Dieshots: Opamp input offsets working in the opposite direction to what I expect

[mikerj]

You are probably right about boiling vs hydrogen generation.

Sidenote: the crossover distortion of these LM324s is particularly annoying.  When cascading the opamps and operating near/in the crossover region their output seems to pick up all sorts of interference (probably because it's going high impedance).  It took me a while to realise that was the cause and throw another resistor in.

You can work around this by adding a pull-up (or pull-down) resistor on the output to bias it into Class A operation.  Obviously this increases quiescent current and limits drive capability before the distortion returns, but for driving high-ish impedance loads it can work.  That said, the best solution is to place them in the round filing cabinet.

[Whales]

Yeah, that's what I meant by throwing a resistor on.  Indeed it only helps sometimes.

Another realisation: a few years ago I tried using one of these "TL074"s to drive a discrete AB push/pull pair.  No matter how I tried to bias things there was still crossover distortion (or I blew the parts up from creative arrangements or over-biasing).  Doh!

I'm not going to throw them out.  Not only do I want to keep them for the opportune moment when a friend asks for a better opamp ( ) but they're also the only non-single opamp I have left at the moment.  I'll deal with the audio distortion until some more parts arrive.

[xavier60]

The counterfeiters apply surface texturing to the top of  ICs after the original markings are ground off. But the dimple also gets textured which usually should be shinny.

[StillTrying]

I have a question though:
Increasing input impedance at IN+ makes the output go down. This looks like the input bias current is flowing into the pin rather than out of it.
It should be the other way around in an LM324, the bias current should flow out of the pin due to PNP inputs.

I think you're right,    for the input voltage offset being corrected with increasing resistance the bias current would have to be flowing in, and the measured 9mV across the 1M would have to be -9mV.

[Whales]

I don't think I bothered checking the sign of the 9mV - it was a measurement just to see the effects of temperature.

[Whales]

These are the best die shots I can get of one of the fake TL074CN using the equipment I have at hand.  Standard handheld optical zoom camera + a random half-convex lens.



In the lower-right it looks like some black epoxy is still covering some pads.

To decap: I used a cyclic heating and cooling technique.  Heat the chip with a pencil torch until it starts to glow and burn, then quickly dunk it in cold water.  Repeat several times.  Eventually the black expoxy becomes very weak and you can gently crush it into powder using a pair of pliers.

This beats having to completely incinerate the epoxy (it stinks like blown opamp for some reason ) and it's reasonably quick (took me about 5 minutes).  Most of the hard work is finding which piece of epoxy debris contains the die afterwards.

I might see if I can get access to some microscopes at my uni.  Otherwise: I'm happy to mail a couple of these dies to anyone who is interested or can offer better photos

[magic]

By the way, I clicked your DIY diodes project. I'm afraid that forward conduction involves some electrolysis, so the diode will have finite lifespan due to material loss even if it's protected from water evaporation.

[Whales]

Yes, the water electrolyses at the terminals.   I think this electrolysis (namely the O2 generated) interacts with the sodium bicarbonate buffer, making it difficult to modulate the conductivity of these devices.  I ended up doing my (undergrad) thesis on trying to find ways they could potentially be modified to create transistor-like devices.

I should upload that report somewhere.  There's some nice pictures of the al-bicarb glow (of which I've only been able to find one other photo of on the internet, and only showing green).  Also lots of pages of me complaining about various published works claiming they have made "transistors" but said devices appear to be useless for amplification (ie they can only variably attenuate, albeit many papers papers fail to prove even that).

Thanks for having a look at my page, magic. 

[macboy]

Beware that TL072 and similar suffer from something called phase reversal when the outputs swing too close to the negative supply rail. This perfectly explains the odd behavior seen in the original post.

Google "TL072 phase reversal" and read.

This highlights the need to read and understand specifications (and the application notes section) in the datasheet of any device used. This phase reversal phenomenon is not a secret and you would have been warned within the datasheet. That should have caused you to either keep the inputs well away from the rails (because this is not an RR opamp, nor a single-supply opamp) or to choose a more appropriate device. The common LM358 can handle inputs right down to the negative rail.

