https://www.headphonesty.com/2024/07/apple-patent-dacs-promises-sound-quality/
Can those who know what they are talking about comment of this?
dave
Can those who know what they are talking about comment of this?
dave
Very interesting @iancanada @dddac might understand better what is the "supposed" revolution... @lrisbo might be able to forward this to Bruno P?
One can always wish! 😀
One can always wish! 😀
Last edited:
Do you have any specific questions about it?
It looks like they're trying to compensate for a process or temperature gradient on die by loading the data into the DAC with the opposite gradient or with a changing gradient such that the process/temperature gradient averages out. That's kinda nifty but it's normally not super hard to get good matching on die. Maybe Apple's desire for miniaturization causes them to push the limits there, for example if the DAC is part of an RF transmitter, it could be hard to avoid a temperature gradient.
It would be nice if their graph had units so one could get an idea of the magnitude of things, including the advantage of this approach.
As pointed out above, this is for RF, where the linearity requirements can be significant to get good channel separation and selectivity. Whether it'll translate to good audio performance should this approach be used there is another story.
Tom
It looks like they're trying to compensate for a process or temperature gradient on die by loading the data into the DAC with the opposite gradient or with a changing gradient such that the process/temperature gradient averages out. That's kinda nifty but it's normally not super hard to get good matching on die. Maybe Apple's desire for miniaturization causes them to push the limits there, for example if the DAC is part of an RF transmitter, it could be hard to avoid a temperature gradient.
It would be nice if their graph had units so one could get an idea of the magnitude of things, including the advantage of this approach.
As pointed out above, this is for RF, where the linearity requirements can be significant to get good channel separation and selectivity. Whether it'll translate to good audio performance should this approach be used there is another story.
Tom
Yes, this is for RF and has little to do with audio. The headphonesty article is very misleading and riddled with hyperboles.
Multibit sigma-delta audio DACs and ADCs usually also use unit DAC cell arrays that need to match well.
It looks like an extension of common-centroid layout, which is a well-known trick to improve matching on a chip. The extension is that they look at the enabled cells together, rather than the individual cells. Whether it is usable for audio will depend on how well or how poorly it can be combined with the dynamic element matching techniques normally used in mutibit sigma-delts audio converters.
It looks like an extension of common-centroid layout, which is a well-known trick to improve matching on a chip. The extension is that they look at the enabled cells together, rather than the individual cells. Whether it is usable for audio will depend on how well or how poorly it can be combined with the dynamic element matching techniques normally used in mutibit sigma-delts audio converters.
I designed a block of a microprocessor in the late 1990s, implementing three 8-bit video DACs (for 24-bit color). This was back in the days before HDMI. The DACs ran at something like 90 or 120 MHz, I don't remember precisely. The physical implementation was a 16x16 matrix of unit cells, extremely similar to the Apple patent. Unsurprisingly, each cell was just a current source which could be steered either to the output or to the dumpster. The fab process was 0.22um digital CMOS. For those who may remember the acronym, this was just a "Brooktree RAMDAC" without the RAM, integrated onto a CPU.
Some prior art for the Apple patent might be the "random walk" style of DAC: (link) .
Some prior art for the Apple patent might be the "random walk" style of DAC: (link) .
Last edited:
It seems pretty clickbaity to me.
I seem to recall an older Philips DAC that switched between several current sources similar to the scheme Mark describes in order to (supposedly) improve the performance (probably DNL). I can't remember the part number.
A DAC like the ES9038 deliver nearly 140 dB DNR and THD below 1 ppm in stereo mode. That's hard to improve on.
Tom
Yes, but ESS claims ES9039 sounds better. How could that be possible if DNR and THD are all that matters? Anyway, ESS seems to think other things matter too.
If sound quality matters, go for the ES9069...It has the MQA renderer to "Reveal the original master resolution" 😝😂🤣
https://www.esstech.com/wp-content/uploads/2023/03/ES9069_Datasheet_v0.1.3.pdf
140dB DNR and or MQA are both good for advertising. The former tends work better on engineers. The latter, more on the general public.
Some of us have enough design experience to realize how hard -140 dB DNR is to get. Think about it for a little. That's not a lot of stray electrons...
Tom
As does the ES9038PRO. The PRO is $62.79. The non-PRO is $13.50. Both prices from Mouser, QTY 100. The difference is the MQA magic. L@@K!! For only an additional $49.29 we'll garble your data for you! You should like this!
Tom
Just pointing out that a couple of numbers don't necessarily mean what some people think it does, like some folks do at that other website. I feel its important to repeat the message now and then just to so everyone keeps aware that a couple of dazzling numbers aren't all there is to it. Its real easy to get focused on the numbers and forget what is missing there.
Ah. Fair. My point was that audio DACs are pretty darn good now, so maybe Apple's trickery isn't needed or wanted. Or maybe they use even fancier trickery. The ES9038 has some THD reduction magic in it (says the data sheet). Yes. I said THD again. I know it's not be-all, end-all. But it is an important metric nonetheless. The fact that it's an important metric does not imply that it is the only metric or the most important metric. Only that it is one of many and a metric that I and many others consider to be important.
There. Maybe I should write a disclaimer for my disclaimers from now on. 🙂
Tom
There. Maybe I should write a disclaimer for my disclaimers from now on. 🙂
Tom