Pavel,
The clip leads are cheap off shore junk. Probaby 26 gauge or thinner. Used to be ten for $1. These days maybe $3.
But the preamp under test used a wiring harness with about 1/2 M per lead length of 24 gauge wire. That in addition to the battery protection circuit gets the battery source resistance to a few tenths of an ohm. So it forms an RC filter with the super caps. The time constant is more than 5 seconds so it is quite an effective filter. As mentioned these chips do not have great PSRR and with gain it drops even more.
Scott,
Commercially available wire usually runs at the minimum diameter to still meet specs. It also cheats a bit on copper purity. So for practical considerations you consider it 1/2 - 1 gauge smaller. I also just got pricing on copper clad aluminum loudspeaker wire, quite cheap.
Wire resistance is a big part of my size sound systems!
So I have to guess Scott is coming around to all discrete circuitry and silver wire.
Does that mean John will now have to go to all IC designs?
Last edited:
I think Ed is interested in more than accurately measuring the resistance but minimizing the added noise (which in the supply lead is not really added to the output except by horrible PSRR). He said the amp has a noise resistance of ~1 Ohm and the gain setting resistor is 15/24 Ohm giving 1.625 Ohm getting a battery and interconnect to a small fraction of that would seem trivial compared to some other problems.
Pretty much, except it also seems to affect the shot noise corner frequency. That is the actual issue among all the noise. My suspicions are the noise test fixture used was giving inaccurate data. There are two other items I am still looking at for a cause.
But I am on the west coast for a few days looking at a large venue sound system. So it will be a bit till I get the rest of the data.
But yes an RC filter on steroids does do a bit.
Scott,
I'll ignore the typo. The super cap issue was is it cheaper to use low noise batteries or super caps. As the lowest noise batteries at FeLiPO and they require a keeper circuit the super caps have the advantage at low voltages.
The second issue that piped up in testing was the çhange in 1/F noise.
So I'll look a bit into that. For future designs of infrequently used equipment I suspect my choice will be alkaline batteries and super caps.
So where does the battery start, the NIST study gives -200dB for alkaline D cells and < -180dB even at 10Hz. The amplifier is specified at 60dB min CMRR and 30Mhz BW at G = 20 so there is very little if any loss at 1kHz. That would still give -34dB or so at the output or .002nV compared with the expected ~2nV at the output without a filter.
BTW Ed did you read the whole data sheet, the noise corner there is at several kHz. Your parts must be exceptional.
Bah, it's just hardware. Run of the mill, boring stuff. I love the sextupole descriptor "polish the ends of the bunches"..cute.
edit..bet nobody saw the rack and pinion drive on the end of the vertical test dewar, eh??
The swimsuit one at the bottom of the page, much better "firmware", so to speak..
John
Scott,
I thought I mentioned that my conclusion is that battery noise really isn't an issue as you get more noise from connecting them than they make. What is the resistance of a typical spring loaded AA battèry holder with those nice bright shiny chrome like contacts, steel springs and riveted connections? A lot more resistance generated noise that you will get from a fresh battery.
So while it was a brilliant bit of work making those measurements the results will always be worse in real applications.
Same thing with data like PSRR in practice the inputs are not shorted right at the chip's input pins.
You can get close but never all the way.
Again I think the data sheet specs may have been wrong on PSRR but will try a few other issues including why the high corner frequency shows up in my tests done with one device. The corner seems to drop with more chips and more bypass capacitance.
Last edited:
Yup.
trying to get good current through oxide interfaces taint a good idea. Might as well pinch the garden hose..
John
If that is a metaphor it sounds very painful.
Who da thunk it you can screw up power supply wiring to make it the limit not the power supply. Next step shunt regulators.
Ewww. Not sure I like the way you think...
Well, leave it to Ed...
badaboom
Sometimes I'm too ashamed to cash the check...But not this time...
John
The AA batteries were measured in a cheap chinese holder of the kind
you described.
< http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf >
But you have have seen that, I have mentioned it in this thread already.
Walls measures even better than me, but he was paid for it. (but not
much longer, I have a XFFT analyzer, too
And the 1/f frequency does not sink with more amplifiers in parallel; it's the noise
level in toto that sinks, and the visibly lower 1/f frequency is only an artefact
because your world exists only above the sea level of the noise. You cannot see
that it falls even more.
See for example Enrico Rubiolas a, b coefficients in the 1/f noise of oscillators.
The 1/f corner makes only sense if you see _only_ the effect that produces it.
Your opamp has a PSSR of 60 or 70 dB IIRC. A LM317 is only 45 dB above 1nV/
sqrt Hz, so even that should be absorbed. I have found ST's data sheets quite
reliable, and 70 dB PSSR is not exactly outstanding, so I would accept it.
regards, Gerhard
Enrico Rubiola home page IIRC
Hey, RNMARCH,
are you in Kathmandu right now? If I go there again next November, what
do you think: will I be more part of the solution or part of the problem?
Last edited:
I think we are in almost complete agreement. Except I suspect it is the battery management card that is the source of the noise.
I will confirm that the 1/F noise is limited by the analyzer when I complete the second stage.
The item to note is that the super capacitors are less expensive than the Li batteries. So I probably will use them with alkaline batteries. (Their noise level does go up with use.)
I will confirm that the 1/F noise is limited by the analyzer when I complete the second stage.
The item to note is that the super capacitors are less expensive than the Li batteries. So I probably will use them with alkaline batteries. (Their noise level does go up with use.)
Thanks Gerhard for linking to Enrico Rubiola's work
attachment is from this presentation
http://www.rubiola.org/pdf-slides/2014T-IFCS--Cross-spectrum.pdf
George
attachment is from this presentation
http://www.rubiola.org/pdf-slides/2014T-IFCS--Cross-spectrum.pdf
George
Attachments
I can imagine a 1/f look alike affect where the finite power supply source impedance interacts with local decoupling. Adding the super cap will lower the apparent noise corner However, this effect is PSRR related and not intrinsic input stage 1/f noise.
It seems to be lost on a lot of folks that if your PSU output Z is low (eg from high feedback), then the PSU noise which is usually wideband, will also be from a low source Z - increases in Z due regulator loop gain reduction with increasing f notwithstanding.
Have you tried adding a 1 or two ohm resistor in series with the supply before the super cap Ed?
It seems to be lost on a lot of folks that if your PSU output Z is low (eg from high feedback), then the PSU noise which is usually wideband, will also be from a low source Z - increases in Z due regulator loop gain reduction with increasing f notwithstanding.
Have you tried adding a 1 or two ohm resistor in series with the supply before the super cap Ed?
Hi... Yes, I am here in Kathmandu right now. You would be very much a part of the solution. The people in the country side are really hurting and suffering really badly. Leaving some money behind here helps a lot. The people are so kind and innocent and beautiful spirits/souls ---- pls come.
-Richard
Li batteries die if you overcharge them or over discharge them. During chargeing you have be sure all the cells get to the same voltage. So there are specialty chips to do that. I bought completely built out circuit cards for under $2 to handle four cells.
You do have to charge the batteries before first use. They then shut them down when the voltage drops and also when fully charged. But it seems they make a bit of noise. Haven't got exact numbers still building the rest of what is needed for a complete packaged low noise preamp.
- Status
- Not open for further replies.