[magic]

The schematic shows ±6V rails and ground-referenced signal, there shouldn't be a problem.

Yes, the water electrolyses at the terminals.   I think this electrolysis (namely the O2 generated) interacts with the sodium bicarbonate buffer, making it difficult to modulate the conductivity of these devices.  I ended up doing my (undergrad) thesis on trying to find ways they could potentially be modified to create transistor-like devices.

I should upload that report somewhere.  There's some nice pictures of the al-bicarb glow (of which I've only been able to find one other photo of on the internet, and only showing green).  Also lots of pages of me complaining about various published works claiming they have made "transistors" but said devices appear to be useless for amplification (ie they can only variably attenuate, albeit many papers papers fail to prove even that).

Thanks for having a look at my page, magic. 

Truth be told, I have no idea how one would even begin to electrically modulate conductivity of an electrolyte. But that's because I'm quite clueless about electrochemistry

I found your "home-made diodes" link intriguing. I remember a few years ago somebody in America built working semiconductor transistors at home, hydrofluoric acid and all the fun stuff included.

[Whales]

Beware that TL072 and similar suffer from something called phase reversal when the outputs swing too close to the negative supply rail. This perfectly explains the odd behavior seen in the original post.

Nothing is near the negative supply rail, have a look at the schematic and text in the first post.  Ground = 0V, supply = +/- 6V.

[Whales]

I was melting at the fingertips whilst using this microscope.  It's a beautiful piece of equipment -- a big thankyou to the people who let me use it.

Logo and text:


Full die:


I tried to remove the large hunk of epoxy in the lower-right, however all I succeeded at doing was gouging out large areas of the die.  Hopefully no text is hiding underneath it.

[exe]

I'm not an expert, but to me it looks like lm324: https://s.zeptobars.com/TI-LM324-HD.jpg.. I sort of see bjt long-tail pair at the input, so it's not a j-fet opamp. However, I dunno what jfets look like on IC. May be like those weird crosses here: https://zeptobars.com/en/read/ST-TL072-JFET-dual-opamp ?

[magic]

Yes, it's more or less the same chip.

The input is a differential pair of PNP Darlingtons. The input transistor of each pair is right next to the bonding pad and its emitter is connected to the base of the second transistor located towards the center of the chip. The latter have their emitters tied together and fed from one of those multi-collector PNPs near the central axis. Other collectors also feed emitters of the input PNPs to bias them in class A. Next up is an NPN current mirror, combined with those inverting PNPs that are some kind of LM358 idiosyncrasy, look up the schematic.

The output is an NPN Darlington plus a PNP, without biasing. The rest is probably equivalent to the bizarre gain stage of the LM358 whose schematic can be found on the Internet if you care (look for full schematics, most of the time simplified is shown). Somewhere there is also hidden the 50µA output current sink.

The big floods of metal are compensation capacitors, that's always helpful in trying to figure out opamp circuits. The transistors between the capacitors and output pads are the final transistors of the internal gain stage which drive the output stage. They also operate in class A, collectors are loaded by another multi-collector PNP in the center.

[iMo]

The TL06/7/8x series die shots are easy to recognize based on the input jfets, which are made in the form of an "X"

[Zero999]

Yes, it's more or less the same chip.

The input is a differential pair of PNP Darlingtons.

Are you talking about the LM358/324? If so, no, the input transistors are not Darlington pairs. They are emitter followers, driving a differential pair. The collectors are connected to the negative rail, not the collectors of the differential pair, so they're emitter followers.

[magic]

And... I found a dual version of this opamp.

Crap quality because the image was take with household items
But it's quite obvious it's the same circuit and the same logo. Kinda looks like letters GLX to me, probably some obscure Chinese jellybean vendor who made this die.

