Just one more completely off topic post
excuse me...@barrows, or anyone else, do you know if the NDK SDA series are AT cut? I tried to find this out but couldn't see anything on their datasheets. I understand AT has the lowest phase noise around 25degrees but SC is better around 65degrees. Big difference.
Nope...
I do not know. Typically SC cut is used in OCXOs and is much more expensive than the NDK SDA series parts, so I think it is safe to assume that these are AT cut.
I think if one wants SC, one has to go to an OCXO, like the NDK DuColon, which is much, much more expensive, and maybe even hard to import into the US? It is difficult to find good OCXO oscillators at audio frequencies, as these are not typically used in audio products. I do know that Ayre has special SC cut oscillators made for their top level DAC(s) by Morion, but these are proprietary to them.
Perhaps a group buy could be arranged for an SC cut oscillator from some supplier at audio frequencies? Or some enterprising individual could special order SC cut xtals at audio frequencies and make a business of selling them to audio manufacturers and DIYers. I am pretty sure that is how the Pulsar Clocks were available, but the demise of Pulsar Clock suggests that such an operation was not viable from business perspective I guess. I was on the waiting list for a Pulsar Clock, but the never got made again int he frequency I needed, missed it by that much!
Just one more completely off topic post
@barrows, or anyone else, do you know if the NDK SDA series are AT cut? I tried to find this out but couldn't see anything on their datasheets. I understand AT has the lowest phase noise around 25degrees but SC is better around 65degrees. Big difference.
I do not know. Typically SC cut is used in OCXOs and is much more expensive than the NDK SDA series parts, so I think it is safe to assume that these are AT cut.
I think if one wants SC, one has to go to an OCXO, like the NDK DuColon, which is much, much more expensive, and maybe even hard to import into the US? It is difficult to find good OCXO oscillators at audio frequencies, as these are not typically used in audio products. I do know that Ayre has special SC cut oscillators made for their top level DAC(s) by Morion, but these are proprietary to them.
Perhaps a group buy could be arranged for an SC cut oscillator from some supplier at audio frequencies? Or some enterprising individual could special order SC cut xtals at audio frequencies and make a business of selling them to audio manufacturers and DIYers. I am pretty sure that is how the Pulsar Clocks were available, but the demise of Pulsar Clock suggests that such an operation was not viable from business perspective I guess. I was on the waiting list for a Pulsar Clock, but the never got made again int he frequency I needed, missed it by that much!
Thanks
Thanks man, just trying to share. I have learned tons from other members, so if i have anything to offer which might be of interest, I am happy to contribute.
I am not interested in just repeating myself over and over, LOL, so no more responses from me on that "other stuff", life is too short to dragged into that nonsense.
Getting back to AKM 4499, does anyone who is playing with the chip have anymore info on the direct DSD path, I am a little curious about the FIR filter at the input, can that be done on a single bit stream (is it treating the single bit stream as "analog" and just filtering it?) it appears it can from the block diagram? I am really only interested in the chip in the direct DSD mode in terms of oversampling in software to DSD 256 (I see from the data sheet that chip performance degrades at DSD 512). Looking at the rolloff of that FIR filter at DSD 256 it appears it would be entirely inaudible anyway (-0.3 dB @ 80 kHz).
Thanks man, just trying to share. I have learned tons from other members, so if i have anything to offer which might be of interest, I am happy to contribute.
I am not interested in just repeating myself over and over, LOL, so no more responses from me on that "other stuff", life is too short to dragged into that nonsense.
Getting back to AKM 4499, does anyone who is playing with the chip have anymore info on the direct DSD path, I am a little curious about the FIR filter at the input, can that be done on a single bit stream (is it treating the single bit stream as "analog" and just filtering it?) it appears it can from the block diagram? I am really only interested in the chip in the direct DSD mode in terms of oversampling in software to DSD 256 (I see from the data sheet that chip performance degrades at DSD 512). Looking at the rolloff of that FIR filter at DSD 256 it appears it would be entirely inaudible anyway (-0.3 dB @ 80 kHz).
Last edited:
Looks like AK4499 may soon be AKM's 2nd best dac: https://www.diyaudio.com/forums/digital-line-level/350733-ak4191eq.html#post6110269
And in combination with the suggested AK4498 DAC the outcome is... a mystery. The AK4498 is spec'd at -116dB, hardly a top of the line product.
Got me, I have no idea why or what the point of this chip is other than to let manufacturers market as a product differentiator. The old "we use our own custom xxx" so it must be better than those who use the canned solution.
These are great parts - I've been introduced to the AK4498 a year ago by AKM japan and at the time the engineers where using it with a FPGA development board - we heard the combination verses there highest end device and I can attest that they sound VERY much better!! Which makes them VERY VERY good indeed!!!
If the AK4191 digital section is as good as the FPGA demo solution I heard, then the combination its going to be very VERY good - I'm pretty sure the best "off the shelf DAC" solution available today when correctly implemented!!!
