't much of a difference and if you want better performance you're moving to a faster part or r2r option or something like that.

[Phoenix]

TL071/2/4 are in everything. They are the go-to low noise opamp for audio almost everywhere in the music industry where designers need an inexpensive opamp. Onboard preamps use the 62 because it draws much less current. The 82 is higher spec but in guitar/bass applications there isn't much of a difference and if you want better performance you're moving to a faster part or r2r option or something like that.

TL08x used to be higher noise than TL07x (I believe binned from same die), but are now spec'd the same with process improvements. But historically, TL07x was the better part, not the other way around.

[Nitefly182]

TL071/2/4 are in everything. They are the go-to low noise opamp for audio almost everywhere in the music industry where designers need an inexpensive opamp. Onboard preamps use the 62 because it draws much less current. The 82 is higher spec but in guitar/bass applications there isn't much of a difference and if you want better performance you're moving to a faster part or r2r option or something like that.

TL08x used to be higher noise than TL07x (I believe binned from same die), but are now spec'd the same with process improvements. But historically, TL07x was the better part, not the other way around.

Ah got it. I've seen them in many places compared to higher performing parts but those may have been errors.

[ElectricDruid]

TL071/2/4 are in everything. They are the go-to low noise opamp for audio almost everywhere in the music industry where designers need an inexpensive opamp.

...and that's absolutely fine where you've got a +/-15V supply and you can run 20Vpp signals and still not hit the situation we're being warned about by the datasheet.

To be honest, it's our own fault for trying to run a chip designed for 30V supply on 9V!

Still, it sounds like you can go close to a volt from either rail without anything too horrible happening. That's broadly the same as the output clipping, which is about two diode drops from the rails, iirc.

[Nitefly182]

TL071/2/4 are in everything. They are the go-to low noise opamp for audio almost everywhere in the music industry where designers need an inexpensive opamp.

...and that's absolutely fine where you've got a +/-15V supply and you can run 20Vpp signals and still not hit the situation we're being warned about by the datasheet.

To be honest, it's our own fault for trying to run a chip designed for 30V supply on 9V!

Still, it sounds like you can go close to a volt from either rail without anything too horrible happening. That's broadly the same as the output clipping, which is about two diode drops from the rails, iirc.

I run 'em +/-12V and they behave real nice 

[DrAlx]

Regarding OPs original question on whether bias level half-way between supply voltages causes a problem, does this not depend on how the opamp is used?
e.g. for an inverting gain stage with supply pins at 0V,9V, then a bias level of 4.5V applied direct to the non-inverting input will cause both inputs to be at 4.5V, and therefore within the spec'd CM input voltage range regardless of signal level.