Very low noise op amp recommendations?

[Craig Sawyers]

Doug Self's idea for paralleling capacitors is that if the values are normally distributed, the parallel combination of N capacitors has an effective tolerance of 1/root(N) times the single capacitor tolerance. In practice taking capacitors out of a batch, there is little evidence that they are normally distributed (as measured on my GR bridge), so you probably will not get as much as root(N).

For the five 10nF (+/-)1% capacitors in parallel you would expect 1%/Root(5) or +/-0.45%, but in practice somewhat more than that.

Apart from that quibble, that is exactly the circuits I use with my Shure V15/IV, and it performs very well indeed. It feeds a headphone only listening system, so noise is a big issue. And without a record playing it is very quiet indeed.

Craig

[paulsherwin]

Maybe Andy should breadboard the VSPS first in his op-amp experiments, and see what he thinks of it. You could lash up a stereo version in less than 20 minutes if you have all the parts to hand, and as the website blurb says, it's simple enough to experiment with individual components. You could use a couple of Poundland PP3s as a PSU.

[Ambientnoise]

A small query - won’t the lack of a polarising voltage across C7 give problems long term ? It’s quite a large value cap. Maybe a special type was specified, I think tang beads were used in the Texan but maybe that’s not one to follow.

[Craig Sawyers]

Tants are not the right thing to use in that application. Cyril Bateman (RIP) measured solid tantalum bead capacitors with 0.3V AC across the capacitors and measured 0.05% distortion from every one he tried, regardless of manufacturer.

If you want to use an electrolytic, the lowest cost way to do it without worrying about the way round it is, is to wire two aluminium electrolytics back to back - it looks like a non-polar capacitor - and has orders of magnitude less distortion than a tantalum bead.

Or use a non-polar electrolytic (AKA bipolar electrolytic). Again exceptionally low distortion. Usual suppliers (Farnell, Mouser etc). Or HiFi Collective https://www.hificollective.co.uk/com...chicon_es.html which are typically 30p for the sort of value/voltage you would need.

Craig

[Radio Wrangler]

The DC voltage across C7 in Self's design will be small and could be of either direction. Aluminium electrolytics would not handle this. Some tantalum types would be OK-ish, but life expectancy might be disappointing.

This capacitor is used to handle the DC offset issue in the design Paul recently cited.

As for Self's paralleling of multiple capacitors...

If the capacitors are evenly distributed acroaa their specified range, the probable error will be reduced as stated. But the worst case error will be the same. You might be lucky or you might not, but he has shifted the odds in favour of lucky.

However, if you buy a bunch of parts at once, then it is likely that they all came out of the same batch and production methods mean that parametric errors are often batch related, so the idea of getting better accuracy turns to tosh. It's still no worse than having a single capacitor of the same tolerance for the full value, but the improvement in accuracy may just be wishful thinking. It's unlikely that anyone will hear the difference for certain, anyway.

If you want more accurate capacitors, they are available. Often they are weeded out from the ordinary grade ones... meaning ordinary grade ones have distorted probability density functions with a gap in the middle, where the accurate ones once were...

David

Would you proper engineers care to explain the drawbacks of minimalist single chip designs like the RJM VSPS?

More a pragmatic engineer here, it's more than good enough and as Sam Dodsworth would (almost) say "Just sit back and enjoy the music (beer)". The only real (almost) problem is the gain won't go below one upsetting the RIAA curve a weeny bit, and that wouldn't be noticeable anyway.

I have solved all these problems by putting on a CD.

[Radio Wrangler]

Not on one of those CD players with a valve sticking out the top?

They give me a fit of the giggles as I think of what a CD player implemented in pure vacuum tube technology would be like. Colossus pales into insignificance. I suppose I'd allow core memory, a glass delay line, but the greenies might have a fit with mercury.

