I would like to have performance at or above OktoDAC level, and the highest possible for the budget. I care both about objective measurements and subjective impressions. Objective because I would like low distortion (no fake warmth) and very low noise to enable EQ and digital volume headroom. Subjective I care about natural and refined sound, with good stereo placement and instrument textures (e.g. on flutes, hi-hats, rides, brass, slap bass). Not bright, artificial or cold.
Budget would be 1000-2000 euro all in. I may already have a case left where I could build it in. And a few UltraBibs...
A friend is building something custom with Diyink ES9038 multi channel, I will listen to it soon. I would like to stick with AKM if possible, though no hard requirement.
Budget would be 1000-2000 euro all in. I may already have a case left where I could build it in. And a few UltraBibs...
A friend is building something custom with Diyink ES9038 multi channel, I will listen to it soon. I would like to stick with AKM if possible, though no hard requirement.
Hi fedde,
There are a whole lot of things to think about here.
First thing to ask might be if there is a commercial unit within your budget that could work?
If not:
Then there is the question of how you envision using such a dac, will it need SPDIF support, USB via Windows or what OS? Something else?
Architecture (e.g. stereo from PC, DSP and DSD conversion in DAC? User interface to do what? Independent 6 or 8 channels as seen by PC?, etc.)
In other words, let's try to start from the top down, then take a look from the bottom up, then see how things might be made to meet in the middle.
Also, there is the matter of your budget. DAC building is not necessarily a way to save money. It can be an interesting journey and give you something custom, but a Mola dac for 2,000 Euros is probably not going to happen. Maybe not even if you only wanted a simple stereo dac.
There are a whole lot of things to think about here.
First thing to ask might be if there is a commercial unit within your budget that could work?
If not:
Then there is the question of how you envision using such a dac, will it need SPDIF support, USB via Windows or what OS? Something else?
Architecture (e.g. stereo from PC, DSP and DSD conversion in DAC? User interface to do what? Independent 6 or 8 channels as seen by PC?, etc.)
In other words, let's try to start from the top down, then take a look from the bottom up, then see how things might be made to meet in the middle.
Also, there is the matter of your budget. DAC building is not necessarily a way to save money. It can be an interesting journey and give you something custom, but a Mola dac for 2,000 Euros is probably not going to happen. Maybe not even if you only wanted a simple stereo dac.
Yes, OktoDAC would probably fit to what I need. At least feature wise. I heard it in action and it sounds good. Though probably, a good AKM4499 solution could be even better.
I would like to do 64 bit float DSP processing on PC. Crossovers and EQ. IIR and maybe FIR. Custom Windows software, maybe also Linux in future. 6 channel out via USB.
I understand a Mola Mola approach would be very hard to make and expensive...
I would like to do 64 bit float DSP processing on PC. Crossovers and EQ. IIR and maybe FIR. Custom Windows software, maybe also Linux in future. 6 channel out via USB.
I understand a Mola Mola approach would be very hard to make and expensive...
IIRC OktoDAC is based on an ESS PRO series dac chip. Heard Okto was out of production due to parts shortages though. Still, it could be instructive to look at pics of the inside of one. Saw one or more pics before but don't know where to find them now.
Another interesting ESS implementation is the DAC board that can be seen at: https://www.merging.com/products/interfaces/option-cards
Seems to me both OktoDAC and the Horus dac board share something in common: The output stage opamps are more far away from the dac chip than most designs I have seen
A expanded pic of the Horus board attached.
Another interesting ESS implementation is the DAC board that can be seen at: https://www.merging.com/products/interfaces/option-cards
Seems to me both OktoDAC and the Horus dac board share something in common: The output stage opamps are more far away from the dac chip than most designs I have seen
A expanded pic of the Horus board attached.
Attachments
I know about Horus, but it is outside my budget.
Below the internals of OktoDAC. Not sure whether this is the latest version, probably latest version has two caps instead of three. It has the ES9028PRO DAC chip and uses opamps on output (OP1612 or so, I do not remember). Not sure about the clock solution used, perhaps it falls a bit short on this. I care about very low jitter. OktoDAC is still available, though only in small batches from time to time. It was still orderable a few weeks ago...
Is anyone on Diyadio considering to create PCBs for AK4499EXEQ, preferably 4-8 channels?
Below the internals of OktoDAC. Not sure whether this is the latest version, probably latest version has two caps instead of three. It has the ES9028PRO DAC chip and uses opamps on output (OP1612 or so, I do not remember). Not sure about the clock solution used, perhaps it falls a bit short on this. I care about very low jitter. OktoDAC is still available, though only in small batches from time to time. It was still orderable a few weeks ago...
