Au Revoir NE5534

[knobtwiddler]

Originally Posted by Craig Sawyers

Although it is only SOIC8 (and tinier packages), the OPA1611/12 (single and dual) give the same input referred voltage noise in 20kHz bandwidth with 10k source resistance (2uV).

I haven't done the comparison in my RIAA calculator yet, but I will do and report back.

Craig

I think you'll find it's nowhere near the 5534 due to current noise. It is very quiet when fed with low Z, but for MM, I don't think you will get much joy. I have tried the 1612 for line-level applications, but didn't think to try it for MM.

[knobtwiddler]

Originally Posted by Valvepower

Tell me about it, obsolescence of linear semiconductors both IC and discrete is a constant battle… there is a void between a lot of Semiconductor companies and the continuing resurgence of record buying and associated equipment. A few semiconductor manufacturers are on board - so to speak - however a lot of the the big boys aren't

Oh well, as Bod Dylan sung…, The Times They Are A-Changin'...

Terry.

One of the most concerning aspects is that the IC makers aren't being straight about the foibles of some of the new uber-ICs... It is as if they expect OEMs to use several thousand of them, pick up on quirks and work it out for themselves. I can think of one very good example where OEMs were left out to dry... Upon pointing out that the IC had 1/f noise, the maker revised its noise spec, so that it maintained the same figure, but the smallprint showed it from 400Hz and up!!!

When companies get really big (having swallowed up their competitors...), something seems to happen to their ethics. No one is responsible for a specific issue. Upon reporting the issue, they will note the data given to them, with the messenger simply getting a shrug of the shoulders and not even a thank you in return!

[Valvepower]

From an incoming email of 30 mins ago from a supplier…, Nisshinbo pulled out of the NE5534 a couple of years ago.

Of late I’ve tended to use the SM version of 5532 in various positions, and as I previously said it’s been OK for me.

Just my two penn'orth, nowt more than that

Terry

[Ted Kendall]

Originally Posted by m0cemdave

I remember somebody showing me the latest Midas input module, I think it was for the original PF mixer. They pointed out the great innovation - a TDA1034, the original 5534 in a TO5 can. They also told me what it cost, something like £13 (and that was to an OEM). Which was getting on for 2 days' wages for most people at the time!

Worse yet, if memory serves they ran the thing on 24v rails, well outside spec, and somehow got away with it, though the dissipation was huge.

[knobtwiddler]

I have a suspicion that the Nisshimbo / JRC 5532 will be a fraction better than the Ti part in terms of THD+N. I hope to find time to test this theory. Note to self: put order through with Profusion!

[Craig Sawyers]

OK - here's the comparison. Cartridge 610ohm/0.47H, 220 ohm lower arm resistor, 47k load resistor. Referred to 5mV and RIAA EQ.

NE5534A - 77.9dB
OPA1611 - 75.3dB

So the OPA1611 is 2.6dB noisier than the NE5534A. That is mainly driven by the current noise, which is 0.4pA/rootHz for the 5534A and 1.7pA/rootHz for the 1611. The 3x lower voltage noise of the 1611 does not compensate for the 4.25 times greater current noise with a real world cartridge impedance.

The SNR calculations do not include 1/f noise. The 1611 has a much lower 1/f knee than the 5534A in both voltage and current noise, and this will certainly narrow the gap between the two devices.

Craig

[Radio Wrangler]

NE5534 is, indeed, a hard act to follow. There are various opamps claiming performance which looks close, but once you get into the details.

Its 600 ohm line driving capability is rather special, and as Craig has found, its low noise behaiour is with a range of source impedances typical of many transducers and microphones.

It can be beaten, but you have to work at it.

David

[knobtwiddler]

Samuel Groner recreated the 5534 as a discrete block and tested it .

Samuel's ground-breaking opamp study paper can be found here, for those interested in such things: http://www.nanovolt.ch/resources/discrete_opamps/

Predictably, Samuel no longer works in audio, but RF. Funny that!

[lesmw0sec]

Before the 5534 et al was invented, I used an LM301 in a mic. amp circuit, but with a single ended low-noise transistor in front of it and included in the feedback loop. This arrangement set back the importance of the 301 noise and eliminated the noise arising from diff. inputs. The result was an amplifier having a Johnson noise within 1dB of the theoretical over the audio bandwidth. The 5534 was excellent - but it's demise is hardly the end of the world.

[Craig Sawyers]

Originally Posted by knobtwiddler

Samuel Groner recreated the 5534 as a discrete block and tested it .

Samuel's ground-breaking opamp study paper can be found here, for those interested in such things: http://www.nanovolt.ch/resources/discrete_opamps/

Predictably, Samuel no longer works in audio, but RF. Funny that!

Samuel's discrete opamp work was superb. And yes - when I asked him a question regarding one of his designs (his high voltage design that can run on +/-40V rails) he said that he was no longer in audio electronics, but RF systems.

Of course since the time of his discrete 5534 design in 2004, the input devices he specified are (surprise not) obsolete.

Craig

[Diabolical Artificer]

For those of us wanted to snaggle a few 5534's before they become unobtainium is there much difference between all the various suffixes? IE AN, P etc? Looked at the datasheets, can't discern anything.

