Replacement for 741

==================================

** Errr - so snap, crackle pop ?

** And never an issue in any but the rarest apps. Certainly not audio ones, where hundreds of millions were used.

...... Phil

all the faults of the 741.

, e

In about 1986, getting rid of a 741 in a particular application - a GaAs s ingle-crystal puller - got rid of a lot of pop-corn noise. I suspect that b ecause the 741 data sheet didn't say anything about popcorn noise, the part number was used as a dumping ground for a bunch of chips that met the rest of the spec (which isn't demanding), but failed pop-corn noise testing.

--
Bill Sloman, Sydney

Bill Sloman wrote: ======================

** Studiously avoids the issue of *how long* the 741s had been there.

The part was released as the " uA741 " by Fairchild in 1968.

Sold for about $10 a piece IIRC.

..... Phil

I bought a couple of uA709's before that, for about $30 each. Largely manag ed to avoid the 741 - the LM301 became available before I might have been t empted by the 741.

By 1986 the semiconductor manufacturers should have been able to make a 741 without pop-corn noise, which was why I was surprised to find that dumping the part had the effect that it did. There were other things wrong with th e circuit I re-designed, but getting rid of the pop-corn noise was the simp lest explanation of the change that the people who used the machine noticed . I was a bit peeved about that - there was quite a lot of clever design (o r what I thought was clever design) in the new circuit, but the disappearan ce of what appeared to be pop-corn noise was what got the operator's attent ion.

--
Bill Sloman, Sydney

Is there a difference between popcorn noise and 1/f noise? or perhaps a better question, is popcorn noise a subset of all 1/f noise sources?

George H. (what I call popcorn noise is when things are mostly quiet, and then you get a burst of some 'large' scale transition (I'm thinking of a

10-20 volt zener diode near it's knee) and then it's quiet again for some random length of time.)

Popcorn noise sounds like, well, corn popping.

It takes two forms. Both tend to be fast jumps between two effective input offset voltages. In one mode, the offset jumps randomly between two values. In the other, it sits mostly at one level and makes short pulses to the other.

An opamp may sit quietly for seconds and then make a burst of pops. The offset voltage can be big. AoE3 has a picture.

I recall something about ionic contamination of oxide layers or something. Early 741s were terrible.

:

e

get a

In the GaAs crystal puller that I worked on, the temperature of the liquid GaAs in its platinum crucible was maintained - very precisely - at it's mel

th the 741 at the front end, the induction heater spent about 30 seconds fu ll on and about 60 seconds fully off.

After I swapped it out, the induction heater ran steadily, delivering about 10kW all the time. Apparently the uV step changes caused the pop-corn noi se were big enough to force the operating system into bang-bang control.

The induction heater was noisy, and it took a day or two to pull a single c rystal of GaAs, so the machine operators were much happier with a machine that generated a constant level of din.

--
Bill Sloman, Sydney

Might be the crossover distortion...

Really? I don't see how. You've snipped out the bit about taking out the 74

1 curing the problem - I didn't replace it with anything had much better be havior at crossover - probably an OP-07 - and the whole control loop was de ad slow - a roughly half-hour time constant. GaAs is like ice in that it ex pands when it freezes, which means the meniscus hit the edge of the single crystal at about 83 degrees rather than 90 degrees and we had to wait long enough for extra weight of the new layer of single crystal GaAs to dominat e the surface tension force.

The guy who had invented the scheme had managed to write it up so confusing ly that he had a patent on it.

The signal that I was passing back to the control loop was just the weight of the GaAs single crystal (which was hung off a precision spring and the m ovement of the spring monitored by a linear variable differential transform er, which we excited with a very stable - feed-back controlled - sine wave oscillator and demodulated the output with a very linear phase sensitive de tector. What had gone back to the control circuit, via the 741 which I'd r eplaced, was a well filtered -10V to +10V DC signal, and the control circui t did a lot more filtering on that.

The 741 had originally directly driven a fairly long screened cable, and th e original "designer" had stopped it oscillating (or at least reduce the os cillation to very low amplitude) by hanging 100nF on it. I put in the usual 100R-to-isolate-the-output -from-the-capacitative-load thing, which worked fine.

There really wasn't any place in the signal chain where cross-over distorti on could have had that kind of an effect.

--
Bill Sloman, Sydney

OK Good. that was what I thought. You see that with higher voltage zeners near the knee... just when they turn on.

TBH I've never listened to much noise, besides the hsss of white noise. (And I wonder how 'flat' the audio speaker and my ear are?) I do recall some sound files with white, pink and 1/f noise... but to my ears, not that much of a difference.

George H.

Yes, IIRC it was process contamination with heavy metal ions, not something special to the 741.

Gerhard

Zeners tend toward relaxation oscillations. That looks a bit different.

uA709s weren't as bad. They were externally compensated. I wonder if the internal comp cap in the 741 was the source of the popcorn noise.

But the 709 had a differential npn input pair, and the 741 was more complex.

--

John Larkin      Highland Technology, Inc 

The best designs are necessarily accidental.

Oxide passivation of silicon results in surface stresses if the oxide is grown rather than sputtered; I'm not sure what the modern approach is, but in the old days, a nitride layer could get rid of it. Those surface stresses make popcorn noise, but '1/f noise' includes long term offset drifts as well as popcorn.

I'm not sure true '1/f' characterization is a good reflection of any random processes, seems more likely to be a dirt-migration phenomenon (which does NOT reverse, so is not random in its noise character).

Popcorn noise is called 'telegraph noise' in the literature, because popcorn is so lowbrow. ;)

1/f noise has a popcorn-ish character. There's a theorem that says that the noisiest kind of stationary process has Poisson statistics, so 1/f noise can't be stationary.

Cheers

Phil Hobbs

Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510

Surface stress doesn't cause noise--it just sits there, just like strained layers in semiconductors. PCN is an ion-migration issue.

I have no idea what you mean by that. Could you say more about it?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Higher voltage "zener" diodes - above about five volts - don't break down by the zener mechanism, but rather by avalanche.

The avalanche process depends on each charge carrier creating more charge carriers as they pass through the junction (which is tiny) which is a statistical process, and stops from time to time when one charge carrier doesn't create any new ones

We had a long thread about this twenty-odd years ago.

It wasn't.

Not in any relevant way.

--
Bill Sloman, Sydney

Grin, well popcorn does seem more like shot noise. Rain on a tin roof. So popcorn noise is associated with a (random) step like transition in the signal. Other types of noise may have amplitude variations at all levels. I once looked at the 1/f noise from a carbon composite resistor (under bias) I don't recall any step like signals. But perhaps they were faster than I could see.

Hmm, OK that makes sense, your ideal Poissonian white noise has a certain predictability about it.

1/f noise is the earthquake or change in conduction path (resistance) waiting to happen.

George H.

Thanks to everyone who provided input. I ordered the Comparators from DigiKey and I was able to find the LT1016 on AliExpress (5x for $2 )

Oh well, order a few lm393s as backup. We once ordered cheaper IC's from china*, they were crap. (rebranded or something.) How will you know if the Ali-express IC's are any good? Make sure to add some hysteresis. (positive feedback)

George H.

*US distributors out of stock and ~3 month delivery time.