TL074 - Lower than expected slew rate

[sebajom]

Hi all,

This stumped me last night. I was working on a project where I need to multiplex the audio-rate output signal of a DAC into several channels. I built a sample-and-hold circuit on breadboard:

ESP32 DAC -> TL074 (buffer) -> 74HC4067 mux -> 0.1uF cap (to gnd) -> TL074 (buffer) -> high-ish z load

The signal from the DAC is the full 3v3 p-p and the opamps are on a +/-5V supply I cobbled together from a 7805/LM337 (tests ok). This works fine but I'm seeing a much lower slew rate than expected on the buffers, so I'm not getting the performance I need.

To test the TL074CN in isolation, I hooked it up to +/-5V, configured as a buffer, into a 2 kOhm load, and fed it a 1kHz 3V square from a signal gen.



The buffered signal rises or falls in about 5.5 us. That's a slew rate of about 0.55 V/us - nowhere near the datasheet's specified 13 V/us.



I don't think it's a bandwidth issue - I'm nowhere near the max for that opamp (~4MHz), and lowering the input amplitude shortens the rise/fall time correspondingly. No amount of decoupling seems to make any difference. Neither does the load (within reason). I tried two completely different batches of TL074 (one DIP, one SMD) with virtually identical results.

There's probably something obvious I'm missing. Why am I not seeing anywhere near the slew rate I'm expecting?

[Andreas]

Shure you have a TL074? (fake device?)
the slew rate would fit to a LM324
How is the supply current draw for the TL074?

with best regards

Andreas

[magic]

Andreas is right. We have seen it many times before.
https://halestrom.net/darksleep/blog/038_fakeopamp/

To test the TL074CN in isolation, I hooked it up to +/-5V, configured as a buffer, into a 2 kOhm load, and fed it a 1kHz 3V square from a signal gen.

Now short IN+ to the negative rail.
True TL074 output will go to the positive rail due to phase reversal.
LM324 output will go to the negative rail, and much closer (<0.1V) than a TL074 ever could.

edit
Sorry, it won't come close to the rail with 2kΩ load to ground. But it will if the load is removed.
A true TL07x will never come that close, even if you load it with 2kΩ to the negative rail.

[aheid]

Just curious for my own sake.

The slew rate is specified at +/- 15V with 2kOhm and 100pF load. He's driving it at +/- 5V and 2kOhm but on a breadboard with a 1X probe if the oscilloscope is to be believed, so I'm guessing more than 100pF.

How much, if any, would that affect the slew rate? Assuming one had a proper TL074.

[magic]

I suppose you would have problems when load current approaches limits of the opamp.
To charge 100pF at the rate of 13V/µs, you need (in addition to any other load current):

13V/µs · 100pC/V = 1300pC/µs = 1300µA = 1.3mA

Maybe not a big deal yet, but getting close.

[Kleinstein]

The slew rate may go down with a lower supply, but usually not very much (e.g. comparable to how much the supply current goes down - so maybe some 10-20% in this case).

With a measures slew rate of a bit under 1 V/µs the capacitve loading is not yet a problem. There can be some effect of the load, but usually not much.

A LM324 would also have considerably more input bias than the TL074 - so the capacitor voltage would drift up quite a bit faster.

[sebajom]

Shure you have a TL074? (fake device?)
the slew rate would fit to a LM324

No, it could certainly be a fake - these probably came from eBay originally. In fact, that came to mind, and that's why I tested two different TL074s that were definitely purchased independently. Since they both tested identically, I kind of rejected the idea, but perhaps it's a more widespread problem, and they just happened to be the exact same fake chip.

Now short IN+ to the negative rail.
True TL074 output will go to the positive rail due to phase reversal.
LM324 output will go to the negative rail, and much closer (<0.1V) than a TL074 ever could.
...
if the load is removed.

I'll try this. Certainly seems to match the LM324 specs pretty well. I'll check if there's any crossover distortion as well.

