@Mark
I totally agree your thoughts at conceptual level, but their implementation approach is slightly different in my mind.
As more complex a system become, as many potential error sources it gets. All the high quality systems are either expensive because of their expensive components, or simple.
Actually this is the reason why a dedicated diy-er can make a system spending 500$ as good as a 2500$ mass-manufactured one: he keeps it simple for minimizing costs, but uses high quality components in order to reduce the potential errors.
I would like to use a low-cost FPGA for learning purpose and for testing its potential, but for a final project it might induce more errors than benefits. Don't blame me, but at this level of timing values as in a DAC digital filter, low-cost concept might have the sense of a 500$ system who sounds like a 200$ mass-produced one.
Writing custom FIR filters is a big option ESS offers and even if it is not a 1000 filter coefficients and complex shapes capable, but only 128+16 2 stage filter, it is still good enough to build a simple and good system.
And yes, after reading about digital filters almost 8 hours yesterday, I agree that an external high quality built filter can bring an important benefit to sound, but is seems to me like we are speaking about a 2500$ system who might sound like a 10000$ factory-produced one.
I totally agree your thoughts at conceptual level, but their implementation approach is slightly different in my mind.
As more complex a system become, as many potential error sources it gets. All the high quality systems are either expensive because of their expensive components, or simple.
Actually this is the reason why a dedicated diy-er can make a system spending 500$ as good as a 2500$ mass-manufactured one: he keeps it simple for minimizing costs, but uses high quality components in order to reduce the potential errors.
I would like to use a low-cost FPGA for learning purpose and for testing its potential, but for a final project it might induce more errors than benefits. Don't blame me, but at this level of timing values as in a DAC digital filter, low-cost concept might have the sense of a 500$ system who sounds like a 200$ mass-produced one.
Writing custom FIR filters is a big option ESS offers and even if it is not a 1000 filter coefficients and complex shapes capable, but only 128+16 2 stage filter, it is still good enough to build a simple and good system.
And yes, after reading about digital filters almost 8 hours yesterday, I agree that an external high quality built filter can bring an important benefit to sound, but is seems to me like we are speaking about a 2500$ system who might sound like a 10000$ factory-produced one.
zenonn67, ES9038Q2M AVCC consumption is < 8mA, OPA2134 output 35mA, hence, you don't need bc413, I guess any 0603 beads 1k@100MHz in-serial, or just 50ohm resistor, may prevent opamp(BTW, quite slow one 8MHz) self-oscillating. bc413 in that circuit will fast charge and slow passively-discharge AVCC caps, and this is not an improvement of regulation for sure.
888777, i have used the TPS7A4700 modules one for each channel of the AVCC.
for me this was the best upgrade for the dac, along with the clock.
i have not tried the method that mark suggests though, it might be much better
here as a picture of the finished dac as i promised few months ago
for me this was the best upgrade for the dac, along with the clock.
i have not tried the method that mark suggests though, it might be much better
here as a picture of the finished dac as i promised few months ago
Attachments
@Thorp,
The price of dacs with external interpolation filters (and upsampling to take best advantage of the filters) starts at around $1,700 here in the US. The cost of the additional chips needed even in small quantities is not so much, maybe $15 to $20 each, for 2 or 3 chips. There will of course be some more support components as well. A lot of the difference appears to be in the engineering effort involved, since some details of the methods remain mostly proprietary.
EDIT: At the $500 price level one can simply buy Allo Katana for $250 or a little more and add a sufficient linear power supply that is compatible with the load it presents. Film caps are pretty forgiving and can work quite well to smooth out that last bit of graininess. No external filtering or upsampling, but for a basic clean low-jitter implementation it is pretty darn good.
The price of dacs with external interpolation filters (and upsampling to take best advantage of the filters) starts at around $1,700 here in the US. The cost of the additional chips needed even in small quantities is not so much, maybe $15 to $20 each, for 2 or 3 chips. There will of course be some more support components as well. A lot of the difference appears to be in the engineering effort involved, since some details of the methods remain mostly proprietary.
EDIT: At the $500 price level one can simply buy Allo Katana for $250 or a little more and add a sufficient linear power supply that is compatible with the load it presents. Film caps are pretty forgiving and can work quite well to smooth out that last bit of graininess. No external filtering or upsampling, but for a basic clean low-jitter implementation it is pretty darn good.
