the effect of mains in audio measurements

[metebalci]

I am playing with a QA403 (audio analyzer) and the effect of mains, 50Hz and harmonics, is very visible in the measurements. The only solution I could find is to power the DUT (PCB or final/enclosed product) with a battery. How are these tests (around -100dB THD+N) done in the industry ? Always with a battery, with a very linear power supply or what am I missing ?

[CaptDon]

Who publishes specs at -100dB except for power amplifiers where they state 100dB below full power output. One method would be to operate the U.U.T. through an isolation transformer with extremely low inter-winding capacitance (they do make them) and to have very good shielded leads for any external connections. You may still see real artifacts of the mains power due to the mains magnetic fields inside the U.U.T. being picked up in the preamplifiers and so forth. In that case 'it is what it is'. IMD and THD can be reduced to some acceptable level but face it, when you turn up the gain there is a noise floor you just can't beat. When speaking of what is 'tolerable' in music, If a passage is averaging around 0Vu with expected peaks of around 10dB above 0Vu and if the music source is stopped any residual better than -80dB lower than 0Vu is about as good as it gets. And then if you grab the volume control and crank it with no music source you're still going to bring up the shot noise hiss to an audible level. I think noise specs for consumer gear better than -100dB below reference are fake anyway. The Telcos years ago had a 'standard' they considered to be the 'noise floor' and they made their measurements as DBRN, decibels referenced to noise. 40dB was considered wonderful. There is a tradeoff also, usually shifting to a quieter signal below reference means sacrificing available headroom. And the opposite is true, allowing 20dB of headroom usually means less room between the noise floor and the reference. A 100 dB span is about all we get with typical audio gear.

[metebalci]

I think I actually take back my question. It seems the things I was testing before was actually not good in terms of power supply isolation. I am looking at another (not expensive) device now and there is nothing visible from the mains.

[radiolistener]

I think noise specs for consumer gear better than -100dB below reference are fake anyway. The Telcos years ago had a 'standard' they considered to be the 'noise floor' and they made their measurements as DBRN, decibels referenced to noise. 40dB was considered wonderful.

It is useless to talking about noise floor with no mention for which frequency bandwidth. Because noise floor depends on bandwidth.

If you're talking about 100 MHz bandwidth, then -100 dBFS looks not possible.
But if you're talking about 1 Hz bandwidth, then -100 dBFS noise floor is too noisy circuit.

Audio electronics standard is 20 kHz bandwidth.
Dynamic range from typical hearing threshold up to the max possible volume in the air environment at pressure 1 bar is about 195 dB, which corresponds to ideal 32 bit resolution DAC/ADC.
High-end audio electronics strives for the dynamic range 146 dB which corresponds to ideal 24 bit resolution DAC/ADC.

The real dynamic range of cutting edge audio technology is limited at about 120-135 dB for 32-bit DAC.

So, -100 dBFS noise floor is possible and is definitely not the best result.

[Coordonnée_chromatique]

operate the U.U.T. through an isolation transformer with extremely low inter-winding capacitance (they do make them)

Do you think that it can be usefull to reduce the magnetic flux (decrease the efficiency) in the transfortmer core ?

[radiolistener]

Do you think that it can be usefull to reduce the magnetic flux (decrease the efficiency) in the transfortmer core ?

For the best result you can power your equipment from a DC battery and keep it away from mains.

[CaptDon]

There is something to be said for testing 'in-situ' where it will all be connected as it is in the customers environment. Powering a piece of equipment from batteries and publishing those specs would be an outright lie if when powered from mains there was a huge hum problem from a power supply with huge ripple and stray voltages induced by magnetic fields that are picked up and amplified. I guess you could say one measurement would be "as good as it could possibly get" and the other measurement would be "as good as it actually is in your environment". 32 bit DAC audio is nice when the band and all recording equipment is in a Faraday cage and battery operated and surrounded by 3 foot thick concrete walls! Of course we never had that luxury even when recording Michael on the Skinner pipe organ for the Telarc series of disks, although you can hear the audible environmental noise above the electronic noise floor as would be expected with ADD recording methods. What is the dynamic range between threshold of hearing and eardrum rupture? Not 195dB!

[radiolistener]

32 bit DAC audio is nice when the band and all recording equipment is in a Faraday cage and battery operated and surrounded by 3 foot thick concrete walls!

it works very good with usual headphones. You can use even cheap Chinese USB dongle with 32-bit 384 kHz DAC which cost about 5-10 USD and it produce pretty nice sound, much better than usual 16-bit sound card. But cheap DAC's have about 100-120 dB SNR, if you want better, there is a sense to buy professional DAC with better frontend.

[metebalci]

There is something to be said for testing 'in-situ' where it will all be connected as it is in the customers environment. Powering a piece of equipment from batteries and publishing those specs would be an outright lie if when powered from mains there was a huge hum problem from a power supply with huge ripple and stray voltages induced by magnetic fields that are picked up and amplified.

I definitely agree with this when testing the final product and producing a spec for publishing. But I think battery etc. can be useful during development.

[metebalci]

it works very good with usual headphones. You can use even cheap Chinese USB dongle with 32-bit 384 kHz DAC which cost about 5-10 USD and it produce pretty nice sound, much better than usual 16-bit sound card. But cheap DAC's have about 100-120 dB SNR, if you want better, there is a sense to buy professional DAC with better frontend.

I wonder; since quite a silent room is ~30dB SPL, and considering maybe a 120 dB SPL transient, this is only 90dB. Is >100dB SNR a safety margin ?

[radiolistener]

I wonder; since quite a silent room is ~30dB SPL, and considering maybe a 120 dB SPL transient, this is only 90dB. Is >100dB SNR a safety margin ?

30 dB is not silent room, that is pretty annoying noise level.
Powerful 12cm fan for PC has about 25 dB noise.

Human ear can listen sounds at about 0 dB.
Gun shot has about 150-170 dB.

So if you want audio system which can realistically reproduce gun shot sound and at the same time has no audible noise, it should have at least 150 dB dynamic range, which currently is not possible with current audio technology.

But the things is much more complicated, because human ear has non flat frequency response, while DAC specification shows SNR for a flat response. It requires about 40 dB more dynamic range for DAC in order to cover full dynamic range of human ear for all frequencies.

And there is needs to take into account that audio system has sound volume level regulator and user can change it depends on the situation and high gain should not lead to audible noise. It adds additional margin requirements for DAC SNR.

In reality there is current technology limitations which limits DAC SNR at about 120-130 dB. So, at this moment, there is no ideal DAC which can cover full human ear dynamic range with no distortions.

But the mains noise can be filtered to the level at which it cannot be audible. I'm sure that 50 Hz hum sound from transformers, wires, electrical panels, etc has a higher sound level than noise floor which a good sound card can provide.