David

[bikerhifinut]

I have enjoyed and appreciated the serious aspects of this thread, so thanks on that folks.
I figure that I possibly asked the wrong question in a way, but I got a set of answers that I can work with and I shall re read the posts as carefully as I can.
I do think that it's possible to get OTT when figuring out how to get the best practical result from this particular exercise but that doesn't mean, in my book, that one should settle for a mediocre solution when for not a lot more effort it is possible to get a surprisingly good result and extract the best musical enjoyment and that little bit of self satisfaction from something constructed in the workshop/kitchen table etc.
And if you want an insight into the way my mind works then "Good enough" translates into "satisfactory but average" in many cases.
That single chip circuit that Paul posted the link to looks so minimalist, but its quite fascinating. It's not a whole lot different to the single chip velleman kit which I have built, although that didn't use the dual power supply, but was simple in the extreme, and didn't like 5534 chips especially when I fiddled the gain up as it did need beefing up to work with my kit. David RW put me wise that time on why the TL072 was better suited to that particular application, and it has made me look differently at the people who claim dropping in "new improved op amps" will guarantee some sort of instant performance boost when in fact the opposite might just happen.
Thanks gents again.
It's good to learn things for their own sake too.

A.

[Craig Sawyers]

Originally Posted by Radio Wrangler

Not on one of those CD players with a valve sticking out the top?

They give me a fit of the giggles as I think of what a CD player implemented in pure vacuum tube technology would be like. Colossus pales into insignificance. I suppose I'd allow core memory, a glass delay line, but the greenies might have a fit with mercury.

David

Voyager 1 and 2 were designed and launched so long ago that the experiment data store was a (space qualified) cassette recorder.

They are still alive after all that time, and transmitting data back at 1k bit per second.

And all the planetary probes use valve-like technology; the source of microwave power for transmitting data back to earth are travelling wave tubes. 60W is all that is needed. A staggeringly small power, but quite enough.

Totally off topic, I'm afraid.

Craig

[turretslug]

Well, resolving signals as low power as that from the very edge of the Solar System still fits with "very low noise amp" I guess.... I'm also slightly stunned at the idea of a cassette recorder still doing its stuff in those lonely little metal boxes so far away. Who made them- Kudelski, or some US exotica concern?

Colin

[Craig Sawyers]

Back when they were doing their grand tour of the solar system, there were a wide selection of instruments turned on (Imaging, spectrometers, radio astronomy and so forth). They generated quite a bit of data during planetary fly-bys which needed to be buffered until the transmitter dishes were pointing back to earth - hence the tape recorder.

But now in interplanetary space, a much smaller suite of instruments are turned on (cosmic ray, charged particles, plasma, magnetometer) that are applicable to the environment. And their trajectory is stable and the dishes are permanently earth-pointing. So no need any more for a data buffer.

Even at their seemingly extreme distance of ~17 light hours, they would (even if they were pointed in that direction) take ~50,000 years to get to our nearest neighbour star - The Alpha Centauri system.

However, using photon pressure using a solar sail, you could (in principle) get to Sirius and get into orbit quicker than you could get to the Alpha C system. Calculations (apparently) come out at 140 years to get to Alpha C and 69 years to get to Sirius.

Once at Sirius the one way light time is 8 years, so asking a question and getting an answer would be 16 years.

Horribly off topic. Other than low noise! And a very slow conversation....

Craig

[Radio Wrangler]

You know the way folk-beliefs grow without any basis in fact.... well the further from the truth, the better they thrive?

Well, one such belief had grown that Morse will get through poor signal to noise ratios better than any other mode? Well it evolved into being hand-sent Morse. Oh well.

Once when giving a talk I showed a slide of an unidentifiable American soldier. I announced that this was Curtis C. Curtis Jr III, who could send clear Morse at over 40wpm. He volunteered for a suicide mission, to go on the Voyager mission and send all the pictures of Jupiter, Saturn and their moons back, coded into Morse, the only mode which could get through at that range. It took a while for some people to twig that it was a wind-up.

With the ice well and truly broken, I went on to the consequences of noise figure, and then correlation techniques, spreading gain and the 'FM advantage' then showed that you could receive signals lost in noise but the available data rate sank and that it was a case of swings and roundabouts. You could do it, but it would be slow.

There is a second thing about noise which costs time. If you measure the level of noise and do it many times, you will get scatter in your results. The RMS of the scatter is the 'standard deviation' If you want lower scatter, you need the average of more results and this takes more time. For an analogue measurement, bandwidth can be considered to be the rate of arrival of measurement results. So you have a trade off. To reduce scatter standard deviation, either measure for longer, or measure a wider bandwidth.