Is anyone on Diyadio considering to create PCBs for AK4499EXEQ, preferably 4-8 channels?
Good. Not quite as I remembered, but helpful to have. A slightly blown up version of the same pic below.
Looks like it has USB in, and what looks like they might be some AES inputs. Hard to tell from the pic but looks like they might be using two NDK SDA clocks? Otherwise the output stage kinda looks pretty straightforward with only I/V conversion and no other active filtering (unless something is hidden under the XLR connectors). Does anyone know for sure the exact architecture and or have a high res pic?
Anyway, if something like this would suit your needs (not sure if you need AES inputs), then it doesn't look too complex.
If I were going to do something like what Okto has done, I would probably want reliable ASIO and WASAPI Exclusive Mode driver support. Only source I know of for a multi-channel USB board like that would be diyinhk. The Thesycon driver is sold separately for a few dollars more.
Do you think this is the type of direction you might want to go?
EDIT: Its that using an ESS chip is probably the easiest way to go, although it may not sound as good as using multiple AKM chips. If, say, a 6-channel USB only dac would suit your needs, then going with AKM might be fairly easy too.
Of course, dacs can be made more complex than Okto is, and that might further help SQ. To say Okto sounds good might essentially be to say that someone hasn't heard anything better so far?
Looks like it has USB in, and what looks like they might be some AES inputs. Hard to tell from the pic but looks like they might be using two NDK SDA clocks? Otherwise the output stage kinda looks pretty straightforward with only I/V conversion and no other active filtering (unless something is hidden under the XLR connectors). Does anyone know for sure the exact architecture and or have a high res pic?
Anyway, if something like this would suit your needs (not sure if you need AES inputs), then it doesn't look too complex.
If I were going to do something like what Okto has done, I would probably want reliable ASIO and WASAPI Exclusive Mode driver support. Only source I know of for a multi-channel USB board like that would be diyinhk. The Thesycon driver is sold separately for a few dollars more.
Do you think this is the type of direction you might want to go?
EDIT: Its that using an ESS chip is probably the easiest way to go, although it may not sound as good as using multiple AKM chips. If, say, a 6-channel USB only dac would suit your needs, then going with AKM might be fairly easy too.
Of course, dacs can be made more complex than Okto is, and that might further help SQ. To say Okto sounds good might essentially be to say that someone hasn't heard anything better so far?
Attachments
Last edited:
To say a bit more, other than selection of dac chips (which set some limit on how good a dac can potentially sound), some of the things that matter for better sound include:
1. Vref design
2. Output stage design (including DC removal)
3. Clocking design (perhaps more so for DSD)
4. Power supplies, especially for the above 3-items.
5. Fully synchronous dac operation, insofar as possible (which implies asynchronous USB)
The above things can elevate dac performance, but at increased cost and complexity.
1. Vref design
2. Output stage design (including DC removal)
3. Clocking design (perhaps more so for DSD)
4. Power supplies, especially for the above 3-items.
5. Fully synchronous dac operation, insofar as possible (which implies asynchronous USB)
The above things can elevate dac performance, but at increased cost and complexity.
Some other things that I would do if I were going all out would include galvanic isolation of I2S, I2C, USB board status signals, and dac MCLK exported to the USB board. Also you can reclock I2S signals after passing them through galvanic isolators. Also would use R-core transformers, and a steel case (aluminum needs 1/2" thick walls to shield as well as something like 18-guage steel -- but leave some space between the dac circuitry and the steel walls since we don't want too much stray coupling there).
Anyway, probably you can start to see a dac project can get time-consuming, complicated, and costly if you want to find out what is the best you can do.
Anyway, probably you can start to see a dac project can get time-consuming, complicated, and costly if you want to find out what is the best you can do.
Last edited:
Yes AES inputs are a nice feature, especially when routed to PC. Though I would have liked an optical in as well, to connect with my TV settop box. Headphone output also nice. And remote control (volume) essential.
I would add to your list:
6. High quality/precision oversampling filters.
7. Good PCB design.
Having good ASIO drivers is handy, but I care more about Kernel Streaming (sounds better IMHO). Indeed Diyinhk board would be considered as USB receiver. Or perhaps a Merging Ravenna ethernet board if I could acquire this.
I did not compare OktoDAC with more high end DACs yet. I may do this begin of next year. But absolutely (and compared with Topping DM7) it sounded fine.
I would add to your list:
6. High quality/precision oversampling filters.
7. Good PCB design.
Having good ASIO drivers is handy, but I care more about Kernel Streaming (sounds better IMHO). Indeed Diyinhk board would be considered as USB receiver. Or perhaps a Merging Ravenna ethernet board if I could acquire this.
