Do you know what clock speed was used for that? I've had my ES9023 running 384 using Amanero's clocks in sync mode but it is far from optimal.
Another update : making progress with casing.
I've removed the xo socket and fitted the Crystek CCHD-957 22.5792Mhz. It works exceedingly well for CD, but it can't keep lock continuously with DVD-A (96Khz). I haven't tried SACD yet.
I plan to try the 50Mhz SAW. I'll fit it from the top using the X1 input and a nearby 3.3V power via. I will try using the en pin as gnd - so when en is grounded to turn off the Crystek, it will prove gnd for the SAW so that switches on. I've also fitted the u.fl socket so I can feed in an external clock.
The I2S lines need improving, but I'm waiting for the board from Acko..... so, it's time to start seriously trying different clocks and op amps. Less soldering, more music. That's a good feeling....
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
I've removed the xo socket and fitted the Crystek CCHD-957 22.5792Mhz. It works exceedingly well for CD, but it can't keep lock continuously with DVD-A (96Khz). I haven't tried SACD yet.
I plan to try the 50Mhz SAW. I'll fit it from the top using the X1 input and a nearby 3.3V power via. I will try using the en pin as gnd - so when en is grounded to turn off the Crystek, it will prove gnd for the SAW so that switches on. I've also fitted the u.fl socket so I can feed in an external clock.
The I2S lines need improving, but I'm waiting for the board from Acko..... so, it's time to start seriously trying different clocks and op amps. Less soldering, more music. That's a good feeling....
Last edited:
no sorry, it wasnt me, you might try PM'ing Bunpei here on the forum, I think it was one of his compatriots a year or so ago, there will be some info on some of the Japanese audio forums if you are a brave google translator, but I believe there was some talk here on the forum back then also. I remember Acko collaborated on a 4399 project around the same time so perhaps he has some more info
sorry I cant be of more help.
Thanks for tip, I think found the thread. In it Bunpei mentions something about a friend testing DXD on an AK4396 and himself on an AK4399 but as I understood it they both got noise due to low fs. Looks like after that he changed to the ES9018. Shame he didn't test the 45.xxx clock with AK4399.
@DIYINHK, have you tested 45.xxx/49.xxx from your CM6631 board into your AK4399 DAC?
KK, looking good!
Over at the other thread, when you said "audiocom" did you mean the audiolab es9018 DAC? (photo in my blog)
I do plan to use some higher value capacitors to mitigate the overshoot and undershoot of all series regulators but certainly not 1000 uF. Likely, they will be Elna silmic II in the range of 10 uF or so. For the analog, there are 8 bypass points, so if using 10 uF, there will be a total of 80 uF plus whatever I put in the board input. If you had to use larger caps, what value would you use?
Of course there is also a ceramic 0.1 uF at each position
Over at the other thread, when you said "audiocom" did you mean the audiolab es9018 DAC? (photo in my blog)
I do plan to use some higher value capacitors to mitigate the overshoot and undershoot of all series regulators but certainly not 1000 uF. Likely, they will be Elna silmic II in the range of 10 uF or so. For the analog, there are 8 bypass points, so if using 10 uF, there will be a total of 80 uF plus whatever I put in the board input. If you had to use larger caps, what value would you use?
Of course there is also a ceramic 0.1 uF at each position
Read some more of the TPS7A47 regulator. It is actually designed for fast transient response and optimally 50 uF ceramic at the output for low noise and fast transient (which are already implemented in eval board.
Perhaps either adding small ceramics or nothing is the best approach if using the TPS
Perhaps either adding small ceramics or nothing is the best approach if using the TPS
ahh sorry I wasnt thinking of the tps, but regardless a 10uf silmic is pretty pointless IMO, only satisfies audiophile urges =) and is lower performance than anything else you mentioned. you are after lowest noise, transient response is not high on the agenda when supplying secondary regs and remote decoupling caps for a fairly distant constant current (if you are meaning to supply the analogue pins) it doesnt struggle with low impedance ceramics or say 100-330uf on the output, but either way forget the silmic IMO.
you do however have to keep in mind resonant effects/ringing from paralleling different types of caps in series with an inductive trace
you do however have to keep in mind resonant effects/ringing from paralleling different types of caps in series with an inductive trace
Last edited:
AVCC doesnt need fast transient response, its a reference voltage.... the digital needs fast transient response more than AVCC, you'll find it was designed for low noise too =)
low impedance is a plus, but its the caps that supply the load transients, an external regulator is there to supply the decoupling caps, they supply the transient current demands if any, but low noise is more important for AVCC than transient response.
low impedance is a plus, but its the caps that supply the load transients, an external regulator is there to supply the decoupling caps, they supply the transient current demands if any, but low noise is more important for AVCC than transient response.
Last edited:
Different ak4399 chip production date may have different limit, I have tested few chip in my stock with cm6631 at 45.xxx/49.xxx and they work flawlessly
according to the ak4399 datasheet it's overclocked.But cm6631A with 22.xxx/24.xxx MCK firmware is replacing cm6631, the overclocking chance is less in the future.
I'l pleased it has a home. I have too many built pcbs sitting around and not enough finished gear.
The choice of cap, its location, its size... aiya. The Walt Jung articles on picking capacitors is really helpful. Here's part1 :
http://waltjung.org/PDFs/Picking_Capacitors_1.pdf and part2:
http://waltjung.org/PDFs/Picking_Capacitors_2.pdf and I enjoyed reading this too :
Humble Homemade Hifi
What I've learned - if you want to reduce noise you need low leakage. If you want to reduce noise over a broadband you need a big(ger) cap with a small(er) bypass. If you want to maintain tonal balance, SilmicII are great, Panasonic FC and FM are good, and then have a 0.1uF x7r bypass. The size depends on the load current. For electros, 10uF is fine for a single channel op amp ~5mA. Solid polymer caps generally sound too bright unless, like Audiolab, you use a ton of them, but then all you get is the character of the cap. Not my taste. I tried this (after Blackgate FKs ran out) with Nichicon PLE.
If you use a bank of caps, they must be arranged so that no one cap is taking the strain of the current demands more than the others. One thing I haven't tried yet is a "wagon wheel" of caps with the wheel hub as the input and output points and the caps arranged radially with identical lead length. Just an idea.
So, for this DAC, AVCC@3.8V, and a separate supply for the 3.3V and 1.2V adp's, I only have the 0.1uF ceramics around the DAC IC, and the 10uF's as the board/ESS recommends on the 1.2V supply. There is a 220Uf on the board power input from the Sigma11. The Sigma11 doesn't need anything else, and I'm even considering replacing that 220uF with something smaller. I've already removed the 100uF I had near the DAC IC and felt that sounded punchier, and then another 470uF at the input of the board has gone too. But this is the Sigma11. It only has 1uF of film cap on it's output - that's all it needs for stability.
The adp151 3.3V has 100uF BC038 on input and output. For the current demands, this is a little low, but it's enough I think and avoids possible instability from too much C, and it also has 1uF ceramics on input/output. The adp151 1.2V after it sees the capacitance on the output of the 3.3V reg and also has local 1uF X7R caps on in/out to keep it stable. It has the caps around the IC to deal with.
As current demands on digital circuits are much more constant than something like a class AB amp, I think you are right to aim for lowest noise on these. For the analogue supply, I think low impedance is the priority, so if your supply doesn't have a low Z then caps are the next best thing. Check out the ESR/ESL etc of a 1uF polyester cap vs a 10uF electro. Personally, I'd lean towards using the 1uF film cap.
Just imo. Cap choice is one of the real "snake oil and dark arts" stuff in audio so for sure people won't agree.
For my bi-amps, I used only Mundorf silver/oil in the audio path, and the psus used only SilmicII or Blackgate FK electros, and Sonicap and Vishay MKP1837 film caps, and only COG ceramics. I got a 1KVA isolation transformer and made input and output filters to clean up the 220V mains for the hifi. I tried metal oxide caps there and my hifi sounded way too bright. I went back to x and y rated film caps - back to normal. Every cap matters.
Last edited:
yes I suppose it supply the voltage reference that is converted to transient current over a dynamic N-Channel element (pretty sure you will find small planar caps are etched into the Die as well) and the current comes from ground in an all VEE+ system... remember?
in any case it still wouldnt be a transient demand supplied by the regulator, it will be supplied from the decoupling caps, the regulator has to keep current up to the decoupling caps for transients.
but yes 'fast transient' is overstating it IMO for such a system, the digital will have faster transients, the clock ideally none aside from start up, which again will be primarily supplied by the internal decoupling/bypass caps in most clocks these days.
also on the greater scheme of things, the tps, while an excellent LDO, is not very fast, or very low noise. the LDOs are getting there, but still nothing compared to a shunt reg, or a flea type buffered reference like shown in the datasheet and the more deluxe version used by the ackodac
Last edited:
Shunt vs Serial
Just read the Walt Jung interview at AX and in the article it says that the advantage of shunt over serial is that it prevents the ground currents returning all the way to the raw DC thus avoiding potential problems. I was expecting noise and speed being advantages but those were not mentioned.
Just read the Walt Jung interview at AX and in the article it says that the advantage of shunt over serial is that it prevents the ground currents returning all the way to the raw DC thus avoiding potential problems. I was expecting noise and speed being advantages but those were not mentioned.
thats correct alright, its why I dont tend to use them that much, or recommend them all that much (it was incidental that I put shunt first) Most places I see people considering them for arent really specifically suited to them and the heat and space make them a bit more difficult to implement well, while a nice flea type reg can take up very little space and practically no heat.
they are low noise and low impedance, but not really any better than linear based. you should buy the copy of linear audio where these were all tested head to head, the salas regs did do very well in noise and low impedance. Jack did a very good job on that article. jackinnj you may know him as here. he also posted a few less involved comparisons here on the forum.
probably why Walt didnt make a deal about those factors is because his linear version of the super regulator already has stupidly low noise and low impedance, so making the shunt version of the reg didnt really net him any improvements in those areas.
it is one reason why parallel regs (shunt) are good for clock supplies though.
they are low noise and low impedance, but not really any better than linear based. you should buy the copy of linear audio where these were all tested head to head, the salas regs did do very well in noise and low impedance. Jack did a very good job on that article. jackinnj you may know him as here. he also posted a few less involved comparisons here on the forum.
probably why Walt didnt make a deal about those factors is because his linear version of the super regulator already has stupidly low noise and low impedance, so making the shunt version of the reg didnt really net him any improvements in those areas.
it is one reason why parallel regs (shunt) are good for clock supplies though.
Last edited:
Different ak4399 chip production date may have different limit, I have tested few chip in my stock with cm6631 at 45.xxx/49.xxx and they work flawlessly
But cm6631A with 22.xxx/24.xxx MCK firmware is replacing cm6631, the overclocking chance is less in the future.
Thanks for the confirmation, much appreciated.
Could you please post a detailed picture of the top side of the AK4399 board? the one in your ebay site shows well the bottom side but it is not possible to read the top one as the camera flash reflection makes it impossible and the resolution is not too good. TIA
- Status
- Not open for further replies.