TL072 minimum supply

[marcdraco]

I've read the spec sheet, even I can figure out the basics - but although it's supposed to use 2.5v and -2.5v (5v single rail with more for better linearity) I've accidentally discovered the ones I have here are operating - albeit as a simple comparator - down at 3.3 single rail.

I know you all want to chew me a new one but if this is reliable it makes the circuit I'm working on that much easier because I won't have to use a divider to get the voltage down to sub 3.3 (from 5v) - no prizes for guessing this is a microcontroller. A SAMD21 in fact which is only 3.3v tolerant. I know I'm pushing my luck but stellar performance isn't called for, just a reasonably low-voltage op amp. The venerable LM10 (to show my age, I remember the fanfare and the price when it came out) would be fantastic in this application but I had some TL072s in the lolly box so... (Also the LM10 is still quite costly at about 10x that of this little jfet jelly bean.)

Obviously if I have to spec a different device then that's what I gotta do: so long as (a) it can work at 3.3v and (b) is ridiculously cheap . Presumably an NE5532, CA3240 etc. would perform similarly? Doesn't need to be a dual, which is why I could use an LM10 at a pinch but given the modest frequency requirement (sub 1Hz) a dedicated comparator isn't really called for.

Cost is more important than amazing performance. You can't have both, sadly. Saying that I do chuckle to hear audiophiles arguing how good their vinyl sounds on their 20k turntable and valve amp - when the master was probably mixed on a professional desk that was (in the early days anyway) driven with 741s! I see your valve sound and raise you ... this 741 harmonic distortion!

[mikerj]

Clearly it's working more by luck than judgement.  Even with a 5v supply the output pin is only guaranteed to get within about 1v of the 0v rail, how does that fit in with your GPIO logic thresholds?  Also the TL072 is infamous for phase reversal if you exceed it's common mode range, who knows what happens when you run it below it's minimum voltage?

If you need a comparator then use a comparator, not an op-amp e.g. the inexpensive LMV331 may be a good fit.  Remember the output is open collector, so it will need a pull-up.  You maybe be able to use the internal GPIO pull-up if you don't need high speed operation.

[TimFox]

With common devices (from multiple sources) like the TL072, it is very dangerous to assume that since the sample(s) you have tested on your breadboard (presumably from the same vendor, possibly the same lot) work at a supply outside the manufacturer's specified range, that other vendors' parts (or even another lot from the same vendor) will do the same.  I would never recommend operating a device outside of the data-sheet operating conditions.

[Manul]

Well, I guess its better then using it above the maximum supply Go and use it if it works for you and you don't care. For production design it would be a bad decision, otherwise if you're happy, then everyone's happy. What else to say? You seem to be inclined to use it anyway despite understanding yourself that it is not a good design.

[Jwillis]

The data sheets max and minimum values are where the device will operate without derating . This does not mean that it will not operate at a lower or higher voltage . Many op amps will operate at higher voltages but the risk of damage becomes very high . At lower voltages  than minimum value the op amp can become unpredictable and prone to distortion and oscillation . Of coarse this is only what I've experienced . I've experimented with dual rail op amps with single rail supplies and found that they do operate but go weird with higher frequencies .
I also found that drawing current to close to the short circuit value also causes them to start derating and even in some cases burn up or explode . But again this is only what I have experienced .
By keeping the current low and operating well with in the specifications I get the best performance with no risk of damage . 

[Kleinstein]

There are fake chip around that are supposed to be other OPs, but are actually  LM358 (or using just 1/2 half of it) or similar. Usually they sell for something better (e.g. NE5534, OPA2134 or similar) on bangbad or similar.

Many real comparators come with open collector ouputs and can thus outout to a lower voltage than there supply. LM393 is not that expenside either.

[Zero999]

Why not use the LM393, run off 5V, with its output connected to 3.3V, via a suitable pull-up resistor?

Or you could just run the TL072 off 5V and connect its output directly to the MCU's input. The output will be 1.5V below the supply, leaving 3.5V, which is well within the maximum rating of with input pins, with respect to V_DD, which is 0.6V: 3.3+0.6 = 3.9V.

