OP is clearly looking for capability outside what he can get from a soundcard. He's chasing after power supply noise. Is 16bits at >100KHz really necessary for the application?
Hi ! yes exactly. I am still confused about how valuable this kind of investigation is but I found an article about smps testing where they were using a SA up to 20 MHz.
I also see in psu datasheet that ripple is referred up to 20 MHz. from this my curiosity to go beyond 100kHz
Taken from a datasheet of a psu ...
Cheap if not exactly calibrated way to measure this sort of HF stuff is just to buy a HF capable SDR receiver, something like Cross Country Wireless SDR-4+ general coverage receiver
It is no substitute for a real SA, but it is cheap and will do for this sort of investigation work (~850Khz - 70MHz).
Note the 50 ohm input however, but you will find that on all the SAs as well.
Basically the thing mixes the RF down to the audio band and feeds it to your soundcard as an I/Q pair for analysis using something like HPSDR or Linrad (Linrad is not particularly user friendly, but does have all the calibration options for doing real measurements).
It is no substitute for a real SA, but it is cheap and will do for this sort of investigation work (~850Khz - 70MHz).
Note the 50 ohm input however, but you will find that on all the SAs as well.
Basically the thing mixes the RF down to the audio band and feeds it to your soundcard as an I/Q pair for analysis using something like HPSDR or Linrad (Linrad is not particularly user friendly, but does have all the calibration options for doing real measurements).
Hi ! thanks a lot indeed for the very interesting advice.
Let me explain a little why I am so interested about the MHz range.
I am 100% digital. And I have no plan to go back to vinyl.
Nevertheless analog audio has always been pleasant to me. Maybe not perfect but always nice. I would call it musical.
Much much different experience with digital. Sometime has been unbearable. And some other fatiguing. Very very few times musical like analog.
Then I read some comments about the opinion that this digital fatigue could come from the extreme sensitivity of digital to RFI, that are in the MHz region.
From this I think that digital should be tested very well also in the MHz region to spot some possible noise/interference that in some way can mix down with the signals (increasing for instance jitter or clock stability).
I have an old idea to try batteries to power dac chip and clock.
If I had to build a dac I would separate galvanically digital and analog circuits and use batteries to power the digital section.
Using batteries and a very good shielding RFI should not enter the digital section.
Instead RFI can enter from the mains, the cabling and of course the power supply.
Thanks a lot again.
Hi !
i wonder how many in this forum own a spectrum analyzer.
If not how can they test their projects ?
The more i read the more it looks to me as an essential tool for any serious analysis of audio circuit.
Can you really live without it ?
i wonder how many in this forum own a spectrum analyzer.
If not how can they test their projects ?
The more i read the more it looks to me as an essential tool for any serious analysis of audio circuit.
Can you really live without it ?
I use a picoscope from the 3000 series . If you are interested in measuring digital signals/MHz bandwidth then you might look here. It's not cheap but still in range and really works. I had an analyzer from Velleman before I got the picoscope and it croaked one day without any warnign or outer reason (cannot be compared to picoscope at any level).
https://www.picotech.com/oscilloscope/3000/usb3-oscilloscope-logic-analyzer
https://www.picotech.com/oscilloscope/3000/usb3-oscilloscope-logic-analyzer
Hi Gino,
I have three spectrum analyzers designed for different jobs. Low frequency models are often called signal analyzers. Spectrum analyzers may go to 40 MHz or more (Tuner IF work) and higher like 500 MHz + (RF amplifier work).
These instruments normally have a maximum input voltage of 5 VDC -+ ACV peak into 50 R. To protect these instruments, you should use them behind some other equipment, like THD analyzer. This is what I do often. My stuff is all HP/Agilent/Keysight. These are often more plentiful on the used market and tend to hold their calibration better than most other ones. I use an HP 339A THD meter that has an output for the residuals after the main frequency has been eliminated/reduced in level.
