Article: Understanding Digital Audio Measurements

[rkt31]

@amirm my question, what is the difference between dynamic range, sinad and linearity and how are they measured ?. i see dynamic range is more than sinad.

 

[pozz]

@amirm my question, what is the difference between dynamic range, sinad and linearity and how are they measured ?. i see dynamic range is more than sinad.

SNR is the ratio of the fundamental's amplitude to noise. SINAD is the ratio of amplitude to all noise, spurs and harmonics. Linearity is how accurately the device can reproduce levels.

 

[trl]

If SNR is the difference between the highest level that can be reproduced (with a low THD) and the background noise while playing, then "Dynamic range (abbreviated DR, DNR,[1] or DYR[2]) is the ratio between the largest and smallest values that a certain quantity can assume" (source: https://en.wikipedia.org/wiki/Dynamic_range).

"The dynamic range of an audio signal is the difference between the loudest (high level) and the softest points (low level) of a sound as illustrated by the diagram above":

Source: https://www.researchgate.net/figure...between-the-loudest-high-level_fig8_307991686
​

That means that SNR will be represented by a higher number than Dynamic Range (usually, the lowest level sound that is reproduced is still having a higher amplitude as the background noise).

A small addendum to pozz's answer regarding Linearity: deviation from the axis should be within +/-0.5 dB, per latest Amir's recommendation. Usually the scale is between 0 dB (max. level a DAC can reproduce) and lowest level (usually close to -120 dB).

As for the SINAD (SIgNal Over Noise And Distortion), well this is basically -(THD+N). So, if THD+N = -115 dB, the SINAD would be 115 dB.

I'm sure there're plenty of documentations and Internet resources to read about this topic.

 

[Electrical]

I joined for the express purpose to say thank you to amirm, to all contributors really, but amir specifically. Thank you!

Clearly this work is a labor of love and passion for audio, but it is still that: work. I wish to express my appreciation with hopes it plays a small part to encourage continued efforts. Also a gentle reminder of the unseen positive impact to thousands of random, unregistered readers who are educated and steered to good buying decisions.

Rarely do you have someone with technical skills combined with the gift of incredibly clear and concise communication, everything you need and nothing you don't, as they say. More so, the drive to share such a combination with the world. And now I will slip back into the random shadows and quietly enjoy this brilliant display.

 

[aj625]

@Amir, please review DCS Bartok and chord tt2 with m scaler.

 

[amirm]

@Amir, please review DCS Bartok and chord tt2 with m scaler.

I am ready and willing. We just need to find their owners willing to loan them for testing. So spread the word!

 

[tvrgeek]

I know it would be difficult to make up a standard test, but in addition to the excellent basic electrical tests, epically with DACs ( anything containing them) an additional "compatibility" test may be in order.

I say this after beating my head against the desk for several days trying to get the D30pro driver to work with the SF ASIO plugin for WMP. This is not even very high performance compared to what these things can do. Many drivers, many OS's, many proprietary.

This is also because the lack of, and often incorrect information out there. For instance, I am pretty sure I was able to run ASIO via the CM driver to my Schiit Asgard. Schiit says nothing about it and a couple of folks said ASIO required hardware compatibility.

Many out there are old hats at this stuff and are not confused, but coming in cold, this is a horrible mess of incompatibilities, patches and hacks. Not everyone is a programmer. ( I only had to hire them and turn them lose) To make it more fun, it is a moving target so what was true last year, may not be now.

 

[Ingenieur]

Not sure if this has been shared but imo an excellent primer on DAC's.. written by Analog Devices 20 years ago.

https://www.analog.com/media/en/training-seminars/design-handbooks/Technical-Tutorial-DDS/Section4.pdf

 

[Plcamp]

I have some questions about modern DACs.

- Is it generally true that you can safely assume a modern DAC like a Topping e30 would have dc isolation on its outputs such that it could driver a power amp with its front end isolation cap removed?
- Is it true that output impedance is stable vs frequency?

 

[Rene]

I have some questions about modern DACs.

- Is it generally true that you can safely assume a modern DAC like a Topping e30 would have dc isolation on its outputs such that it could driver a power amp with its front end isolation cap removed?
- Is it true that output impedance is stable vs frequency?

Not at all true. Many dacs have no DC blocking.

 

[Phylos]

I'm sorry, how do we get that chart?