Page 865 of the SAMD21 data sheet.
https://ww1.microchip.com/downloads/en/DeviceDoc/SAM_D21_DA1_Family_DataSheet_DS40001882F.pdf

Pages 19 & 34 of the TL072 data sheet.
https://www.ti.com/lit/ds/symlink/tl071h.pdf

[guymo]

I've read the spec sheet, even I can figure out the basics - but although it's supposed to use 2.5v and -2.5v (5v single rail with more for better linearity) I've accidentally discovered the ones I have here are operating - albeit as a simple comparator - down at 3.3 single rail.

The regular TL072's minimum supply voltage is +/-5V (total range of 10V between the supplies). There's a new variant called TL072H which seems to be quite a different device, whose minimum supply is 4.5V.

The regular TL072 suffers from phase reversal if its input terminals get within 4V of the negative supply which can make it very problematic as a comparator in particular. Phase reversal means that when you would expect the output to be rising, it in fact falls (or vice versa) and ends up at the saturation level, a couple of volts from the supply voltage. I haven't studied the datasheet for the TL072H very closely so I can't tell you if that does it or not.

If the op amps you have are working in your circuit then that's ok I guess -- but if you have a choice, a TL072 is not the right one to use here, as others have said.

[magic]

TL072H is a CMOS opamp supposedly guaranteed free of phase reversal but not guaranteed to work with inputs near ground. Never tried it in practice.

LM358 is a cheap opamp that will work near ground but not near VCC. LM393 is a similar thing with open collector output.

I wouldn't feed the output of a 5V opamp into 3.3V circuits. I see nothing on the schematic suggesting that the output can't swing up to 1V below VCC.

"Phase reversal" is a misnomer, the concept really is only meaningful in unity gain configuration. What exactly happens in TL072 is that when either of the inputs is pulled down too close to ground, the output shoots up to VCC. It doesn't matter which input is taken to ground and what the other input is doing at the time.

[TimFox]

Real TL072’s are JFET input, not CMOS, devices.

[magic]

TL072H is a counterfeit TL072 made by Texas Instruments

For further confusion, the marking on SO8 devices is supposed to be TL072D.

[Zero999]

Or you could just run the TL072 off 5V and connect its output directly to the MCU's input. The output will be 1.5V below the supply, leaving 3.5V, which is well within the maximum rating of with input pins, with respect to V_DD, which is 0.6V: 3.3+0.6 = 3.9V.

This doesn't apply to the TL072H, which has a rail-to-rail output. Only try this with the TL072.

TL072H is a CMOS opamp supposedly guaranteed free of phase reversal but not guaranteed to work with inputs near ground. Never tried it in practice.

Does it have a CMOS input stage? I thought the TL072H still has a J-FET input stage, but the output is CMOS. I couldn't find any schematics. Has anyone taken any die shots?

[magic]

It's listed as CMOS in their parametric search and they removed every mention of "JFET" from the datasheet. Also noisier than the original and T3sl4co1l noticed that leakage over temperature is more typical of MOS than JFET.
https://www.eevblog.com/forum/chat/chips-with-built-in-rfi-protection/

Or you could just run the TL072 off 5V and connect its output directly to the MCU's input. The output will be 1.5V below the supply, leaving 3.5V, which is well within the maximum rating of with input pins, with respect to V_DD, which is 0.6V: 3.3+0.6 = 3.9V.

This doesn't apply to the TL072H, which has a rail-to-rail output. Only try this with the TL072.

No, don't try this

[marcdraco]

Thank you everyone for the useful replies. It's always worth getting a second opinion and a third, fourth... This design is destined for the SAMD21 but 3.3v is everywhere these days (much lower on the some modern dies, my mind boggles!) but will easily translate to other MCUs like the RP2040 which is also 3.3 off the top of my head. 5v is so last century.

