• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Variation between same model dac

ShiZo

Addicted to Fun and Learning
Forum Donor
Joined
Sep 7, 2018
Messages
673
Likes
437
#1
How much variation do you think exists between multiple samples of the same DAC? I recently had a DAC replaced and without any analyzer would have no idea how much worse or better this could be.

I'm really big into overclocking and with intel processors, there is huge variation between CPUs referred to as the "silicon lottery". Is it possible that there is the same thing with DACs?
 

Doodski

Major Contributor
Forum Donor
Joined
Dec 9, 2019
Messages
5,813
Likes
4,860
Location
Canada
#2
How much variation do you think exists between multiple samples of the same DAC? I recently had a DAC replaced and without any analyzer would have no idea how much worse or better this could be.

I'm really big into overclocking and with intel processors, there is huge variation between CPUs referred to as the "silicon lottery". Is it possible that there is the same thing with DACs?
Variations between the same DAC ICs should be so small as to be imperceptible.
 

pozz

Data Ordinator
Forum Donor
Editor
Joined
May 21, 2019
Messages
3,492
Likes
5,517
#3
I would assume that the differences wouldn't be large, but then I've never seen data about it either. It's a good question.
 

maverickronin

Major Contributor
Forum Donor
Joined
Jul 19, 2018
Messages
1,849
Likes
2,191
Location
Midwest, USA
#4
I'm really big into overclocking and with intel processors, there is huge variation between CPUs referred to as the "silicon lottery". Is it possible that there is the same thing with DACs?
With DAC chips I doubt that would be an issue. AFIK, variation in overclocking potential in CPU is more or less about "binning". Not all come out at the same quality, but they test that each one can at least stand up to the rated multiplier. Even with unlocked chips some still end up better than others but they at least do what they're rated to. Any IC manufacturer needs similar quality controls.

If audio DAC chips were less application specific and could be "reprogrammed" to perform another task, convert at higher sample rates, or something else then we'd probably see the kind of variation seen in CPUs.

This is still a valid question, but I think the potential variation would be in the quality and consistency of the final assembly rather that the quality of the DAC chip itself.
 
OP
S

ShiZo

Addicted to Fun and Learning
Forum Donor
Joined
Sep 7, 2018
Messages
673
Likes
437
Thread Starter #5
With DAC chips I doubt that would be an issue. AFIK, variation in overclocking potential in CPU is more or less about "binning". Not all come out at the same quality, but they test that each one can at least stand up to the rated multiplier. Even with unlocked chips some still end up better than others but they at least do what they're rated to. Any IC manufacturer needs similar quality controls.

If audio DAC chips were less application specific and could be "reprogrammed" to perform another task, convert at higher sample rates, or something else then we'd probably see the kind of variation seen in CPUs.

This is still a valid question, but I think the potential variation would be in the quality and consistency of the final assembly rather that the quality of the DAC chip itself.
I was also reffering to the final assembly of the dac and QC. I should have been more clear.

My 9900k runs 5ghz at 1.28 :p, not the best but not the worse.
 

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
2,692
Likes
4,208
#7
Margins of variance would be too small to fall under categories of the sort of silicon lottery as experienced in CPU's/GPU's for example (those are pushing the thermal and power limits of just how much power is realistically allowed for the material). Audio processing would never reveal these limits. Not just because it's so low-powered, but because we don't have tools that could realistically test for such in the first place.

Silicon lottery yes, is part of nearly all electronics, but in some areas it's simply inconsequential due to the pervading paradigm on what such electronics actually do in the first place.
 

solderdude

Major Contributor
Joined
Jul 21, 2018
Messages
9,161
Likes
19,547
Location
The Neverlands
#8
How much variation do you think exists between multiple samples of the same DAC? I recently had a DAC replaced and without any analyzer would have no idea how much worse or better this could be.

I'm really big into overclocking and with intel processors, there is huge variation between CPUs referred to as the "silicon lottery". Is it possible that there is the same thing with DACs?
Unless there has been a silent revision or firmware implemented that actually changed performance or during production some incorrect components were placed and not be picked out by QC chances are slim there will be substantial differences. Of course there is tolerance in all components but shouldn't reach audible levels.
 

restorer-john

Major Contributor
Joined
Mar 1, 2018
Messages
7,133
Likes
18,189
Location
Gold Coast, Queensland, Australia
#9
Back in the day of ladder/multibit DACs, there was a tremendous difference, hence all the different gradings and guaranteed minimum specifications. Off the same wafer, even after laser trimming, grading was required. Within a given grade, manufacturers would often grade again for their products.

Having recently checked the low level resolution of some identical NOS vintage CD players in my collection, I can confirm they:

a) had no evidence of drifting from their factory adjustments (top four MSB trims in each channel) ie I could not improve their adjustment over factory settings from 30 years ago.
b) had differing levels of THD at -60dB but not significantly different to one another.

Note they were identical selection (grading) Burr Brown R2R chips from the same batch in identical machines. Both were within their rated (specification) numbers.
 

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
2,692
Likes
4,208
#10
Oh and one thing I forgot that restorer-john just reminded me of.. Intel CPU's and AMD/Nvidia GPU's revelation of silicon lottery is only evident when pushing the processors WAY beyond rated specifications. Even they cease to exhibit any discernible phenomena of silicon lottery if used out-the-box with no overclocking settings applied. Since we don't have access to "overclocking" tools of ADC/DAC hardware as consumers, we would never find any performance variance based on silicon evaluations. Do they still exist? Yes, but never where rated specifications would ever be violated (unless a company was buying silicon at the edges of a wafer or something silly like that which wouldn't be performance based silicon lottery, but instead would just contribute to many chips failing rated specification, that would never pass QC and sent to companies for final sales..).

So audio silicon lottery doesn't exist in any relevant semblance beyond the atomic/material science of silicon wafer production realities itself.

I also realized your question was also asking at the same time about DAC revisions. Which is a whole other issue to the silicon lottery phenomena of on-die architecture power/efficiency envelopes. That question on the other hand is, actually very possible to exhibit varience that a DAC maker might not be aware of changing between batches of parts ordered from companies.

Like if for instance you take a look at the RME V2 DAC, you can see potentially that the IEM port is possibly worse than the V1. Or one other area that seems to be spreading like wildfire of performance degradation, being jitter (more spikes around the main frequency tested). You can see issues like this arise when OEM's possibly not informing manufacturer customers of revisions of SKU's that are sharing the same name as the prior. This is just a whole other issue to silicon lottery that is mostly determined by the quality of silicon based on from what portion of the wafer your die is made from (the further you deviate from the center of the wafer, the lower quality the silicon essentially, and is why Enterprise industries pay and arm and a leg for realiability and consistency, to which someone like Intel complies by reserving the center wafer for Xeon parts or things like that, while Nvidia reserves it for their highest end GPU's costing thousands).
 
Last edited:
Top Bottom