Measurement of R2R DAC

[gvl]

That graph only shows linearity to a bit more than 120 dB/20 bits though. But sure, if I can one for free to test, I will.

They will be releasing a relatively low cost minimalistic Metrum DAC called Flint soon, if you ever dreamed to have a decent R2R NOS DAC in your collection that may be the one to get



PS: even 20 bits for a R2R is pretty darn good.

 

[Purité Audio]

‘Decent r2r NOS dac’ oxymoron.
Keith

 

[gvl]

You can always filter/upsample externally using HQPlayer or even SoX for example, chances are external digital filters are superior to those embedded on DAC chips due to not having the same computational constraints. Decent here means high-speed R2R D/A converter with good linearity, which is what the latest Metrum DACs are, or said to be. Perhaps Amir can check if the claims are true.

 

[amirm]

Perhaps Amir can check if the claims are true.

I hope to test the effects of external resampling to that of the DAC itself.....

 

[gvl]

I hope to test the effects of external resampling to that of the DAC itself.....

hat would be interesting but probably dependent on the filter/algorithm used, which there are many?

 

[amirm]

hat would be interesting but probably dependent on the filter/algorithm used, which there are many?

And that's the challenge and hence the reason I have not done it yet.

 

[svart-hvitt]

And that's the challenge and hence the reason I have not done it yet.

Use Jussi’s recommendation of focusing on the different poly sinc filters.

 

[Arnold Krueger]

I hope to test the effects of external resampling to that of the DAC itself.....

Software should always win or tie because it need not run on a limited hardware configuration and need not operate in real time.

I know of no hardware algorithm that can't be executed as software.

 

[Arnold Krueger]

hat would be interesting but probably dependent on the filter/algorithm used, which there are many?

Many? Just one extensive collection analyzed here: http://src.infinitewave.ca/

Then many software resamplers have adjustable parameters so they represent multiple families of different resamplers: Example Sox: http://sox.sourceforge.net/

 

[svart-hvitt]

Software should always win or tie because it need not run on a limited hardware configuration and need not operate in real time.

I know of no hardware algorithm that can't be executed as software.

MQA...

 

[gvl]

New R2R on the block: https://www.massdrop.com/buy/massdrop-x-airist-audio-r-2r-dac?utm_source=linkshare&referer=5242TX

 

[mindbomb]

thd+n relatively high on the massdrop r2r dac. Is that hinting at linearity issues?

 

[amirm]

thd+n relatively high on the massdrop r2r dac. Is that hinting at linearity issues?

It is two orders of magnitude worse than classical DACs. So something is going on there.

Love to measure this and compare it to Schiit Modi 2 Multibit. Anyone will be buying it? If not, I may have to.

 

[Veri]

It is two orders of magnitude worse than classical DACs. So something is going on there.

Love to measure this and compare it to Schiit Modi 2 Multibit. Anyone will be buying it? If not, I may have to.

If it even releases. It looks like massdrop has copied the design without licensing it... (author just commented)

https://www.massdrop.com/buy/massdrop-x-airist-audio-r-2r-dac/talk/2098625

 

[bunkbail]

They have some measurements for the Massdrop RDAC already over at SBAF http://www.superbestaudiofriends.org/index.php?threads/rdac-measurements.6418/

And it doesn't look pretty. It looks like the DAC has just 11 bit of linearity, yucks.

 

[Veri]

They have some measurements for the Massdrop RDAC already over at SBAF http://www.superbestaudiofriends.org/index.php?threads/rdac-measurements.6418/

And it doesn't look pretty. It looks like the DAC has just 11 bit of linearity, yucks.

Was about to post this too Also note that over at SBAF they don't think this is a "bad" thing per se:

Again, in a nutshell, hard to say how 10.5 bits (ENOB) is going to sound. It likely depends on the type of noise limiting the ENOB. This is one reason I kind of criticized Amir and his performance characterization using Amir-bits. The other is the fact that Amir-bits are not even ENOB.

 

[amirm]

Was about to post this too Also note that over at SBAF they don't think this is a "bad" thing per se:

Again, in a nutshell, hard to say how 10.5 bits (ENOB) is going to sound. It likely depends on the type of noise limiting the ENOB. This is one reason I kind of criticized Amir and his performance characterization using Amir-bits. The other is the fact that Amir-bits are not even ENOB.

If he thinks this is my invention, he needs to go and read Robert Harley's article on this topic from 1995 on linearity measurements and how to interpret them: https://www.stereophile.com/content/quality-lies-details-page-5

When stereophile used to run this test routinely, they would always attempt to interpret the linearity as I do.

 

[amirm]

I see CEE TEE there. Why doesn't he present his own measurements to counter what is there? Does he not have any?

 

[DonH56]

10.5 bits is 1 part in 1448 or 0.07% so in terms of something like THD you could argue it is below our threshold detection for a full-scale signal. But it also means noise or whatever is only 63 dB down so may be audible at lower levels depending upon the cause of the distortion and how it tracks with amplitude. The point I (and many others) have made before is that either way, given the number of inexpensive DACs that measure much (MUCH!) better, why pay for one that does not?

I suspect "Amir bits" is in reference to Amir's linearity deviation plots which are not based on ENOB per-se (he does not measure ENOB directly IIRC and the metric for deviation is not 6 dB but something lower). I have known and used ENOB for ages but it is not the end-all be-all spec and as long as I know what Amir is doing his scheme provides a good relative reference for comparisons among products. Which is all most of us want anyway.

IMO! - Don (well, the math is not opinion, but check it...)

 

[derp1n]

1) We do use some integer-rate upsampling with linear phase FIR filters. This gives us a smoother output waveform and thus a smoother sound. Without the upsampling, we found that lower sample rates could have a harsher edge to the sound from the larger steps between samples.

Does this William Tse guy of Airist understand digital audio at all?