Perhaps I should been clearer that this is a single-rail job and it's only comparing a something passing a voltage at about 50% VCC so the design isn't picky - by, er, design... and I can't envisage any way it could go all weird on me in such a configuration. Sure I *could* use a negative rail generator but that comes with its own problems. I learned long ago (Art of Electronics, 1st Edition) that relying on hFE is a like wading into a pool of alligators on a warm Florida evening with a couple of carrier bags of fresh chicken, and shouting.... "Come on if you think yer 'ard enough!"

These TL072s might well be a counterfeit - I got some XR2206s recently for a related project and they were (gasp) also counterfeit. I mean is there ANYTHING the Chinese won't copy? The fake XR2206 cark out at about 11.5 volts I'm told but seem to be OK on a 9v battery so that's why I haven't seen the problem. Thanks eBAY!

Hell I even picked up some (cough) germanium point contact diodes (cough) recently... which turned out to be... Schotkys ...    You knew that didn't you...

Testing a for a "real" TL072 is presumably more difficult without a proper harness and I don't have anything remotely good enough to do that. I'm assuming there's no easy way?

Back in the days when the LM10s came out, a 741 cost me a week's pocket money. We really were that poor - and the local electronics store wasn't exactly local either!  The pride of my test equipment is a Fluke DMM that's about 30 old years old (I daren't look) that I was given when my buddy upgraded to a new fangled Fluke autoranger. Still, what ain't broke, don't need fixing!

I'll get some of those comparators ordered (thank you to those who mentioned it) - I'm happy enough to use an op amp but if there are low-cost comparators, then when in Rome! Never let it be said that I ask for advice and then ignore it!

Clearly the 393 is FAR and away more suitable with a specified 2.0v single supply voltage, there's no chance of this thing coming a gutser! (I hope I spelled that correctly this time, sorry Dave.)

So thank you one and all I hope you enjoy my sideways, self-effacing humour and I wish you all safe grounding and freedom from all things Covid!

[nigelwright7557]

If  I have analogue signals greater than micro can cope with I put a 1k in series with micro input.
If micro is fussy about over/under volts then a schottky to each rail to clip signal works for me.

[magic]

There is a simple way to test if a TL072 has a chance of being genuine. On single supply, connect IN- to OUT and IN+ to ground. If the output is at VCC-1V, it may be a legit TL072 or something close to that. If the output is at <0.2V, it's most likely a rebadged LM358 which is a common thing on auction sites these days.

LM358 will work correctly as a comparator on 3.3V supply with 1.6V threshold and input range from 0V to 30V - well above VCC.

TL072 - I'm really not sure. I think yours may be fake if it works. It could have a chance if it's IN+ which is tied to a fixed voltage and IN- that changes, but 1.6V seems very low anyway. Such operation certainly is far out of spec.

edit
Actually, one other opamp which has phase reversal at the negative rail is RC4558, and those too are sold as other opamps on auction sites. So passing the simple test doesn't guarantee that you have a TL072.

[marcdraco]

Hey thanks! I'll get that on the breadboard when I finish clearing the decks and check it out.

I had a look (out of curiosity) for a 16-bit DAC and found some Burr-Brown chips as you might expect... and some *really* dodgy looking ones too. Prior to this experience I wouldn't have given these a second thought. I can't even tell if the ones with the the typical gold tops seen on most (all?) BB chips might be fake!

[marcdraco]

If  I have analogue signals greater than micro can cope with I put a 1k in series with micro input.
If micro is fussy about over/under volts then a schottky to each rail to clip signal works for me.

I've got some of those in my spares - from eBay... described as germanium point contact diodes. At least they'll come in handy now.

[Kleinstein]

I mean is there ANYTHING the Chinese won't copy?

Probaly not : there are reports of fake 0 ohms resistors.  Not the usual out of 10% tolerance joke, but way higher resistance > 10 Ohms.

Starting with unknown part is not that much fun - unless you are in the guess what chip I got game.
There are plenty of 5 V rated CMOS rail to rail OPs available also for a chep price from a reliable source.  A common one is the MCP6001, but other manufacturers have similar ones too.

