Thanks Hans, this is what I thought I saw. I'll have a look through my Picoscope measurements to see if they are any better. Otherwise, there is no use in trying to get a meaningful SNR measurement with either of these set-ups. A last option (if I want) is to change the head-amp to +30 or + 40dB, and see waht that brings. Otherwise, I can only use this set-up if I want to see if I can get rid of the 15-20 kHz junk (and the 50 Hz hum), once I know that this isn't aliasing I am looking at. My thoughts are that the turntable (which is direct drive) is generating noise, or that my tone-arm grounding leaves room for improvement. Having said that, the record's own groove noise is much higher than what is coming from the RIAA pre-amp / cart combo.
Previously I was using the Picoscope in a valve environment to trim phase-inverters and balance cathode currents on PP outputs to minimize 2nd harmonic distortion, for which it worked well, since the distortion levels are much higher than the noise floor.
Previously I was using the Picoscope in a valve environment to trim phase-inverters and balance cathode currents on PP outputs to minimize 2nd harmonic distortion, for which it worked well, since the distortion levels are much higher than the noise floor.
Last edited:
Is the annotation "Soundcard + 20 dB head-amp, input soundcard shorted" correct? It seems more logical to short the 20 dB preamplifier input.
At 1 kHz:
-112 dBV with shorted phono preamplifier input and 20 dB of extra amplification
-125 dBV with (presumably) shorted 20 dB amplifier input
Subtracting noise powers: 10 dB log10(10-11.2 - 10-12.5) ~= -112.2233 dBV with shorted phono preamplifier input and 20 dB of extra amplification, corrected for measuring equipment noise
Corrected for the extra 20 dB: -132.2233 dBV at the phono preamplifier output with shorted phono preamplifier input
44.1 kHz sample rate, DFT length 65536 samples, Hann window, presumably scaled such that a tone in the middle of a bin is shown with the right voltage -> noise bandwidth 1.5 * 44100 Hz/65536 ~= 1.00937 Hz, I think
Noise density at the phono preamplifier output when its input is shorted: 10-132.2233/20 V/sqrt(1.00937 Hz) ~= 243.67 nV/sqrt(Hz)
Equivalent input noise voltage density: 243.67 nV/sqrt(Hz)/1030/20 ~= 7.7057 nV/sqrt(Hz)
Theoretical estimate: thermal noise of 953 ohm plus thermal noise of 100 ohm plus op-amp voltage noise: 6.4844 nV/sqrt(Hz)
All in all, the measured noise with shorted input is close to the calculated noise with shorted input. It would be interesting to see how much worse it gets with the real cartridge connected again.
At 1 kHz:
-112 dBV with shorted phono preamplifier input and 20 dB of extra amplification
-125 dBV with (presumably) shorted 20 dB amplifier input
Subtracting noise powers: 10 dB log10(10-11.2 - 10-12.5) ~= -112.2233 dBV with shorted phono preamplifier input and 20 dB of extra amplification, corrected for measuring equipment noise
Corrected for the extra 20 dB: -132.2233 dBV at the phono preamplifier output with shorted phono preamplifier input
44.1 kHz sample rate, DFT length 65536 samples, Hann window, presumably scaled such that a tone in the middle of a bin is shown with the right voltage -> noise bandwidth 1.5 * 44100 Hz/65536 ~= 1.00937 Hz, I think
Noise density at the phono preamplifier output when its input is shorted: 10-132.2233/20 V/sqrt(1.00937 Hz) ~= 243.67 nV/sqrt(Hz)
Equivalent input noise voltage density: 243.67 nV/sqrt(Hz)/1030/20 ~= 7.7057 nV/sqrt(Hz)
Theoretical estimate: thermal noise of 953 ohm plus thermal noise of 100 ohm plus op-amp voltage noise: 6.4844 nV/sqrt(Hz)
All in all, the measured noise with shorted input is close to the calculated noise with shorted input. It would be interesting to see how much worse it gets with the real cartridge connected again.
Last edited:
Marcel,
The 243nV/rtHz@1kHz corresponds perfectly with the simulation in #119.
And with the 30dB gain this calculates of course into 7.7nV/rtHz@1kHz EIN.
It shows again how reliable noise prediction can be with LTSpice and proves that the Amp works perfectly.
Arjen got the spectrum (almost) right in the end after earlier attempts that were way off, bravo.
For the perfectionist a bit more gain from the second amp would also have shown the spectrum’s high end correctly
Hans
The 243nV/rtHz@1kHz corresponds perfectly with the simulation in #119.
And with the 30dB gain this calculates of course into 7.7nV/rtHz@1kHz EIN.
It shows again how reliable noise prediction can be with LTSpice and proves that the Amp works perfectly.
Arjen got the spectrum (almost) right in the end after earlier attempts that were way off, bravo.
For the perfectionist a bit more gain from the second amp would also have shown the spectrum’s high end correctly
Hans