So for a given bandwidth of noise, there is a trade-off between how long it takes to do a measurement and how much you can trust the result.

We once made an instrument to servo control the signal to noise ratio of a signal for testing digital microwave receivers. We had a fast and precise attenuator and a noise source being monitored by a slow and accurate thermal power meter. So we knew the long term average going into the attenuator, and we could shift attenuation quickly. So quickly that downstream of it, it was impossible to measure the noise to the level accuracy we thought we had! was the noise at the level it was because of the attenuator's activity, or was it a quiet or loud patch in the noise? Impossible to tell. So a signal you couldn't measure and there was a mathematical proof it was so.

At this point you begin to wonder what it all means!

It worked, though and did the wanted job. I'm just not sure how

Noise is interesting stuff.

We don't want it, we don't like it, but know thine enemy.

David

[Craig Sawyers]

Going up a few orders of frequency magnitude you get photon bunching and anti-bunching, which are correlation phenomena between two beams of photons. It is all a weird consequence of the quantum nature of light. I used to understand that stuff a few decades ago.

However, if this happens with light frequencies, it must happen at lower frequencies too, and might explain some of the observations in the microwave region that David was measuring.

Craig

[paulsherwin]

I think we need to get back on topic. This is all some way from RIAA preamps.

[bikerhifinut]

Paul, as a mod could you split the equally fascinating topic of the space noise stuff into a new thread?
And do you think perhaps we should alter this thread title to reflect the actual application my question was aimed at, i.e. chip based/silicon RIAA preamps? I think you are equally as interested as me on an optimal way of getting an affordable and not too complicated bang for ones buck?
regards.

Andy

[bikerhifinut]

This wee beastie actually does work quite well, its been commented on in the past.
I bought one in a maplins sale years ago and built it up for an acquaintance, I ran it off 2 pp3 in series and tweaked up the gain slightly to get a more usable signal into a modern line stage. That and swapping the electrolytic coupling caps for low voltage polywotsits along with the ceramic (ugh!) capacitors in the EQ network.
The basic idea seems similar to the VSPS to me.
http://www.velleman.co.uk/manuals/k2573.pdf

Sorry I couldn't split the circuit out of the PDF document. Its proably the portable device I am using.
A

[saddlestone-man]

I produced this list a few years ago for an article in the QRP Club's magazine Sprat. I think it's still relevant, and DIL versions of the very low (my definition) and ultra low (again my definition) devices are available on eBay.

Would be interesting to know if there are now any ultra-ultra low noise devices available at a reasonable price?

best regards ... Stef

[mhennessy]

Originally Posted by bikerhifinut

Sorry I couldn't split the circuit out of the PDF document. Its proably the portable device I am using.
A

Here you go, with annotations...

Mark

[bikerhifinut]

Originally Posted by saddlestone-man

I produced this list a few years ago for an article in the QRP Club's magazine Sprat. I think it's still relevant, and DIL versions of the very low (my definition) and ultra low (again my definition) devices are available on eBay.

Would be interesting to know if there are now any ultra-ultra low noise devices available at a reasonable price?

best regards ... Stef

Nice one stef,
Concise and easy to use.

Ah I see the AD797 is there, that seems to be the op amp that many DIY audio types rate for low noise low signal stuff.

Blimey £12.80 each from RS in 8 pin DIL …………….. I'll stick with NE5532/4 at around a quid a pop for one offs ...………...

A.

[Craig Sawyers]

Interesting list Stef - thanks for that. And it is relevant for low resistance sources where voltage noise is dominant. In audio, in particular with moving coil cartridges with a handful of ohms coil resistance, way sub-mV output and hardly any measurable inductance.

With moving magnet cartridges, with resistances in the mid hundreds of ohms and inductances of hundreds of mH the current noise is a critical parameter.

That is why the NE5534A is so good. It has just about the optimum combination of vn and in for the job. It can only be bettered by using a discrete front end with multiple, low rbb' transistors in parallel. And you only gain 1dB or so as compared to just the NE5534A.

It is just not worth the circuit design effort for 1dB.

Craig