NE5532 Distortion - What Happened?

[knobtwiddler]

Hi,
Having used the 5532 for as long as I can remember, I have become somewhat guilty of taking its reliability for granted. However, this has bitten me in recent weeks, where I've experienced a few with raised distortion- as per attached photo. I use numerous other ICs in AF circuits and am no stranger to their foibles, but the 5532 seems to exhibit a particular harmonic behaviour that is unique to it (not so much the spread, but the level seems the same in other cases, and is uniform). Note that all odd harmonics are equal in amplitude. I wonder what failure mode would cause this? And was it a known issue prior to TI relocating production to PRC? Did you ever see this in a Signetics part?

[cmjones01]

I don't know about "genuine" 5532s but I do know there are a lot of fake ones around. A few weeks ago I bought a cheap tone control module from an online marketplace here in Poland. It's obviously imported from China. It proudly advertised that it used NE5532, and the chips do indeed say NE5532 on them, but even the most cursory testing reveals that their supply current and slew rate are nothing like the real thing. They behave much more like humble 1458s. I didn't expect anything else to be honest!

Chris

[knobtwiddler]

Hi Chris,

I should add that the ICs in question were bought directly from TI. I have noticed on recent shipments that, for some lines, they come directly from PRC. There has been much consternation WRT to other ICs when fab lines have been relocated.

One of the first things I wanted to do when I got my first R+S FFT analyser was to find out if the claims I'd read on the Internet about the Signetics 5532 having slightly better performance was true. Of course, cynicism about claims on Internet groups made me deeply suspicious. Well, it transpires that the Signetics does have very slightly better THD+N when compared with TI and other types. I haven't compared the JRC edition - I have a feeling it could be better than the TI and should look into using it.

NB - just to reiterate: odd harmonic distortion is commonly what's seen when an IC has an issue. What's striking here is the uniformity in level and distribution. I'm sure that someone more learned than myself will be able to pinpoint it to a specific failure mechanism.

[daviddeakin]

What circuit? What test conditions? What does the waveform look like when the distortion actually becomes visible to the eye? Hard to say anything from an FFT.

[knobtwiddler]

The IC was originally being used in a Baxandall tone control circuit (a-la Quad tilt). I replaced it with another and all seems well. To be sure I got the culprit, I then took the faulty IC and put it into a test jig, which was when I took the photo. It shows an identical harmonic pattern as when it was in the Baxandall-type 'tilt' EQ circuit. Test jig is wired for inverting, standard circuit with 1KR input R, 10K feedback. The fact that the distortion is the same level for each harmonic, with a flat spread at odd harmonics, makes me think there is a particular failure mode present. I think I've seen it on other 5532s and might have a dig around to see if i kept any.

I don't think a schematic for the 5532 has ever been published. I think its 5534 cousin's topology is out there, and believe it has more parallel input BJTs for lower noise. But am going off memory...

[Cruisin Marine]

Have you looked at it on a scope?

[knobtwiddler]

Thanks. I haven't... What would you look for on a scope? We already know the harmonic pattern, so I can only think if phase is affected?

I ought to have said that the FFT is into a 100K load, so it's not as if I am loading the IC. The reason I am asking is simply out of curiosity, for the sake of learning. To my knowledge, there has never been a schematic published for the 5532. I am wondering what kind of topology and fault could give such a flat -68dBV for all odd harmonics. I am used to seeing harmonics at all intervals and with uneven profiles when something fails. The pattern suggests something to me.

edit - I could look for ringing with a square wave?

[Cruisin Marine]

I would suspect a push-pull stage producing a square wave. I doubt it is triangular as that would tail off at the higher harmonics.
I would like to see the waveform though.
You are not accidentally overdriving and "railing" the output are you? If a feedback resistor has become unsoldered or something, that could cause it too.
Just a silly thought.

[kalee20]

A series of harmonics which don't fall, is not consistent with a square wave though (harmonics fall as harmonic number).

Ultimately they must start to tail off, else the Fourier series would not converge!

'Scoping the output could be educational, if the distortion is large enough to see.

[Radio Wrangler]

Being a fully paid-up sceptic (with oakleaf cluster and bar) I think I'd stick an RF spectrum analyser on the thing and look for traces of hooting, either continuous or in bursts. Spectra that look to go on forever are rather suspicious.