I did not compare OktoDAC with more high end DACs yet. I may do this begin of next year. But absolutely (and compared with Topping DM7) it sounded fine.
ASIO may be necessary for Native DSD on Windows.
The built-in filters in AK4499 were fine for DSD256 mode, which is where it sounds best anyway. However, one could always add a Sharc chip or FPGA to add more options. Then reclock after it, would be my suggestion.
Good PCB design is essential, of course. At least 4-layers. 2 oz. copper is nice.
We didn't talk about it yet, but my point of view is that dacs often have a sweet spot for SQ. For AK4499, IMHO it was DSD256. So everything gets converted to DSD256. The output stage can be optimized for DSD256. No need for brick wall output stage filtering for 16/44 CD audio, etc.
The built-in filters in AK4499 were fine for DSD256 mode, which is where it sounds best anyway. However, one could always add a Sharc chip or FPGA to add more options. Then reclock after it, would be my suggestion.
Good PCB design is essential, of course. At least 4-layers. 2 oz. copper is nice.
We didn't talk about it yet, but my point of view is that dacs often have a sweet spot for SQ. For AK4499, IMHO it was DSD256. So everything gets converted to DSD256. The output stage can be optimized for DSD256. No need for brick wall output stage filtering for 16/44 CD audio, etc.
True, DS cannot be done with KS. But I do not want to send DS anyhow from PC as I will do crossover and EQ processing there. So PCM 24 bit or preferably 32 bit it is. 44-192k, up to 384k nice to have.
NOS aside, you will need a brick wall filter anyhow with CD material. At least if you want to avoid aliasing. Regardless of PCM or DSD (and upsampling rate).
If DAC sounds better in DSD256 it could be considered to convert from PCM to DSD on the board. But this would need to be in 6-8 channels and either require a conversion chip or a custom FPGA/SHARC solution. 2nd probably too much effort to all do myself. Though I saw some nice FPGA projects already on Diyaudio.
But the question would also how much a AK4499 would bring over the discrete averaging resistor solutions with DSD256...
NOS aside, you will need a brick wall filter anyhow with CD material. At least if you want to avoid aliasing. Regardless of PCM or DSD (and upsampling rate).
If DAC sounds better in DSD256 it could be considered to convert from PCM to DSD on the board. But this would need to be in 6-8 channels and either require a conversion chip or a custom FPGA/SHARC solution. 2nd probably too much effort to all do myself. Though I saw some nice FPGA projects already on Diyaudio.
But the question would also how much a AK4499 would bring over the discrete averaging resistor solutions with DSD256...
The brickwall filter for CD can be linear phase digital if converting to DSD256. The reason Chord DAVE dac output stage runs up at around 170MHz is so the analog output filter can be very simple.
I do the DSD conversion on board, but IIRC HQ player can do it after PCM DSP is done on a PC. Roon has a fairly decent DSD modulator too, although not as good as HQ Player.
Regarding discrete resistor DSD dacs, don't know of a good low-cost one. Maybe one of Marcel's designs is world class, don't know that either. Do know AK4499 can compete in the big league if carefully enough implemented.
I do the DSD conversion on board, but IIRC HQ player can do it after PCM DSP is done on a PC. Roon has a fairly decent DSD modulator too, although not as good as HQ Player.
Regarding discrete resistor DSD dacs, don't know of a good low-cost one. Maybe one of Marcel's designs is world class, don't know that either. Do know AK4499 can compete in the big league if carefully enough implemented.
Last edited:
True, analog filter requirements are eased up with the upsampling.
Getting a discrete DSD PCB is cheap. But it requires a lot of SMD soldering. A ready made solution would be:
http://www.audiodesignguide.com/DSC2/index.html
It includes a CT7302PL chip for PCM to DSD conversation. I presume this would scale up to six channels with a Diyinhk XMOS board. But no idea how it sounds...
Getting a discrete DSD PCB is cheap. But it requires a lot of SMD soldering. A ready made solution would be:
http://www.audiodesignguide.com/DSC2/index.html
It includes a CT7302PL chip for PCM to DSD conversation. I presume this would scale up to six channels with a Diyinhk XMOS board. But no idea how it sounds...
That particular DSD dac got some unfavorable review posted in its thread. I auditioned CT7320PL and worked with their factory try to get the best out it. John Westlake also wanted to try it. We gave the chip a thumbs down, but more recently some of the guys have said they figured out how to get it working better.
this DAC SACD player is something I really want to hear.
made by engineers of Marantz who used to be Marantz Phillips and who also know analog super well.
im pretty much almost done with DAC companies who don’t know analog well.