The most important instruments for your bench are a good bench meter, a good oscilloscope, low distortion audio oscillator and THD meter. Add-on equipment might be the spectrum/network analyzer, RF oscillator that can be modulated, power supplies and a decent LCR meter. You really want to know about dissipation (or quality factor) rather than ESR. A frequency counter is another useful item. More voltmeters, you almost can't have enough of these. The hand held ones are good for working with high voltages. There are a number of useful instruments that might fit you specifically, depending on the work you want to do as well. Selective voltmeters are great for setting traps for 19 and 38 KHz in FM stereo demodulators for example. Clamp on DC or AC meters are quite handy to have also. The HP 428 is an excellent example of this. Don't forget various probes and cable adapters. Your own jigs that make some jobs easier will add to the list. But your equipment needs are personal beyond the basic set mentioned earlier.
-Chris
I have three spectrum analyzers designed for different jobs. Low frequency models are often called signal analyzers. Spectrum analyzers may go to 40 MHz or more (Tuner IF work) and higher like 500 MHz + (RF amplifier work).
These instruments normally have a maximum input voltage of 5 VDC -+ ACV peak into 50 R. To protect these instruments, you should use them behind some other equipment, like THD analyzer. This is what I do often. My stuff is all HP/Agilent/Keysight. These are often more plentiful on the used market and tend to hold their calibration better than most other ones. I use an HP 339A THD meter that has an output for the residuals after the main frequency has been eliminated/reduced in level.
The most important instruments for your bench are a good bench meter, a good oscilloscope, low distortion audio oscillator and THD meter. Add-on equipment might be the spectrum/network analyzer, RF oscillator that can be modulated, power supplies and a decent LCR meter. You really want to know about dissipation (or quality factor) rather than ESR. A frequency counter is another useful item. More voltmeters, you almost can't have enough of these. The hand held ones are good for working with high voltages. There are a number of useful instruments that might fit you specifically, depending on the work you want to do as well. Selective voltmeters are great for setting traps for 19 and 38 KHz in FM stereo demodulators for example. Clamp on DC or AC meters are quite handy to have also. The HP 428 is an excellent example of this. Don't forget various probes and cable adapters. Your own jigs that make some jobs easier will add to the list. But your equipment needs are personal beyond the basic set mentioned earlier.
-Chris
Hi !
thanks a lot to You all for the very helpful advice.
I am still in the gathering advice phase.
I understand that the first decision is about the needed bandwidth ?
I am only interested in dacs.
So I wonder which would be a good bandwidth to investigate dacs performance.
Maybe 20 MHz ? more ?
I understand that dacs are very sensitive to RFI coming with the power supply and from the space around the chips.
But I do not know up to which frequency.
I understand the selection of bandwidth and precision is always a compromise.
And that to get both high bandwidth and precision is very very expensive.
What specs would you recommend to test dacs ?
Thanks a lot again.
Kind regards, gino
thanks a lot to You all for the very helpful advice.
I am still in the gathering advice phase.
I understand that the first decision is about the needed bandwidth ?
I am only interested in dacs.
So I wonder which would be a good bandwidth to investigate dacs performance.
Maybe 20 MHz ? more ?
I understand that dacs are very sensitive to RFI coming with the power supply and from the space around the chips.
But I do not know up to which frequency.
I understand the selection of bandwidth and precision is always a compromise.
And that to get both high bandwidth and precision is very very expensive.
What specs would you recommend to test dacs ?
Thanks a lot again.
Kind regards, gino
If VHF is a problem because it causes artifacts in the audible range, then why not just measure in the audible range to see if that's the case? (Spoiler: you're chasing a ghost again)
Hi and thanks a lot for the kind effort of injecting some sanity in my mind 😱
You are right. I should see those artifacts in the audible range.
I am currently testing/comparing different power supplies, both smps and linear ones.
I get very similar same noise floor graphs up to 96kHz (checked using a software SA).
And still I like the sound with linear power supplies better. Significantly better actually.
In particular i hear a cleaner sound with better rendering of ambient noises embedded in the recordings (like people chatting in a live recording in a club and so on).
From what I understand this is one of the main benefits from a low noise playback system (with the other very important quality being low distortion).
This is very important to me because I could not even need a proper spectrum analyzer (however I like the instrument immensely).
For now I am replacing parts in some psu kits I bought, like caps and diodes.
Just to get used to soldering and de-soldering parts on a pcb.
The stock caps are generic/unheard brands.
I am using mostly Nichicon and Panasonic 105C.
I am waiting for HER508, SB5100 and SR5100 parts I bought.
But as I said I like the sound using linear supplies better, even if in terms of noise the result is very very similar.
