John Curl's Blowtorch preamplifier part II

[elektroj]

Originally Posted by Esperado

OPA with offset.

Care to show it how it will work?

Originally Posted by Esperado

With 32 bits, you can attenuate by 60db without any audible artifacts. What the hell ?

Indeed ..

http://www.jamminpower.com/PDF/48-bit Audio.pdf

[PMA]

Anyone ever seen true 32 bits without errors, skip codes, glitches and uncorrelated spuriae? Real worls is somewhere at 20 bits, if excellent.

[Esperado]

Originally Posted by elektroj

Care to show it how it will work?

Nothing complicated: Applying a +2.5V *low noise* reference to the + input of the OPA (gain 1) , while you enter your signal to the -input.

[PMA]

.... yes, low swing (<5Vpp)

[Esperado]

Originally Posted by PMA

Anyone ever seen true 32 bits without errors, skip codes, glitches and uncorrelated spuriae? Real worls is somewhere at 20 bits, if excellent.

PMA, this 32 bits is not sampling rate, 24 bits is large enough. It is precision for calculation. Then rounded for 24 bits with dithering.
Anyway, i do not know so much AD converters in 32 bits, and i worry what kind of previous analog stage able to use all the bits, indeed :-)

[SY]

If 32 bit calculation is problematic, your computer won't work.

[Waly]

Originally Posted by PMA

For the reasons you have just described, I built a measuring amplifier with gain 40dB to be able to measure noise level of preamplifiers and make FFT analysis (noise level which is normally below common audio measuring system threshold) and to be able to measure at low output level. My SW computes unweighted integral noise in 20Hz - 20kHz band, so no tricky "noise bottom" like -150dB etc, which depends on bin width (number of samples in FFT, sampling frequency and possibly averaging as well). Noise of the measuring amp is attached. So I amplify the preamp output by 40dB, measure noise and ten subtract -40dB to get preamp's noise. I also measure distortion, at about 10 - 20mVrms of preamp output voltage. I have not measured any increase of harmonic distortion at low signal output level, but found differeces in noise of preamplifiers and its spectra.

With this method, I can measure noise level down to -126dB/ref 1Vrms (20Hz - 20kHz, unweighted). The best preamp I measured had noise -115dB/1Vrms unweighted, 20Hz - 20kHz. The gain of that preamp was 14dB.

I have to admit I am still in the dark with what exactly you are measuring. In particular the lack of 1/f noise (increasing by 10dB/decade with decreasing frequency) is intriguing.

What are the units on the Y axis? dB ref. 2V of what? It's certainly not the standard nV/rtHz. Could you draw the noise of a 1Kohm resistor on the same graph?

[Kindhornman]

Sy,
Don't we just daisy chain a bunch of 8086 chip until we get to 64 bits.........

[bcarso]

What's often interesting about datasheets is what is left out. After that, what the conditions of measurement and specification are.

I miss the characteristic curves in the old databooks for various transistor "processes". But I'd conjecture that people were referring to them and holding the manufacturer to them, rather than the guaranteed specifications for a particular part selected from that process.

[SY]

I use an abacus. None of these newfangled calculational methods.

Just the other day, I was having lunch with some of my colleagues, and we were talking at a level that people here wouldn't understand. There were questions about wire and bead materials and how they can impact the accuracy of accounting. But if you're happy with wood and steel, more power to you. We selected jade and nichrome after extensive testing, but that's the difference between a Bentley and a Yugo. Works for me!