[marcdraco]

I think I'll nip back to the 1980s. At least a 5% tolerance (or even 20% carbon) was actually that. 

[magic]

A rail to rail opamp is really not the optimal option. Typically, they will not tolerate input voltages above 3.3V if running on 3.3V and will output more than 3.3V when running on 5V.

LM358 or LM393 is actually a decent choice. As comparators, they will work correctly as long as at least one input is less than 1.5V below VCC and both inputs are above ground and no more than 30V. Note that LM358 output doesn't swing very close to VCC, so a pullup resistor may be necessary. LM393 of course won't work without a pullup at all.

BTW, Atmel's AVR chips contain built-in analog comparators. Check if your SAM doesn't have the same. Then you could simply divide the signal and the reference threshold by half and feed them into the MCU's comparator. Or possibly even use the MCU's internal reference for the threshold.

[marcdraco]

There is a simple way to test if a TL072 has a chance of being genuine. On single supply, connect IN- to OUT and IN+ to ground. If the output is at VCC-1V, it may be a legit TL072 or something close to that. If the output is at <0.2V, it's most likely a rebadged LM358 which is a common thing on auction sites these days..

Well there *is* a surprise. This reads at 2v1v on a 3v3 supply. I would have sworn on my breakfast that I had some 358s in the lolly box, but turns out I must have used (or blown) all of them. Out of curiosity, I tried an NE5532 in this to see what it would do and it outputs about a volt. Clearly that's operating way out of design tolerance but I thought it made for an interesting comparison. All the same, I've got some low-voltage comparators on order since they can operate comfortably well below what I need and with open collector output too, I can run them at the 5V USB (regulated-ish) line and save loading on the 3v3 using a high-value pull up, so that's worked out well!

[marcdraco]

BTW, Atmel's AVR chips contain built-in analog comparators. Check if your SAM doesn't have the same. Then you could simply divide the signal and the reference threshold by half and feed them into the MCU's comparator. Or possibly even use the MCU's internal reference for the threshold.

You're right Magic, they do. Low-end Atmel Mega series, like the 328p only have an 10-bit ADC and the SAM has 12-bit and Teensy 4 (I forget which chip Paul uses) has up to 14. But this isn't about the voltage, it's about generating an interrupt when the threshold is crossed. ADCs on MCUs have a bundle of problems, not least accuracy at the upper and lower limits, but the Arduino framework in particular is horrendously slow and the timings are all over the place. (Even Arduino's delay routines are known for being a bit ... sloppy, presumably due to interupts.)

According to Atmel itselt, the reliable way to get a halfway accurate reading from the ADC is to apply a very well regulated voltage and decouplped source to the external vRef and then shut the CPU down until the ADC completes and wakes it up again. It's a really neat feature but given the extra complexity in programming not to mention that the rest of the code is running concurrently while the comparator waits for the crossing, you can see why I'm using this design paradigm.

[magic]

I'm not talking about the ADC. On AVR there is a separate AC (analog comparator) peripheral which is simply a comparator. You can enable it, read its output state and set it to fire an interrupt when the state changes. All inside the MCU.

[marcdraco]

Sorry, my bad! 

You're quite correct of course... yes it's on D6 and D7 (on Arduino boards) . I obviously need to read those datasheets with more care (I've read that one --- well it's more of a huge book than a sheet -- more times than I care to remember but somehow that escaped my notice. Time to break out a nice smelly haddock and beat myself silly!

I would imagine the SAMD21 has the same but I haven't looked that closely as I've been "designing" the device to work across a number of platforms in order to be MCU agnostic (ish). The SAM21 is a little more useful in this regard because it has a 10 bit DAC too which is useful for generating various signals without having to go mucking around with various oscillators.

Of course the always wonderful, Nick Gammon has a piece about it: http://www.gammon.com.au/forum/?id=11916

Time to break out the coffee and the SAMD21 book of many words then!

EDIT: Woot! Turns out I don't have to!

https://microchipdeveloper.com/32arm:samd21-ac-overview