I want to measure the noise power supplies down to 1uv or less.
I also want to be able to measure phase noise of crystal oscillators.
I have never used a spectrum analyzer before so I don't know what specs I require for this task.
Trying to spend as little money as possible.
Any advice?
I also want to be able to measure phase noise of crystal oscillators.
I have never used a spectrum analyzer before so I don't know what specs I require for this task.
Trying to spend as little money as possible.
Any advice?
In that case you should consider a good USB-soundcard, ARTA and some PC setup for noise measurements.
Phase noise measurement is not trivial business. I suggest you study this thread.
Measuring power supply noise, by contrast, is relatively easy. Most of what you need is what an off-the-shelf balanced microphone preamp already does. Honestly, you could probably make an adapter box (with some coupling caps - especially for voltages outside 0...+48 V - and misc. protection to keep big transients off) and use an USB audio interface with a decent built-in mic pre plus software FFT... even a run of the mill -125 dBu input noise floor is just 0.43 µV after all. The Mackie 402 VL4 I recently bought is closer to -132 dBu at full gain.
Potential gotchas:
You need to carry out an absolute level calibration somehow, maybe using a soundcard output. I am guessing that you'll be more interested in relative changes anyway.
Amplifier bandwidth at high gains - where noise is lowest - may not extend an awful lot beyond 20 kHz. Using a prepre like a Cloudlifter and correspondingly reduced gain on the main pre may prove helpful. If in doubt, you could always roll your own instrumentation amp (e.g. THAT1512 or similar) based mic pre, these tend to have plenty of bandwidth.
Measuring power supply noise, by contrast, is relatively easy. Most of what you need is what an off-the-shelf balanced microphone preamp already does. Honestly, you could probably make an adapter box (with some coupling caps - especially for voltages outside 0...+48 V - and misc. protection to keep big transients off) and use an USB audio interface with a decent built-in mic pre plus software FFT... even a run of the mill -125 dBu input noise floor is just 0.43 µV after all. The Mackie 402 VL4 I recently bought is closer to -132 dBu at full gain.
Potential gotchas:
You need to carry out an absolute level calibration somehow, maybe using a soundcard output. I am guessing that you'll be more interested in relative changes anyway.
Amplifier bandwidth at high gains - where noise is lowest - may not extend an awful lot beyond 20 kHz. Using a prepre like a Cloudlifter and correspondingly reduced gain on the main pre may prove helpful. If in doubt, you could always roll your own instrumentation amp (e.g. THAT1512 or similar) based mic pre, these tend to have plenty of bandwidth.
I concur with voltwide's suggestion; a sound card could be useful for low-frequency spectrum analysis.
I recall that Jim Williams (legendary guru at Linear Technology) wrote some app notes (featured in EDN and/or Electronic Design) about measuring noise in power supplies.
From personal experience, measuring phase noise in oscillators is interesting and very challenging. Using a spectrum analyzer alone to measure crystal oscillator phase noise is likely a futile exercise--- the analyzer's Local Oscillator will probably have far higher nose than the the source under test. See link below. If you're up for developing your own instrumentation, you can make very sensitive measurements without breaking the bank. The sound card FFT analyzer could benefit close-in noise measurement.
Good luck!
https://www.keysight.com/upload/cmc_upload/All/5952-8286E.pdf
I recall that Jim Williams (legendary guru at Linear Technology) wrote some app notes (featured in EDN and/or Electronic Design) about measuring noise in power supplies.
From personal experience, measuring phase noise in oscillators is interesting and very challenging. Using a spectrum analyzer alone to measure crystal oscillator phase noise is likely a futile exercise--- the analyzer's Local Oscillator will probably have far higher nose than the the source under test. See link below. If you're up for developing your own instrumentation, you can make very sensitive measurements without breaking the bank. The sound card FFT analyzer could benefit close-in noise measurement.
Good luck!
https://www.keysight.com/upload/cmc_upload/All/5952-8286E.pdf
In regulators yes, at least through the audio range, some regs will even peak at hf.
But in an unregulated supply, mostly 120Hz and harmonics, decreasing with frequency.
Simple Voltage Regulators Part 1: Noise - [English]
Last edited:
Just so I'm clear. A sound card is good enough to measure noise on a regulator intended for audio oscillators and dacs?
In a word, yes.
I'd wager most engineers use common linear regulators (e.g. LM7815) and do little more than perform a perfunctory check with an oscilloscope to confirm the supply is clean. Many ICs (opamps,etc.) have power supply rejection properties (PSRR) that provide additional rejection of power supply noise.
I'd wager most engineers use common linear regulators (e.g. LM7815) and do little more than perform a perfunctory check with an oscilloscope to confirm the supply is clean. Many ICs (opamps,etc.) have power supply rejection properties (PSRR) that provide additional rejection of power supply noise.