Re HV opamps, Linear(Analogue) did or do a +75v 0v -75v opamp.

Andy.

[Craig Sawyers]

The A version is lower noise than the non-A

NE5534A at 1 kHz vn 5.5nV/root Hz typical 7nV/root Hz maximum
in 0.4pA/rootHz typical

NE5534 at 1kHz 7nV/root Hz maximum typical
in 0.6pA/rootHz typical

Craig

[m0cemdave]

Originally Posted by Ted Kendall

[
Worse yet, if memory serves they ran the thing on 24v rails, well outside spec, and somehow got away with it, though the dissipation was huge.

Not the Midas desks. They were split rail, + and - 16V.
Must have been somebody else (and I wonder who? )

[Craig Sawyers]

A bit of good news - the OPA210 (single) and OPA2210 (dual) have a lower voltage noise, crucially the same current noise, as the NE5534A. They also has lower 1/f corners and stupidly low distortion. Introduced in 2018 it looks like it will be around for a good long time.

The critical number is - how does an RIAA stage with an opamp with vn and in compare with a noiseless amp?

With an NE5534A - 2.32dB noisier than no amp at all

With an OPA210 - 1.61dB noisier than no amp at all

Both numbers with a realistic MM cartridge load of 610 ohms and 0.47H

Both amps are very close to the maximum possible SNR with a noiseless amp - but the OPA210 is even closer than the NE5534A.

SOIC (or smaller) only and about 2.5 times the price of the now defunct NE5534A - but it looks like it does the job very nicely.

Craig

[Nickthedentist]

That's nice.

I see that SOIC to DIL adaptor boards exist e.g. https://www.mouser.co.uk/ProductDeta...RoCz2kQAvD_BwE

I reckon I could fit an SOIC with my big Weller, so long as I had some desoldering braid handy to sop up the inevitable bridges.

[knobtwiddler]

When the OPA210 was discussed over at diyaudio.com, one member claimed that the part 'doesn't have a guaranteed noise spec'. What on earth does that mean? One of the datasheets implies that the 210 can be used for medical imaging applications. How on earth could it not have a guaranteed noise spec if they expect people to use them in that application?

I did find out from bitter experience that some Ti ICs do not meet their noise spec, and when tens of users complained, Ti simply shrugged their shoulders, admitted nothing and revised the datasheet to give specs above 400Hz... Ergo, I have very little faith in the company, as per my previous post regarding huge corporations mislaying their ethics once they get above a certain size.

I have been hand-soldering SOIC parts for decades (not just me, but associates at various firms), since similar packages first came about. I have found that you get higher failure rates than when everything gets done by the P+P machine. I haven't quite worked out whether it's due to ESD or heat... In the case of ICs with bipolar inputs, they seem to handle hand-soldering a lot better, so I might have answered my own question there...

NB Ti obviously have some of the cleverest designers in the game. But that's not the same thing as being honest about foibles.

[Craig Sawyers]

I'm not sure what that link means either. No part as far as I know has guaranteed specs. Typical, minimum, maximum etc.

If you ask for 100 percent specification screening prior to delivery of parts you add two or three orders of magnitude to the unit price. A good example is the hermetically sealed and oil filled z-foil resistors from Vishay Precision. Each one is calibrated and the resistance etched into the little metal can to 6 digits. £190 a pop.

But as far as the 2021 datasheet for the OPA210 is concerned the table data for noise and the noise spectral density plots are entirely consistent. The OPA210 data is more comprehensive than the data for the NE5534 family.

I do not think that TI is any worse or different to the data from other manufacturers.

Craig

[Radio Wrangler]

Originally Posted by knobtwiddler

one member claimed that the part 'doesn't have a guaranteed noise spec'. What on earth does that mean?

Thermal and semiconductor noise is random, and therefore unpredictable.... So he was after something that Douglas Adams' philosophers demanded: "We demand rigidly defined areas of doubt and uncertainty!"

Any measurement of the averaged power of noise over a finite time leaves you with the uncertainty of whether you happened to hit a loud patch or a quiet patch and repetitions of the measurement will show scatter which diminishes as the averaging time is increased. Certainty takes infinite time, and once you've made one of those measurements, there is no time left in which to use the result

David

[Diabolical Artificer]

Thanks Craig, i'll look for some of those before they vanish completely.

Thermal and semiconductor noise is random, and therefore unpredictable.... So he was after something that Douglas Adams' philosophers demanded: "We demand rigidly defined areas of doubt and uncertainty!"

The man was a genius, genius is an often overused epithet but regards DA entirely deserving. sad he's no longer with us.

Re SMD versus DIP IC's, do the SMD versions have a better performance? When I build circuits using opamps I always use a socket, would the socket have an adverse performance re noise to a soldered straight to PCB?

Andy.

[DrStrangelove]

I'd be surprised if the parts are 100% tested for noise though I suppose the all singing & dancing ATE these days might do it.

In the Siliconix databooks back in the day there used to be "* parameter tested on a sample basis on each batch" sort of thing.

The sheer excitement of testing a PAD1 has to be seen to be believed.