Maybe just a little better at lower Hz with linear ones.
Or maybe I have just that ghost on my back smiling at me ... 🙁
Thanks a lot again, gino
Last edited:
I didn't read the entire thread so hopefully I'm not repeating: Certain modulation effects (in particular jitter and noise/ripple of the converter reference voltage) are only fully seen if the DAC is operated with a signal. This is a demanding test as the analyzer should have high (higher than the DUT) dynamic range. Using sinusoidal drive and an analog notch filter with FFT of the residual is a good basic approach.
Samuel
Samuel
Hi ! thanks a lot for the very interesting advice.
My problem is very basic.
To understand if a linear supply is on principle better than a smps to power a dac/soundcard.
To explain a little better i have a usb soundcard under testing.
With a software i can see the spectrum of the noise floor up to 96kHz.
Passing from a decent smps (a 12V genuine power adpater for laptop) to a kit of a linear power supply based on a LT1083cp the spectrum looks extremely similar.
Practically identical.
Still the sound with the linear kit is better.
I can hear more clearly the details in the recordings.
For instance the "live" feeling is much more evident.
Question: is this possible or i am hearing things ?
As these kits are made with extremely cheap parts my next step would be to replace as a start caps and diodes and listen.
But i like the sound better with the linear power supply.
Maybe there is something i cannot see only with the noise floor test.
For instance noise beyond 96kHz can have an impact on jitter 😕
But i cannot see beyond 96kHz. From this all the questions about a proper spectrum analyzer.
From what i read a smps should be much more noisy than a linear psu above 100kHz. But i cannot be sure of this without a proper instrument.
Thanks a lot again, gino
Last edited:
P.S. in the web i have found that the best linear psu can have a wideband noise (0-20MHz) so low that it is difficult to measure.
This is never the case with smps instead. Some are very noisy and others just noisy.
This is never the case with smps instead. Some are very noisy and others just noisy.
Hi ! sorry but i am slow to arrive at the point.
I can only see the noise floor up to 96kHz.
but what about the jitter for instance. The distortion ?
I cannot say anything about jitter and distortion.
I am pretty sure that with the linear psu the sound is better.
So that i am abandoning completely the idea of using smps with digital units, even if they are very convenient and practical.
And moreover i am thinking to replace all the smps inside equipment i already own with linear ones.
I guess that jitter measurements are not easy to carry out.
Thanks a lot again.
Last edited:
No, but as you've been told again and again, if there's significant jitter in a significant place in the circuit, it will be evident in the analog output.
Hi Gino,
A Signal Analyzer would show that clearly. You need high dynamic range as SY pointed out, and fast update times along with more "bins", dividing the frequency span into smaller sections. You might be able to do this with a really good sound card and software for your computer.
If you wanted track down the exact source of the interference, a higher bandwidth spectrum analyzer might be the way to go. Look at what the smallest division of frequency bandwidth for the video filter. For comparison, the HP 3580 has a 1 Hz filter (!!! excellent) whereas the HP 3585A has a 3 Hz filter. Still pretty good, but nothing can equal an HP 3580 up to 51 KHz (or is it 104 KHz?). That's an old instrument in case you try and look it up.
-Chris
A Signal Analyzer would show that clearly. You need high dynamic range as SY pointed out, and fast update times along with more "bins", dividing the frequency span into smaller sections. You might be able to do this with a really good sound card and software for your computer.
If you wanted track down the exact source of the interference, a higher bandwidth spectrum analyzer might be the way to go. Look at what the smallest division of frequency bandwidth for the video filter. For comparison, the HP 3580 has a 1 Hz filter (!!! excellent) whereas the HP 3585A has a 3 Hz filter. Still pretty good, but nothing can equal an HP 3580 up to 51 KHz (or is it 104 KHz?). That's an old instrument in case you try and look it up.
-Chris
Hi and thanks a lot for the very helpful advice.
I am just starting and i intend to do all the possible measurements provided by the software
But i am not sure to have a really good soundcard
Thanks again for the advice. I will try some distortion measurements, at first with the soundcard i have. Just to understand the procedure.
I guess that a distortion measurement could give information also on the overall noise of the soundcard ?
i really have to study more seriously this software Arta.
- Status
- Not open for further replies.