What kind of noise measurement are you interested in? If you're looking for the noise spectral density versus frequency, you'll need a spectrum analyzer of some kind. If you're looking for the integrated noise across some bandwidth, you can do that with a bench top voltmeter such as the HP 34401A.
You should be able to get to 80 kHz bandwidth with a sound card and 192 kHz sampling frequency. Something like an HP 3562A will get you 100 kHz bandwidth. You can go with an actual spectrum analyzer (the HP 8700-series is awesome) but do beware that they really frown on DC on the input (it fries the input stage). Also note that many spectrum analyzers start at 30-50 kHz, so they're useless for audio.
Tom
You should be able to get to 80 kHz bandwidth with a sound card and 192 kHz sampling frequency. Something like an HP 3562A will get you 100 kHz bandwidth. You can go with an actual spectrum analyzer (the HP 8700-series is awesome) but do beware that they really frown on DC on the input (it fries the input stage). Also note that many spectrum analyzers start at 30-50 kHz, so they're useless for audio.
Tom
Or LM317/337... Also, as long as the end-to-end (blackbox) performance of the equipment is good, it doesn't really matter how low noise the supply is. If the performance meets spec, it's obviously good enough.
It's for debugging that an analyzer of sorts can be handy. I bet you'd be able to get pretty far with a bench top meter, oscilloscope, and a sound card.
Tom
Those are two separate problems.
For power supply noise measurement on a budget, I would suggest this pdf by a DIYAUDIO contributor:
https://cdn.shopify.com/s/files/1/0635/1487/files/noise_meas_amp_filter.pdf
There are some real, real noise measurement gurus on DIYAUDIO --
So I'm a bit confused.
I thought when people talk about the requirement of "low noise" power supplies on dac and XOs used in dacs they refer to ultra wide-band noise?
I thought when people talk about the requirement of "low noise" power supplies on dac and XOs used in dacs they refer to ultra wide-band noise?
At first we are talking about audio, not radio frequencies. Which ends at 20kHz. I do not exclude the possibility of audible interference between a sensitive DAC sampling frequency with some pwm clock ripple of smps but - but this can be filtered with some passive LC-filter in the supply rails effectively. So I do not see the need to specify broadband noise of supply in terms of uVrms.
You would need to ensure that your grounding, shielding etc. is good down to 1uV. Not easy! At this level many people end up 'measuring' the noise in their instrumentation.hellokitty123 said:
I suspect that in most cases they are simply repeating what they have seen others say. For a crystal oscillator the most important noise is low frequency, as that can make it around the oscillator feedback loop and create jitter. The main source of jitter in a typical digital audio crystal oscillator is the cheap crystal, not the power supply.
Interesting paper from Texas Instrument about low noise PSU measurments - SLYY076 How to measure LDO noise I think it covers most of your questions.
Noise, per se, seemed to be the least important of the factors affecting the influence of power supply regulation when I did the work for the LinearAudio article. What one of the correspondents pointed out was that the regulator's error amplifier was probably dominant as it can modulate the power supply rails. (So, I put a small signal on the power supply rail and looked at the harmonics, and indeed regulators like the Jung, Sjostrom etc which were ranked tops also had lower 2nd and 3rd harmonics.) You want a fast error amplifier, tight layout and high gain transistor.
I've used a variety of power supplies to test out the PassDIY Pearl II phono pre-pre and whether one uses an LM78/79xx or Jung, Sjostrom or Linear Tech regulator it still sounds great and very quiet.
Noise is a real problem for oscillators, nevertheless.
If you want to measure noise with a sound-card spectrum analyzer I would suggest building a front end with a low noise operational amplifier like the LT1028, AD797, SSM2019, THAT1580, operate it off batteries, and shield it in a Danish cookie tin.
I've used a variety of power supplies to test out the PassDIY Pearl II phono pre-pre and whether one uses an LM78/79xx or Jung, Sjostrom or Linear Tech regulator it still sounds great and very quiet.
Noise is a real problem for oscillators, nevertheless.
If you want to measure noise with a sound-card spectrum analyzer I would suggest building a front end with a low noise operational amplifier like the LT1028, AD797, SSM2019, THAT1580, operate it off batteries, and shield it in a Danish cookie tin.
Last edited:
HiJack! I do not expect much noise reduction with such pre-amp provided your soundcard includes a mic-input - nowadays these are pretty good. For low noise measurements I put the hole setup into an aluminium suitcase bought for little money in a hardware store.
Aluminum isn't much help -- the cookie tin is better -- you'll want to sand the edges of the lids so that it conducts -- Linear (before its marriage with Analog Devices) had these suggestions. Besides, the Danish cookie tin provides other benefits while being emptied.
https://www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf
https://www.analog.com/media/en/technical-documentation/application-notes/an124f.pdf
Jim Williams on Youtube: YouTube
https://www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf
https://www.analog.com/media/en/technical-documentation/application-notes/an124f.pdf
Jim Williams on Youtube: YouTube
Attachments
Last edited:
- Status
- Not open for further replies.