David

[Cruisin Marine]

Originally Posted by Radio Wrangler

Being a fully paid-up sceptic (with oakleaf cluster and bar) I think I'd stick an RF spectrum analyser on the thing and look for traces of hooting, either continuous or in bursts. Spectra that look to go on forever are rather suspicious.

David

I am a fully paid up member too, RF work makes you become a compulsory member lol
I would like to know the cause of it, cos' it could be very useful for an RF Comb Generator

[Diabolical Artificer]

I've hd real issues with 5532's in a preamp oscillating like the proverbial dithering politician, sub it for a TL072 all things are quite silent. The circuit in question is the one attached. The 5532 works ok as U1 & 2 but not U3. The IC's AFAIK were not fakes having come from Farnell not Mr dodgy from HK.

This is with a decent layout and double sided board with the rails on side two and very well de-coupled with both electrolytics and ceramic's on the Vcc pins.

Andy.

[Craig Sawyers]

I've had similar problems with LM4562's doing weird things. Replacement with 5532's cured the problem.

As the 'mericans would say - go figure!

Craig

[Cruisin Marine]

OP Amps don't like capacitive loads, so put a resistor of 100-560 ohm on the output, also often a 22pF-100pF or so feedback capacitor is included in designs to stop HF takeoff.
There is another approach shown here https://e2e.ti.com/blogs_/b/analogwi...pacitive-loads

[daviddeakin]

Originally Posted by knobtwiddler

Thanks. I haven't... What would you look for on a scope? We already know the harmonic pattern, so I can only think if phase is affected?

The phase is half the information, which we don't have. The appearance of the waveform can give us clues to what is happinening inside the IC, though you may have to push it closer to clipping for something to become visible to the eye

[Radio Wrangler]

RF comb generators are usually done as narrow pulse circuits. The PRF sets the comb spacing, the pulse width sets the frequency where amplitude has rolled to a minimum with a sinc envelope to comb amplitude. Circuits may be things like deliberate logic races with a delay element to set pulse width, or to go higher, Step Recovery Diodes.

Baxandall tone controls can be set to give less than unity overall gain at some frequencies. Many opamps have a minimum stable gain figure buried in their datasheets, especially quicker ones. So the Baxandall circuit, depending on the flat gain and cut it is scaled for can be rather iffy with some parts.

NE5532 has internal, fixed, compensation.
NE5534 has an external compensation capacitor allowing you to tune amplifier roll off for stability under a wider range of conditions. Freedom! as they say... from my bedroom window, I can look across fields to the town and see King Robert the Bruce's palace and abbey.

Watch out when circuits act as active attenuators.

NE5532 is a dual, and therefore cheaper per amp than NE5534. It saves a couple of capacitors and it saves having to think about what value to use... so the NE5532 is more popular. But there is a reason for the NE5534.

David

[John_BS]

If I've interpreted your spectrum correctly, those harmonics are at about -70dBc? The fact that they're flat and odd only suggests a symmetrical issue and a very "spiky" hf discontinuity of some kind. There are a couple of possible routes to investigage which might provide more clues:
- if you drop the input level 10dB how much does the distortion change?
- if the op-amp gain is increases 10-fold, what happens to the relative distortion?

My guess is it's some kind of zero-crossing glitch, but so fast it might not be visible on a scope.

[kalee20]

If that's the case, it would be worth seeing what happens with a pull-down or pull-up resistor (a few kilohms to either the +ve or the -ve supply), so that the op-amp's output stage always has either its upper or its lower output transistor doing the work - there's never a crossover.

It will increase the circuit's power consumption, but it may prove a point.

[Diabolical Artificer]

OP Amps don't like capacitive loads, so put a resistor of 100-560 ohm on the output, also often a 22pF-100pF or so feedback capacitor is included in designs to stop HF takeoff.
There is another approach shown here https://e2e.ti.com/blogs_/b/analogwi...pacitive-loads

If that's the case, it would be worth seeing what happens with a pull-down or pull-up resistor (a few kilohms to either the +ve or the -ve supply), so that the op-amp's output stage always has either its upper or its lower output transistor doing the work - there's never a crossover.

Thanks both, et al, Andy.

[Craig Sawyers]

Originally Posted by Cruisin Marine

OP Amps don't like capacitive loads, so put a resistor of 100-560 ohm on the output, also often a 22pF-100pF or so feedback capacitor is included in designs to stop HF takeoff.
There is another approach shown here https://e2e.ti.com/blogs_/b/analogwi...pacitive-loads

Well that is absolutely standard practice.

Craig