Last edited:
@paulmarinis,
Regarding AVCC supplies, I have not done any real comparisons of different regulators myself. I wanted to focus my time in other areas so just used something that was known to work, although at one point I found a need to improve it a little from the original ESS recommendations. It is clear that other methods can work quite well, but I don't know that any public comparisons have been published.
Regarding AVCC supplies, I have not done any real comparisons of different regulators myself. I wanted to focus my time in other areas so just used something that was known to work, although at one point I found a need to improve it a little from the original ESS recommendations. It is clear that other methods can work quite well, but I don't know that any public comparisons have been published.
As a genuine hobbyist, I prefer to spend 2000$ or more and many hours experimenting for making a DAC better than Allo Katana with only 150$.
Right, there is that. But, what happens you hit the end of the line with what you can do with your experimentation? Reducing jitter and custom filter coefficients can only help so much. And, still the music can definitely sound audibly better. What is recorded on better recorded and mastered CDs is better than can be heard with built-in dac filters, no matter how low jitter is. The dac chip manufacturers know this, but they are trying to sell large volumes of chips for use in many different dacs which are made to sell at different price points. They can't cater to the very high end manufacturers only, or the chips will cost too much to compete in the high-volume chip market. They make it possible to produce audibly better dacs if desired, by allowing external DSP, for one thing.
Well it two steps forward and one step back for me. The added capacitance on the AVCC rails did really help a lot and things were sounding really nice. Then the crappy low quality RCA output jacks broke. So I took things apart to fix things by adding some new higher quality ones to the back of my enclosure. This would also allow me to hook up the new output stage. I also decided to try the film cap thing that Mark had been recommending for the =/- 15 v rails. I made up a separate board for the caps and scrounged around in my parts bin for film caps, I managed to put together 18 uF. I needed to rearrange the inside layout to fit everything in the case. What I didn't find out until it was too late was that a tiny blob of solder landed on the pins of the dac chip and shorted it out when I powered back up. I managed to remove it, but the damage was done and the chip is dead. I've got a new board on the way.
Attachments
Ouch! Which pins got shorted out on the dac chip? Just wondering because the chips might not be that easy to kill in most cases if all the pin solder connections are returned to normal. The reason I ask is because it can take some very tiny and sharp needle tip probes to actually come into good contact with individual dac chip pins. Ringing out back to solder joints or connectors at the far end of affected traces may show opens, or there could still some solder shorting two pins together. If by chance you don't have needle probes, you can use a jumper clip to make a sewing pin or needle into a probe tip for your DVM to probe for good connections at the actual dac pins. Easiest with a DVM that can be set to give an audible tone for continuity testing so you don't have to take your eyes off the probing.
To remove excess solder, usually adding a bit of paste flux and using good quality solder wick and a larger, but sufficiently sharp chisel solder iron tip to get down right next to chip and heat quickly may work best. In fact that's one way to solder high density SMD pins, simply use enough low temperature leaded solder to short the pins together a little out at their ends, but not so much solder as to get it way back under the pins, then remove the excess with solder wick. Usually it wicks out all the solder shorts very effectively while leaving the pins nicely soldered.
Last edited:
How about one of these finished boards 16x Digital interpolation filter - drive PCM56, PCM58, AD1865 and so on up to 768 kHz for external filtering?
Hi wealas,
Close, but not quite. Sabre dacs require 8x interpolation filters, and we would want it to run at non-standard sample rates in order to allow a more relaxed transition-band between stop-band and pass-band. That makes it possible to better attenuate aliasing, while allowing the filter to be very low ripple in the passband and with very gentle rate of phase shift with respect to frequency (group delay). Sabre dacs also use a standard word length of 32-bits, which is likely the required input format needed for compatibility purposes when the dac is configured for external filtering.
But, the basic idea looks right. Program a low-cost Spartan 6 for double precision PCM interpolation filtering, with a good number of taps available. I image 3lite could make the right kind for us or maybe share some code if he were motivated to do either of those things. Actually, I asked about something to that effect and the response back was basically one of no interest: https://www.diyaudio.com/forums/dig...pcm56-pcm58-ad1865-768-khz-2.html#post5538494
Close, but not quite. Sabre dacs require 8x interpolation filters, and we would want it to run at non-standard sample rates in order to allow a more relaxed transition-band between stop-band and pass-band. That makes it possible to better attenuate aliasing, while allowing the filter to be very low ripple in the passband and with very gentle rate of phase shift with respect to frequency (group delay). Sabre dacs also use a standard word length of 32-bits, which is likely the required input format needed for compatibility purposes when the dac is configured for external filtering.
