The plots appear to have a frequency point every 6 Hz. The noise is about -108 dBV at 1 kHz. Assuming the noise bandwidth to be 6 Hz, that's -115.8 dBV in 1 Hz, or 1.625 μV/√Hz. Divided by the gain of about 31.6 at 1 kHz, that's 51.4 nV/√Hz equivalent at the input, including the terms due to the amplifier's noise voltage and current and cartridge thermal noise. It is actually higher than I would expect.
I tried finding out about how Pico treats the FFT, and from their forum I read that indeed you can't use their FFT analysis for noise measurements. Apparently it's optimized for plotting spectrums of signals because they apply Coherent Gain to the window function. No idea what it means, other than I can only use the FFT to look at the level of the 50 Hz signal and the junk at the top-end. source: FFT analysis using Picoscope Anyhow, the SNR (apart from the 50 Hz hum and top-end junk?) is baked into the design, and is a given. I don't have any 100 Hz issues.
I will report back after running the pre-amp from 2 x 9V batteries.
When I look at the left parts of the plots, they look like collections of straight line segments between points that correspond to frequencies of 12 Hz, 18 Hz, 24 Hz and so on. Sometimes a line segment appears to span 12 Hz rather than 6 Hz, but that can mean that two adjacent line segments happen to have nearly the same slope.
That's absolute nonsense, but you have to know what you are doing.
Seeing spectral lines around 10Khz, doesn't sat they are indeed around 10Khz because you restricted the spectrum to 24kHz.
When using a much higher frequency range instead of 24Khz , these spectral lines won't fold back in the spectrum below 24Khz when indeed at a much higher frequency.
But you are so close to the Picoscope's own noise, that it would be wise to add another amp with at least 20dB gain behind your phono preamp and to measure behind this second amp.
Then use a much higher bandwidth like 100Khz and see where the peaks are now.
When magnifying the spectrum to the range up to 20Khz, you will get a true noise spectrum of your amp.
When knowing the frequency bin width, the noise per rtHz can be calculated and compared to the simulation.
Hans
P.S. Yes you have used a regulated PS, that would do in this case.
Is it safe to assume that the Picoscope is always sufficiently dithered by its own noise?
Was everything connected with 1:1 probes and/or coaxial cables or with 1:10 probes?
Was everything connected with 1:1 probes and/or coaxial cables or with 1:10 probes?
Quite. Measuring noise from an FFT is by far the best way to do it, but getting a correct result is not trivial.
Picotech is one of the most advanced digital PC scope manufacturers, so one may assume they fully know what they are doing.
That’s a good question.
Arjen should check wether the switch on his probe is in the 1:1 position.
The +/-50mV setting the screen showed is for the 1:1 setting.
Hans
Arjen's Pico 2000 scope with a BW of 10MHz has a specified noise of 150uV.
These scope's figures calculate into 150e-6/sqrt(10MHz) = 47.5nV/rtHz or -146,5dBV for a 1Hz filter bin.
When accidently having the probe in the 1:10 position, this will display as -126.5dBV.
Using a 6Hz filter bin as Marcel found out, raises the level by 7.8dB resulting in 118.7dBV.
So this is almost the -116dBV that Arjen showed.
Ergo, the switch on the probe seems in the wrong position.
When translating the simmed version from nV/rtHz into dBV fot a 1Hz filter bin, we get the figure in the attachment.
At 20Khz, noise is ca.-150dB with shorted input, so to stay at least 10dB above the Scope's -146,5dBV noise, Arjen should add a 20dB amp behind the Phono preamp to get a representative noise spectrum.
Of course when using a wider filter bin width, the displayed noise will raise accordingly, but the ratio between preamp's and scope's noise will stay the same.
Hans
These scope's figures calculate into 150e-6/sqrt(10MHz) = 47.5nV/rtHz or -146,5dBV for a 1Hz filter bin.
When accidently having the probe in the 1:10 position, this will display as -126.5dBV.
Using a 6Hz filter bin as Marcel found out, raises the level by 7.8dB resulting in 118.7dBV.
So this is almost the -116dBV that Arjen showed.
Ergo, the switch on the probe seems in the wrong position.
When translating the simmed version from nV/rtHz into dBV fot a 1Hz filter bin, we get the figure in the attachment.
At 20Khz, noise is ca.-150dB with shorted input, so to stay at least 10dB above the Scope's -146,5dBV noise, Arjen should add a 20dB amp behind the Phono preamp to get a representative noise spectrum.
Of course when using a wider filter bin width, the displayed noise will raise accordingly, but the ratio between preamp's and scope's noise will stay the same.
Hans
Attachments
Hans, how do you explain the differences between the first and second numbered plots of post #137, https://www.diyaudio.com/community/threads/load-capacitance-on-mm-cartridge.426083/post-8013556 ?
That's most likely because of folding back of content at higher frequencies.
The Picoscope has no brickwall filter at 1/2*Fs in fact no filter at all.
Hans
Last edited:
Then instead of building a 20 dB amplifier, a simpler test would be to make sure that the probes are 1:1 and to remeasure the first two numbered graphs with a much higher sample rate. If the aliasing hypothesis is right, the results should look better then.
I must say I'm a bit sceptical about the aliasing hypothesis, because the phono amplifier's output noise density is supposed to drop with frequency and should keep dropping up to a very high frequency thanks to the passive filter at its output. That is, there is not much to alias. Then again, I don't have a better hypothesis.
I must say I'm a bit sceptical about the aliasing hypothesis, because the phono amplifier's output noise density is supposed to drop with frequency and should keep dropping up to a very high frequency thanks to the passive filter at its output. That is, there is not much to alias. Then again, I don't have a better hypothesis.
I did not understand the spectra either, right from the beginning.
Hans
P.s. As long as we don’t see a Riaa curve in the noise spectrum with shorted input, there is no guarantee that everything is o.k.
Last edited:
This is now turning into a measurement seminar for me. 🙂
Some prelimenary remarks / questions:
All measurement were carried out with a 1:1 probe.
Regarding bin-width: I can set the FFT Bandwidth Window and Number of Bins in the PicoScope software. Can I determine the Bin Width from those 2 variables?
Some prelimenary remarks / questions:
All measurement were carried out with a 1:1 probe.
Regarding bin-width: I can set the FFT Bandwidth Window and Number of Bins in the PicoScope software. Can I determine the Bin Width from those 2 variables?
I will try that. I also have a -40 / -20 / 0 / +20 dB attennuator / amplifier to try out.
I will try to get that measurement as well, to qualify the measurement set-up
Yes you can. You can also display all settings in a separate menu at the right.
That way we can see what exact settings you are using.
Also the bin width will be displayed.
Just as a precaution to check your probe setting.
You supply is 18V DC.
When you measure this with your scope in the DC setting, does the display show 18 Volt?
Reason I ask is because some probes when switched in the 1:1 position are still sticking to 1:10.
Switching the slider firmly a few times from one position into the other mostly cures this problem.
Hans