 

[pos]

Not sure this is the right thread, but here is a suggestion for @amirm regarding balanced DAC measurements.

The APx555 analog inputs have SOTA CMRR figures, but that might not necessarily be the case for the device the DAC will eventually be connected to.

For example its CMRR sits between 80dB and 90dB under 2.5V, whereas an Hypex nc400 amp has a 55dB figure there.
This means that potential artifacts that do cancel out properly on the APx inputs (eg DC offset, or maybe some more intrinsic things like the famous ESS hump) might not do so to the same extend on the end user device.
This is especially important when considering manufacturers will typically specify, measure and adjust there product using an APx or similar device, so they might be taking a good CMRR for granted.

Would it be possible to do a few additional measurements (eg SINAD and IMD vs level) by either measuring each leg independently, or using a balanced to unbalanced converter with a low CMRR (and otherwise good specs) in front of the APx?
I would also love to see DC offset measurements in general, as I think these might have more impact in practice than going from a 110dB to a 120dB SINAD.


Same goes (to a much lesser degree) for the input independence, where the APx might be more forgiving to a low output DAC than an amplifier with a higher input impedance.

 

[pilau]

I have some questions about modern DACs.

- Is it generally true that you can safely assume a modern DAC like a Topping e30 would have dc isolation on its outputs such that it could driver a power amp with its front end isolation cap removed?
- Is it true that output impedance is stable vs frequency?

These are such great questions. Have you figured them out?

 

[Plcamp]

These are such great questions. Have you figured them out?

I can’t say I have figured them out.

I suspect that I could in fact remove the isolation cap at the front end of my Adcom 535II power amp, because it has a LF servo, and any offset from the DAC would be easily corrected, if it even exists.

I was asking the output impedance stability because I am considering passive line level crossovers (enabled by the very high input impedance if the Adcom)…have no reason to suspect they are not stable.

Such things aren’t measured nor specified, so it’s impossible to generalize I guess.

 

[pilau]

I can’t say I have figured them out.

I suspect that I could in fact remove the isolation cap at the front end of my Adcom 535II power amp, because it has a LF servo, and any offset from the DAC would be easily corrected, if it even exists.

I was asking the output impedance stability because I am considering passive line level crossovers (enabled by the very high input impedance if the Adcom)…have no reason to suspect they are not stable.

Such things aren’t measured nor specified, so it’s impossible to generalize I guess.

Shouldn't DACs ensure they're not letting DC out?? Or am I crazy?

 

[Rene]

I suspect that I could in fact remove the isolation cap at the front end of my Adcom 535II power amp, because it has a LF servo, and any offset from the DAC would be easily corrected, if it even exists.

I was asking the output impedance stability because I am considering passive line level crossovers (enabled by the very high input impedance if the Adcom)…have no reason to suspect they are not stable.

Such things aren’t measured nor specified, so it’s impossible to generalize I guess.

Removing the isolation (or blocking) cap from the input of the ADCOM 535 and connecting it to a DC coupled output from a DAC will disable the ADCOM servo.

Output impedance from most any solid state DAC will be very low - on the order of 100's of ohms or less. So no problem designing a passive crossover to feed the ADCOM.

 

[Overseas]

In theory, until properly measured and compared, I may very well claim to be hearing a difference between 2 DACs for ex., I mean 2 devices that were not measured at all. Heck, I can even claim that for a comparison between one measured DAC vs another un-measured DAC.
Am I right?

 

[LTig]

In theory, until properly measured and compared, I may very well claim to be hearing a difference between 2 DACs for ex., I mean 2 devices that were not measured at all. Heck, I can even claim that for a comparison between one measured DAC vs another un-measured DAC.
Am I right?

You can claim it if you passed a proper controlled blind test.

 

[Overseas]

You can claim it if you passed a proper controlled blind test.

Nope. Blind test beats measurements. I said 'not measured'.
Otherwise, you may very well measure 1 single device and blind-compare anything to it, whatever hears the same can be put on pair with the neasured device.

 

[BDWoody]

In theory, until properly measured and compared, I may very well claim to be hearing a difference between 2 DACs for ex., I mean 2 devices that were not measured at all. Heck, I can even claim that for a comparison between one measured DAC vs another un-measured DAC.
Am I right?

You can claim to be hearing differences between anything and anything else. People do it all the time.

Claims are easy. Providing evidence is what is generally missing.

What point are you trying to make?