But, the basic idea looks right. Program a low-cost Spartan 6 for double precision PCM interpolation filtering, with a good number of taps available. I image 3lite could make the right kind for us or maybe share some code if he were motivated to do either of those things. Actually, I asked about something to that effect and the response back was basically one of no interest: https://www.diyaudio.com/forums/dig...pcm56-pcm58-ad1865-768-khz-2.html#post5538494
It has been awhile since I did listening comparisons between Benchmark DAC-3 and my modded dac. Decided it was time to try it again in order to keep myself honest about how much progress I have actually been making.
For the first time I think there is probably reason to believe my modded dac may pretty much have all the problems essentially solved except for the PCM interpolation filter. However, I can't be sure until I have such a filter and can use it to directly compare the modded dac with DAC-3.
Still thinking about what to do next. Undecided for now.
For the first time I think there is probably reason to believe my modded dac may pretty much have all the problems essentially solved except for the PCM interpolation filter. However, I can't be sure until I have such a filter and can use it to directly compare the modded dac with DAC-3.
Still thinking about what to do next. Undecided for now.
Hi Mark,
So, how would you rate now your modded DAC vs the Benchark? What is better, what is still missing?
Re future projects, once you will be done with the interpolation filters, I believe you could start a very interesting new project in form of putting into practice all what you have found (and that is a lot!) nut this time around the new AK4599... while getting perhaps even farther by finding its very own tricks and perhaps even beating the Benchmark fare and square.
I use an old AKM DAC chip and love it, love also their openess re specs and what is inside etc.
Just thinking loud, perhaps even more people would follow you and you would then for sure have a unique expertise as ESS and AKM are the 2 makes that rule all the world, unless you want to go very specific ways (R2R, own DAC, medical chips etc.) that may or may not prove better and that are marginal in terms of numbers.
Just thinking loud, as I follow quietly your thread with lot of interest, so many lessons learned... and perhaps still to learn by finding out how the AKM reacts differently to what we discovered around the ESS?
Have a nice WE
Claude
So, how would you rate now your modded DAC vs the Benchark? What is better, what is still missing?
Re future projects, once you will be done with the interpolation filters, I believe you could start a very interesting new project in form of putting into practice all what you have found (and that is a lot!) nut this time around the new AK4599... while getting perhaps even farther by finding its very own tricks and perhaps even beating the Benchmark fare and square.
I use an old AKM DAC chip and love it, love also their openess re specs and what is inside etc.
Just thinking loud, perhaps even more people would follow you and you would then for sure have a unique expertise as ESS and AKM are the 2 makes that rule all the world, unless you want to go very specific ways (R2R, own DAC, medical chips etc.) that may or may not prove better and that are marginal in terms of numbers.
Just thinking loud, as I follow quietly your thread with lot of interest, so many lessons learned... and perhaps still to learn by finding out how the AKM reacts differently to what we discovered around the ESS?
Have a nice WE
Claude
Yep, typo, meant of course 4599
Well, complicate, that's the challenge, and then you want to improve on ESS!
And perhaps some affordable boards will see light, after all ESS is also expensive - or some AKM new chip range?
Yes, love AKM's sound signature - altough given Mark's excellent work one can prove that ESS can sound fine, or indeed perhaps any modern DAC chip if rightly implemented.
Having said that, I would love a DIY AKM DAC to beat ESS on all accounts.
If one wants to avoid tthe AK4499, Why do you suggest AK4493 instead of AK4497 for development work?
Have fun
Claude
Well, complicate, that's the challenge, and then you want to improve on ESS!
And perhaps some affordable boards will see light, after all ESS is also expensive - or some AKM new chip range?
Yes, love AKM's sound signature - altough given Mark's excellent work one can prove that ESS can sound fine, or indeed perhaps any modern DAC chip if rightly implemented.
Having said that, I would love a DIY AKM DAC to beat ESS on all accounts.
If one wants to avoid tthe AK4499, Why do you suggest AK4493 instead of AK4497 for development work?
Have fun
Claude
How about modding the Chinese 9038Pro board discussed in another thread started by idir? Sounds like a natural progression to me (Q2M->Pro). Due to its use of the flagship DAC chip, comparing the modded result with Benchmark DAC-3 seems to me to be more "apple-to-apple." There is even a chance that the modded DAC can beat Benchmark DAC-3, since the latter uses 9028Pro.