It reminds me of the SAA7350 / TDA1547 chipset combination and the original Crystal CS4328 which used dual separated die internally on the same IC leadframe
As having been involved in many ADC / DAC IC designs I can see MANY advantages in splitting the Digital / Analogue sections - so hats off to AKM for bring what will be a relativity (in semiconductor terms) low volume solution to the market!!!
I'd be more then happy to design a PCB around them
If the AK4191 digital section is as good as the FPGA demo solution I heard, then the combination its going to be very VERY good - I'm pretty sure the best "off the shelf DAC" solution available today when correctly implemented!!!
It reminds me of the SAA7350 / TDA1547 chipset combination and the original Crystal CS4328 which used dual separated die internally on the same IC leadframe
As having been involved in many ADC / DAC IC designs I can see MANY advantages in splitting the Digital / Analogue sections - so hats off to AKM for bring what will be a relativity (in semiconductor terms) low volume solution to the market!!!
I'd be more then happy to design a PCB around them
Last edited:
May I ask why -116dB THD worries you? THD this low has zero impact on sound quality! When THD is this low, its no longer significant in the sound quality matrix!
Also -116dB is a limitation of the AKM Demo circuit, we have seen MUCH better results when certain design decisions are corrected - and would expect the same improvements with the AK4498 Analogue array.
Then why not just copy the design from the AK4499 EVB and throw it on the AP and update the PDFs. It can't be that hard. The chip isn't even out yet.
So if distortion and DNR aren't important, then why do you think this part is better than AK4499? Is there anything you actually measure or is it just the typical "sounds better because it's less integrated" unsubstantiated claims?
Chris,
If I thought for a single moment that you where truly receptive to the reasoning behind separating Digital / Analogue Die - the impact this has on performance (and the design compromises / decisions which have to be taken by the IC design team with mixed signal digital design) I would spend the time to write a concise reply - but alas I have no intent in wasting my time with you - I'm not going down a pointless and all too predictable rabbit hole!
We can just say that these parts are not for you!
I'll not be seeing any reply from you as I'm just adding you as my first to the ignore list
Last edited:
There's a lot to be said for device integration mind you as separating things introduces other factors that can add to the problems of getting things to work as intended.
Surely the output of the AK4191 is going to run at high speed anyway so to say you're separating the digital and analogue sections wouldn't be entirely accurate. You're still going to have to contend with high speed signals around the analogue part.
As far as I can see the AK4191 outputs a parallel data-steam running at the oversampling frequency so the interface will run into the MHz.
In terms of digital noise the analogue section will have 8 high speed switching lines going into it. Certainly it won't do the processing inside it but you've still got high speed signals being delivered to the chip and then the return currents of those signals to contend with.
I wouldn't say the separation of the two allows for greater performance because of the separation itself. It's more the separation allows AKM to focus on smaller individual parts when it comes to development. In other words produce one very high and RnD costly modulator, but then have the freedom to produce a variety of analogue companion parts at a variety of price points.
It's not an eggs in one basket approach, like with the AK4499. Which, coming back to that part, if keeping all DAC aspects monolithic was a problem, would not manage the crazy specifications that it does.
Surely the output of the AK4191 is going to run at high speed anyway so to say you're separating the digital and analogue sections wouldn't be entirely accurate. You're still going to have to contend with high speed signals around the analogue part.
As far as I can see the AK4191 outputs a parallel data-steam running at the oversampling frequency so the interface will run into the MHz.
In terms of digital noise the analogue section will have 8 high speed switching lines going into it. Certainly it won't do the processing inside it but you've still got high speed signals being delivered to the chip and then the return currents of those signals to contend with.
I wouldn't say the separation of the two allows for greater performance because of the separation itself. It's more the separation allows AKM to focus on smaller individual parts when it comes to development. In other words produce one very high and RnD costly modulator, but then have the freedom to produce a variety of analogue companion parts at a variety of price points.
It's not an eggs in one basket approach, like with the AK4499. Which, coming back to that part, if keeping all DAC aspects monolithic was a problem, would not manage the crazy specifications that it does.
I think we need more information before we can conclude if the AK4191 + AK4498 is better than the AK4499. And in what way. Better in terms of audio performance? Better (lower) cost? Or?
Splitting the design into two chips can be a benefit, since it allows AKM to use a process suitable for high speed digital logic for the filter part and another process, suitable for audio (low noise, good linearity etc.) for the analog parts. Combining everything on one process does give some compromises. Perhaps the clock speeds are limited by the process node, limiting the oversampling ratios available, or perhaps the analog performance is limited by the use of a high speed logic process. Splitting the design gets the best of both worlds.
In principle AKM could now use a 7 nm process for the filter (I doubt that they have taken it to that extreme though, since the process will probably be too expensive).
The high oversampling rates of 256x compared to 8x as AKM describe may only be possible by using a very high speed process (28, 22, 14, 12 nm ???). Such a process will probably not be suitable for analog circuits. The 1.2 V supply used by the AK4191 (probably for the core) also indicates a fine geometry IC process.
I agree that the published specifications are not impressive compared to the AK4499. But perhaps the target is not to make it better than the AK4499, but to make is "good enough" and a lot cheaper?
And perhaps the higher overclock rate gives other benefits besides THD and noise?
Splitting the design into two chips can be a benefit, since it allows AKM to use a process suitable for high speed digital logic for the filter part and another process, suitable for audio (low noise, good linearity etc.) for the analog parts. Combining everything on one process does give some compromises. Perhaps the clock speeds are limited by the process node, limiting the oversampling ratios available, or perhaps the analog performance is limited by the use of a high speed logic process. Splitting the design gets the best of both worlds.
In principle AKM could now use a 7 nm process for the filter (I doubt that they have taken it to that extreme though, since the process will probably be too expensive).
The high oversampling rates of 256x compared to 8x as AKM describe may only be possible by using a very high speed process (28, 22, 14, 12 nm ???). Such a process will probably not be suitable for analog circuits. The 1.2 V supply used by the AK4191 (probably for the core) also indicates a fine geometry IC process.
I agree that the published specifications are not impressive compared to the AK4499. But perhaps the target is not to make it better than the AK4499, but to make is "good enough" and a lot cheaper?
And perhaps the higher overclock rate gives other benefits besides THD and noise?
I will have to reread their poorly written preliminary datasheet but I had the impression that the 256x is just the modulator rate, just like in AK449x and the text is marketing fluff to sound impressive.
Ah I see now that the rate is apparently double the other parts, which were 128x. Still has the interpolator block before the modulator.
Last edited:
I only had a few hours with AKM engineers, we where discussing a different subject and they where also very keen to demonstrate there HiEnd AK4499, I got talking about my discrete DAC designs and one of the AKM engineers got very excited jump up and brought me the AK4498 with an FPGA as the companion digital IC design had not been completed at the time.
I was lucky to be able to directly compare the AK4499 / AK4498+FPGA in the same system with the same music.
As I said in my earlier post, the AK4498+FPGA clearly sonically outperformed the AK4499.
As it turns out, there was a another person in the room who I originally thought was an AKM employee - later we bumped into each other and he mentioned to me how much better the AK4498+FPGA was to the AK4499 - I was surprised that a "complete stranger" was compelled to discuss the audio quality with me - so they too must also have been very impressed with the difference
IMHO, when compared to ESS DAC, AKM DAC's sound much more natural, real and organic... to me ESS DAC's sound artificial...
The AK4498+FPGA verses the AK4499 had a much wider / deeper sound stage, more texture to instruments / vocals etc. Bass was very nice... IMO the AK4498+FPGA was a big step up over the AK4499.
WRT costs, I believe that the AK4498+AK4191 solution will be higher cost then the AK4499 - but I really dont know... just an impression I formed...
I didn't have an opportunity to measure either solution - however I doubt any differences in "standard measurements" would help to explain the sonic differences I heard... suspect I have one of the best equipped Labs in the entire industry, and I've not had much success equating sound quality with measurements (atleast not the standard measurements)- if I could rely simply on measurements, then designing audio would be so much simpler and "automated"!
I've often said that its far easier to design a state of the art measuring unit then one that sounds good - they are not mutually exclusive, but conversely to a degree they need not be dependent on each other.
I was lucky to be able to directly compare the AK4499 / AK4498+FPGA in the same system with the same music.
As I said in my earlier post, the AK4498+FPGA clearly sonically outperformed the AK4499.
As it turns out, there was a another person in the room who I originally thought was an AKM employee - later we bumped into each other and he mentioned to me how much better the AK4498+FPGA was to the AK4499 - I was surprised that a "complete stranger" was compelled to discuss the audio quality with me - so they too must also have been very impressed with the difference
IMHO, when compared to ESS DAC, AKM DAC's sound much more natural, real and organic... to me ESS DAC's sound artificial...
The AK4498+FPGA verses the AK4499 had a much wider / deeper sound stage, more texture to instruments / vocals etc. Bass was very nice... IMO the AK4498+FPGA was a big step up over the AK4499.
WRT costs, I believe that the AK4498+AK4191 solution will be higher cost then the AK4499 - but I really dont know... just an impression I formed...
I didn't have an opportunity to measure either solution - however I doubt any differences in "standard measurements" would help to explain the sonic differences I heard... suspect I have one of the best equipped Labs in the entire industry, and I've not had much success equating sound quality with measurements (atleast not the standard measurements)- if I could rely simply on measurements, then designing audio would be so much simpler and "automated"!
I've often said that its far easier to design a state of the art measuring unit then one that sounds good - they are not mutually exclusive, but conversely to a degree they need not be dependent on each other.
Last edited: