The small ones will have lower phase noise than Crysteks... if NDK's in SDA or better flavour.
Again you don't understand, 4 poor oscillators become 2 good.
Just measure them with the FFT analyzer.
Wow, there is a PLL + VCXO and FIFO for SPDIF inputs. Who cares.
I thought you said that we need your clock with SC cut crystal hand-polished by virgins to get close-in phase noise that is acceptable.
I thought you said that we need your clock with SC cut crystal hand-polished by virgins to get close-in phase noise that is acceptable.
No, mine are good for satellite, GPS, radar but not for audio.
Indeed I'm selling my Timepod to buy an FFT analyzer.
I'm not so sure. I have two pairs of 22/24MHz NDK SDA hand picked by Jocko (RIP). I haven't been able get them to sound as good as I can get with very carefully implemented Crysteks (used with AK4499). That was with Jocko's advice on how to use the NDK SDAs too.
Of course, maybe the Crysteks don't measure as well. If so, there must be something that is being missed about how clocks affect sound quality.
Last edited:
Have measurements of NDKs ever been published? NDK doesnt even include a spec in the datasheet which is an unusual choice if they really are better, aswell as cheaper, than the crysteks.
NDK has published graphs, but they don't guarantee all clocks meet that performance level. For one example, please see the attached.
There was also a graph that somebody made, not sure who. It showed NDK SDA having lower phase noise than Crystek. Without more information about how the graphs were obtained, I would take it with a grain of salt. I know that Crysteks are very sensitive to power and to buffering. They also take 3-days of power-on time to fully stabilize. It those things were not properly done when measurements were taken, then I personally wouldn't put much credence in it.
There was also a graph that somebody made, not sure who. It showed NDK SDA having lower phase noise than Crystek. Without more information about how the graphs were obtained, I would take it with a grain of salt. I know that Crysteks are very sensitive to power and to buffering. They also take 3-days of power-on time to fully stabilize. It those things were not properly done when measurements were taken, then I personally wouldn't put much credence in it.
Last edited:
Uhm.. I am not proud of this measurement, naturally I did try my presently available NDK (supposedly SDA) oscillayor. 'Presently' because I have assembled quite a number of dac's (Mirand) with it, - but did not had the tool to measure, yet.
There is dirt in the middle, I was experimenting with a LMH6702 buffer, plus the setup not optimized.
22,576MHz Fc
It is showing like - 83dBc/1Hz; - 114dBc/10Hz; - 144dBc/100Hz.
That is quite much compatible with either the NDK doc's blue curve, or the tests of Jocko (of selected SD oscillators..)
So for me, at least with this one random example, the quality and originality of the SDA clocks from Lorien, the group buy here in Diya, - is having one positive data point added..
There is dirt in the middle, I was experimenting with a LMH6702 buffer, plus the setup not optimized.
22,576MHz Fc
It is showing like - 83dBc/1Hz; - 114dBc/10Hz; - 144dBc/100Hz.
That is quite much compatible with either the NDK doc's blue curve, or the tests of Jocko (of selected SD oscillators..)
So for me, at least with this one random example, the quality and originality of the SDA clocks from Lorien, the group buy here in Diya, - is having one positive data point added..
Attachments
Last edited:
Nobody seems to know if it will ever be back. It may have taken a special production line setup to make them.
Why would it be special other than their marketing BS? ESS is completely fabless and the chip is basically a clone.
Are the best sounding DAC made with a dac chip of the both manufacterers?
I'm even not sure !
I'm even not sure !
Which chip is a clone?? The AKM is quite different from ESS. It's not that easy to just move a chip to a new Fab.
I just meant that AK4499 is a conceptual "clone" of the ESS parts, at least. The point was, it's not a switched-cap DAC with internal op-amps like the other AKM V-out parts. If ESS has no special fab, then it stands to reason that UMC or whoever else ESS contracts with could easily produce the same thing. Mark said it required a "special production line", whatever that means. As far as I can tell, this DAC would be the easiest to move to another fab of all the parts they made. I don't mean to imply moving it is easy, but whatever the requirements are, there appear to be contract manufacturers with the capabilities to do it. I am aware the design must be done with the process in mind, so I don't mean that the design could be moved as-is, but clearly there are people making competitive or better parts without their own in-house special sauce.
Cirrus doesn't even have their own fabs as far as I know, so they are also clearly not relying on magic.
Chris, when you have your own fab, you are also running your own processes. Even worse, there will always be some derivative “tuned” processes for specialty chips, which is what the AK4499 was using.
Moving from your own fab to a foundry is extremely difficult, if possible at all, without re-designing the chips. Fabless companies have a huge advantage in flexibility and minimum risk, since there is no foundry with a single fab, they can also switch foundries since they all run generic similar processes, but also some potential limitations in reaching the best performance, since processes cannot be “tuned”.
Moving from your own fab to a foundry is extremely difficult, if possible at all, without re-designing the chips. Fabless companies have a huge advantage in flexibility and minimum risk, since there is no foundry with a single fab, they can also switch foundries since they all run generic similar processes, but also some potential limitations in reaching the best performance, since processes cannot be “tuned”.
Right, I meant to imply that since ESS is fabless and reaches the same or better performance, then how special can the AK4499's process need to be?
The AK4499 and ES9038 appear to be quite similar. I always figured one of the reasons that ESS used the switched-resistor output was because of their reliance on outside foundries. While still hard to move, this design seems to be the most compatible with a standard CMOS process. I completely understand why AK4497, AK4493, the ADCs, etc. would be incredibly dependent on process.
I don't mean to argue it would be easy or possible for AKM to port AK4499 to an outside foundry's process, just questioning the narrative that this is more of a special snowflake than AK4497. Maybe I'm wrong, but it seems like AK4499 is a Sabre-like design, which was prototyped in FPGAs and is fabbed by someone else (UMC?).
On a separate note, I am wondering if AKM's audio group will survive this? A lot of their customers will have already switched to their competitor's parts. Will they trust AKM in the future if they are still completely dependent on that one fab? I don't know where TI's audio converter parts are fabricated, but both Cirrus and ESS are fabless.
The AK4499 and ES9038 appear to be quite similar. I always figured one of the reasons that ESS used the switched-resistor output was because of their reliance on outside foundries. While still hard to move, this design seems to be the most compatible with a standard CMOS process. I completely understand why AK4497, AK4493, the ADCs, etc. would be incredibly dependent on process.
I don't mean to argue it would be easy or possible for AKM to port AK4499 to an outside foundry's process, just questioning the narrative that this is more of a special snowflake than AK4497. Maybe I'm wrong, but it seems like AK4499 is a Sabre-like design, which was prototyped in FPGAs and is fabbed by someone else (UMC?).
On a separate note, I am wondering if AKM's audio group will survive this? A lot of their customers will have already switched to their competitor's parts. Will they trust AKM in the future if they are still completely dependent on that one fab? I don't know where TI's audio converter parts are fabricated, but both Cirrus and ESS are fabless.
Last edited:
Maybe all of this is are "fake news", but:
It was a deal between AKM and Renessas.
It was written that AKM got a first samples in the early February.
It was an earthquake at the Renesas site.
They said that in Mart the factory was recovered and starts to operate.
But where the AKM chips?
ESS needs a competitor, Cirrus is one step down in performance, the Rohm DAC is a hack, TI and AD are obsolete in this respect. IMO, the AKM best strategy would be to re-enter the market with a new line of reference DA and AD chips, they have the time to design them.