A keen observer will notice that right channel bonding pads look differently than all the others. That's because the opamp was bonded as a single channel one and sold as OP-07

Clearly the counterfeiters are increasing sophistication of their operations, now being able to special-order a different pinout or perhaps buy bare dies and package them themselves. I wonder how long till somebody makes a dedicated half-358 opamp just for this purpose.

[Twoflower]

I think the text is actually up-side-down. And then it says 971. So could that be a (wannabe) TL971 or TS971?

[magic]

And the dual says 956. These numbers have to be something else.

The topology is clearly LM358/324. It isn't rail to rail, the outputs are emitter followers and the input has two diode drops in series with it.

[Whales]

Crap quality because the image was take with household items

Thats a really good photo if you're only using household items.  I suspect you're using a standard biology microscope of some sort?  The lack of chromatic aberration says you are using some form of reasonable/coated lens.

A keen observer will notice that right channel bonding pads look differently than all the others. That's because the opamp was bonded as a single channel one and sold as OP-07

Wait, floating inputs.  Do you have any spares?  Tie the known inputs somewhere safe, scope the power suppy current draw and "tap" the chip (on the epoxy) with your finger   Could be a new detection technique 

Clearly the counterfeiters are increasing sophistication of their operations, now being able to special-order a different pinout or perhaps buy bare dies and package them themselves. I wonder how long till somebody makes a dedicated half-358 opamp just for this purpose.

Sidenote: The fab people could be separate to the packaging people who could be separate to the marking people (who are likely separate from the sales people you got the chip from).  The OEM of the chip (shall we call them GLX for now?) could be completely legit & doing nothing wrong.

[magic]

It was a webcam lens stacked onto the fixed lens of a 1/3" sensor compact camera.
The digicam needs to be tiny to get close to the webcam lens and see anything through it.
The webcam lens needs to have relatively large diameter glass to reduce vignetting and diffraction.
LED lighting from the side.

[yunLad]

Made my first die shot!

Not worth creating a new topic but I got burned on aliexpress recently, not expecting someone would counterfeit a cheap opamp like TL074. Turns out they were rebranded 324s (at least that what the die states - 324n24)

Couldn't find this die anywhere on the internet so not sure about manufacturer.

[magic]

Yeah, LM358/LM324/RC4558 is basically all you get ordering the cheapest opamps from Ali.
A recently observed addition is the LM393 dual comparator.

The die looks quite small, is it 1×0.75mm or so?
That's another thing about those fake chips, they are distinctly smaller than products of big manufacturers (about 1.5×1.5mm for LM324).
There have to be some unknown factories in China which specializing in making those miniaturized jellybean chips. Their products are so cheap that you will rarely even find recycled LM358 or similar on auction sites anymore.

[yunLad]

Oh, forgot to mention the size. Absolutely tiny, 0.9×0.7mm.

Decided to decap LM324s (that seem to behave up to spec) and it turns out they are different too... Found a more fitting thread so I posted them there https://www.eevblog.com/forum/projects/whats-inside-the-cheapest-and-fakest-jellybean-opamps/msg3884651/#msg3884651

[Cerebus]

5) The LM324 has trouble above a few kHz and the TL074 works fine to 100kHz.

Oh lordy, not this rubbish again.

Yes, the LM324/LM358 makes a lousy audio amplifier but just because it's lousy for audio isn't an excuse to misrepresent its actual parameters at every opportunity you're given to bring it up. You literally do this in every thread you spot where someone mentions the LM324 or LM358.

Just to prove a point, here's Bode plots of an actual genuine ST LM358 (the dual version of the LM324) and an actual genuine Texas TL072C pulled out of the parts draws a few minutes ago. Both configured for a gain of +1 and tested with a 1V signal.:



As one can see the TL072 has a slightly higher 3dB point but at 100kHz they are both quite happy, whereas according to you the LM358 should have run out of steam at a "few kHz"  but it hasn't. As far as slew rate is concerned the TL072 is far superior, but both will happily pass a 2V ptp signal at 100kHz.