Hi Guys,
An AKM dac that works with technology much like that in Benchmark DAC-3 already exists. It is called Crane Song Solaris, and it uses upsampling to 211kHz and external interpolation filtering, same as DAC-3 in those respects.
I think the best thing that could happen next for what I am interested in doing would be if I can find some way to get or make the filter I need. Don't know how long that might take, but the best sounding chip dacs can't be made without it.
In the meantime, I think I mentioned before I have a mostly bare ES9028PRO board made by diyinhk that I have sketched out the schematic for. I have a number of power supplies and maybe another output stage to compare. There is plenty of work to do in those areas.
Getting back to AKM dacs for a moment, the same kinds of ideas I have used with ES9038Q2M can work for AKM too. It is just that that upsampler and dac clocks would have to be the same clock signal, and it would have to be non-standard sample rate to make some room for a transition band in the external interpolation filter. Since the best clocks are only made for standard frequencies unless maybe one can order large quantities, it might be necessary to use a standard frequency clock and synthesize a very low jitter non-standard frequency from that using some combination of low jitter frequency division and multiplication. 6/5 seems to be about the ratio used in DAC-3 and Solaris using crystal clocks to make room for a filter transition band.
EDIT: By the way, it may interest people to know that DAC-3 and Solaris are said to sound very similar by some people. I know a guy who makes a living as a quiet pro audio dealer who basically sells to the industry, primarily to recording and mastering facilities. He carries many different brands and says he can't tell the difference himself between the sound of DAC-3 and Solaris, although I expect there are some people who can.
An AKM dac that works with technology much like that in Benchmark DAC-3 already exists. It is called Crane Song Solaris, and it uses upsampling to 211kHz and external interpolation filtering, same as DAC-3 in those respects.
I think the best thing that could happen next for what I am interested in doing would be if I can find some way to get or make the filter I need. Don't know how long that might take, but the best sounding chip dacs can't be made without it.
In the meantime, I think I mentioned before I have a mostly bare ES9028PRO board made by diyinhk that I have sketched out the schematic for. I have a number of power supplies and maybe another output stage to compare. There is plenty of work to do in those areas.
Getting back to AKM dacs for a moment, the same kinds of ideas I have used with ES9038Q2M can work for AKM too. It is just that that upsampler and dac clocks would have to be the same clock signal, and it would have to be non-standard sample rate to make some room for a transition band in the external interpolation filter. Since the best clocks are only made for standard frequencies unless maybe one can order large quantities, it might be necessary to use a standard frequency clock and synthesize a very low jitter non-standard frequency from that using some combination of low jitter frequency division and multiplication. 6/5 seems to be about the ratio used in DAC-3 and Solaris using crystal clocks to make room for a filter transition band.
EDIT: By the way, it may interest people to know that DAC-3 and Solaris are said to sound very similar by some people. I know a guy who makes a living as a quiet pro audio dealer who basically sells to the industry, primarily to recording and mastering facilities. He carries many different brands and says he can't tell the difference himself between the sound of DAC-3 and Solaris, although I expect there are some people who can.
Last edited:
Hi Claude,
The sound of a dac is highly influenced by the sound of its interpolation filter, and right now my dac still sounds ever so slightly bright or not bright enough depending on AK4137 filter selection. An external interpolation filter after AK4137, but before ES9038Q2M could dominate over the sound of AK4137 filters since the external filter passband would be at a bit lower frequency and I could probably tune it to make the modded dac sound just right. By way of comparison, DAC-3 has a very nice, just about right, warm sound that is just probably good for most listeners. The right filter could well account for that. (There is a little more to the filter story as well, since it reduces aliasing artifacts. A good custom external filter should be able to result in more attenuation of any aliasing, and also prevent most or all intersample over distortion.)
Where the difference may be slightly more in favor of the modded dac is that using phased-locked clocks for the upsampler and the dac chip may result in a bit less jitter as compared to DAC-3's free running separate crystal clocks. Less jitter can make the sound a little more focused and slightly less blurred, even though DAC-3 jitter is already quite good. Its just that the modded dac might turn out to be just a little bit better. Also, I don't know that jitter is fully optimized jet. I think I might be able to get it a little lower with some more work.
Last edited: