Must know spec when considering a "discrete" R2R ladder dac

[georgehifi]

I don't know the % of what's the best is to get for the R2R "resistors" in the ladders, maybe others here know.

 

This from what I've read in the past, can make a big difference in the sound quality of an R2R "discrete" ladder dac.

(that's why they were laser trimmed in old chip based ones like the PCM1704, and got PCM1704"K" if they really trimmed up to a high accuracy)

 

Before purchasing, find out from the "advertising blurb" or "ask the manufacturer directly", the % of accuracy (I believe/hope it'll be below .01%) of each of the discrete resistors used in their resistor to resistor (R2R) ladder dac. The resistors get exponentially more expensive as the percentage of accuracy goes up. This it's said will make a big difference in their sound accuracy of that dac. 

 

Cheers George

 

Edited June 27 by georgehifi

[davewantsmoore]

37 minutes ago, georgehifi said:

I don't know the % of whats the best is to get, maybe others here know.

It is problematic (at best, totally misleading at worst) to go just by the % number.

 

It is true, that "if it is not enough" then it will be a problem for distortion... but "what is enough" depends on the exact scheme/code used.

 

In the olden days.... converters like this were made (eg. Philips, BB for low speed... MSB for high speed, etc.) were there was serial mode, and everything depended on the tolerance of the resistances.   This makes performance very difficult as you have switching glitches, and the resistors are super difficult if not impossible to be high enough tolerance.

 

46 minutes ago, georgehifi said:

I believe/hope it'll be below .01%

That's about what you need to get 16bit SNR

 

0.00001%+ for 24bit.... but this is not only impossible for the resistor, it gets swamped by all the other circuit.

 

It is why for example that TI/BB won't tell what the accuracy of their "laser trimmed resistors" are ... because they will be 0.01% or whatever (not enough), and people will make the wrong judgement.    They know that you should not be asking about the % (because you don't know what it truly means) ... so they don't tell.

 

1704 uses old (but nothing wrong with) scheme like 2x DACs internally to drive down SNR .... but more modern schemes can do much more parallel arrangement, and then optimise the driving of this through changing the code (computer program), to get the best output.

 

So one manufacturers distortion/SNR with 0.05% could be the same as another with 0.01% ... it is hard to generalise.

 

46 minutes ago, georgehifi said:

This it's said will make a big difference in their sound accuracy of that dac. 

People say a lot of things.  Heh.

 

I think what we know with DACs from long history is that in gross SNR terms ....we need "enough" and nothing more.    The SNR really just doesn't tell much about the sound quality...... sure, it says a lot about how well you could use the DAC in the studio in an AD-DA-AD-DA whatever arrangement, and/or when you cannot have inaudible (quiet) signal distorted ('cos you might amplify it into the mix later).....  but for audio playback.   16 clean bits is more than enough and then some.

 

Spikes of % THD/IMD, etc.... can get bad an maybe audible depending on the severity of what is causing them... but that is more about the overall stability and robustness of the scheme under real world signals (complex signals) ... than any contrived distortion (or even IMD) test.    32 tone, or whatever, IMD type tests like is done these days, is closer to real world, but.

 

 

To cut a long post short.... I really do not think these (SNR, SINAD, etc.) specs are audible... unless your use case (something that's not just "listening to music" pushes up against them..... and we can see this in the leaderboards on places like ASR.    There are plenty of blameless sounding boxes at both ends of the spectrum.    There are cheap boxes (eg. the miniDSP Flex) topping the leaderboard .... and they don't sound better than the middle of the pack....  and this is the proof in the pudding for R2R, they don't dominate on those specs but they sound good.

 

I do like "R2R" for the big reason that it isolates away the foibles on the clocking side.... where on the other hand, the clocking of "SD" DACs (whether they be "DSD" ... or with more bits, 6, 7, 8, 9) seems to be a deep rabbit hole for sound changes....  and I think this is the main reason that R2R sound good (pedestrian SNR/% distortion .... but not as liable to bad clock problems)