The slew rate is specified at +/- 15V with 2kOhm and 100pF load. He's driving it at +/- 5V and 2kOhm but on a breadboard with a 1X probe if the oscilloscope is to be believed, so I'm guessing more than 100pF.

Could very well be. I was probing at 1x, yes, though I see the same result at 10x. But I'm sure there's plenty of capacitance in the breadboard etc. And as magic pointed out, output current is a factor in my original application, though I had made some rough calculations that indicated that I should be ok and the value of the holding capacitor is not set in stone.

Thanks all, I'll do some more experiments tonight.

[magic]

FYI, two years ago I collected over 12 unique samples of dodgy jellybean opamps from numerous vendors on AliExpress and in Poland.

I deliberately picked "too cheap to be real" deals and auctions with photographs showing fake logos etc. I achieved 100% success rate: not even one of those ICs was genuine, not even one was a genuine chip recycled from scrap boards

This is what you will reliably get going to AliExpress, eBay, Amazon, etc. and sorting by lowest price.
https://www.eevblog.com/forum/projects/whats-inside-the-cheapest-and-fakest-jellybean-opamps/

[sebajom]

Now short IN+ to the negative rail.
True TL074 output will go to the positive rail due to phase reversal.
LM324 output will go to the negative rail, and much closer (<0.1V) than a TL074 ever could.
[if the load is removed.]
A true TL07x will never come that close, even if you load it with 2kΩ to the negative rail.

Sure enough, both of my "TL074" go all the way down to the negative rail (no load). They also exhibit crossover distortion for small signals when pushed a little, similar to the examples linked by magic (see images below).

I'm thinking you were all spot on and all of them are actually LM324s or similar. 

At least it means I'm not crazy, I just underestimated the industriousness of the chip fakers out there. I'm almost surprised it's worthwhile for something like a TL074 which is still a pretty inexpensive opamp. Oh well.

Anyway, thanks everyone! Time to order some real parts.

[sebajom]

Just a quick followup on this: I got some TL074s from a more reputable source and set up the exact same experiment as before. Sure enough, this time I measured a slew rate of about 12.7 V/us, almost perfectly matching the datasheet's 13. 

If there was any doubt before, at least we can say for sure now that every single batch of TL074 I had were fakes.

Lesson learned. Thanks all!

[rstofer]

I think I now know why I order parts from DigiKey and Mouser.  I have enough problems getting things to work without fighting fakes.  The real deal at DigiKey is around $1.50 (or a little less) and I can't possibly save enough money to deal with the problems.  And it's in stock!

Thanks for the info!

[CMTan]

Its not a fake.  You need to read the datasheet more carefully.
The spec of 13V/us is specified at Vcc+ =+15V Vcc- =-15V.

If you apply a lower voltage of +5V and -5V for Vcc, the slew rate will drop.
You also have to look at VOM and ViCR, with a swing of +/-3V, the margin is not sufficient.
Using typical case ViCR = -12 to 15, mean margin on the negative must be at least 3V
VOM = +/-13.5V, margin is 1.5V which is not a problem with +/-3V

With Vcc +5/-5, if it only safe to have a swing of +/-2V, and that if only for typical case. 
It would still fail for worst case.  For worst case, only +/- 1V swing is allowed.

Try to raise the Vcc and you have less of these concerns.

[rsjsouza]

Despite you are correct in saying the 13^V/_µs is specified at the ±15V, the reduction in the slew rate wouldn't drop to 4% of the specified value (something closer to what a LM324 would be). On the other hand, the OP mentioned making a drop in replacement and the circuit worked as expected, thus isolating any probing/breadboard/supply effects from the scenario.

[CMTan]

Could you show you test circuit?

[CMTan]

Please note that specification is tested with unity gain amplifier


If you are having any gain on your amplifier, you should make sure your gain-bandwidth product is not exceeded.
If you do, you slew rate will be limited by the highest frequency it can pass so limiting your slew rate.