8 × AK5578EN + 8 × AK4499EQ ADC/DAC Boards

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
USB Board

Here is a very crude sketch of the USB board. It's 35mm × 35mm so that we could put two of them alongside the 70mm × 35mm PSU board within a 1 × 1 brick (aka "unit brick"). It is designed around the XUF232-1024-FB374, which is currently the most powerful chip offered by XMOS (32 cores). It also includes three Si866x low power six-channel digital isolators in order to get up to 18 isolated channels. The red line on the chip outlines the boundary between the two isolated sides of the board. In fact, there will be three isolated sections: one for the XMOS chip, one for the ADAT and S/PDIF outputs, and one for the oscillators (currently shown on the right hand side).

At the top, we have 14 pins (if we stick to exactly 35mm, we'll be limited to 12), and 28 at the bottom, for a total of 42. As a point of comparison, the MCHStreamer Kit has 44 of them, while the DIYINHK has 66, but these do not include the USB input. Therefore, we'll probably need more pins, but we should have space for them on the top left and right sides of the board.

Of course, this sketch does not show any of the capacitors and resistors that we will need. That being said, it's hard to say how many of them are really necessary, because if you take a close look at the PCBs for the DIYINHK and MCHStreamer kits, you'll see that they are quite different on that front: the DIYINHK has very few capacitors and resistors, while the MCHStreamer has many of them. That being said, we're not too worried about these, because we should be able to mount most of them on the other side of the PCB.

Quite clearly, we're really pushing the envelope there, for I don't think there is any such small XMOS board on the market today, and certainly not one with 32 cores. But this is all part of the fun, so we'll give it a serious shot.

Last but not least, the BoM for this board should be well under $75.

Now on to the PSU board...
 

Attachments

  • USB.jpg
    USB.jpg
    529.8 KB · Views: 163
PSU Board

Here is another crude sketch, this one for the PSU board. It is inspired from the OSVA AAPSU01, with an extra LT1965 for the +3.3V supply that we need for the DAC brick. That being said, we'll make sure that our PSU board could be used for many other bricks, including the ADC brick. The board is 70mm × 35mm and has a 35mm × 35mm section similar to the one of the AAPSU01 that is covered by the WE-SHC shield. We also included 6 pins on the input side and 14 pins on the output side. This board is probably the easiest of the three that we have to design, and I am confident that we can fit everything we need on its allocated real estate.
 

Attachments

  • PSU.jpg
    PSU.jpg
    475.8 KB · Views: 156
...with an extra LT1965 for the +3.3V supply that we need for the DAC brick...

Why regulate down to 3.3v for a whole dac brick? There may be a good reason, don't know. However, might make more sense to pre-regulate down to somewhere in the range of 5v to 8v, then do the final regulation locally where it will be needed. At the lower end of that voltage range there will be less power dissipation at secondary regulators. At the higher end of that range they may have a little better performance characteristics.
 
Agree, at least for the digital supplies. For analog supplies, I would say perform listening tests and measurements before finalizing an alteration to the eval board design. If this were a low cost project I wouldn't worry too much, but an ultra-high end part like AK4499 needs some care to enable its intended performance level. An old part like 7805 doesn't actually sound too bad when used for analog audio. Don't know so much about all the new regulators on the market, but haven't had much success so far getting the best sound out of LDOs for analog circuits. That's not for having failed to try either. At the same time, I have not done an extensive effort. There is a paper at Linear Audio you might want to study: https://linearaudio.nl/sites/linearaudio.net/files/v4 jdw.pdf

Regarding capacitor choices, ESR may not be the only consideration. I don't have the desire to scour the eval board schematic looking for each capacitor you might ask about from the BOM. At a minimum, I would want to know where on the schematic it is used.

There is no real external difference between an LDO and a 7805. They are both series regulators, except the 7805 is a design older than I am. The 7805 is not remotely high end and does not belong in a high end product. It is low performance and has a very low bandwidth control loop.

I've read that LA article years ago, it's not very useful.

Edit: I will say, the NJM part appears to be better than an old 7805 according to the datasheet. It's adequate, I am sure. I'd use something else, but it's not terrible.
 
Last edited:
Why regulate down to 3.3v for a whole dac brick? There may be a good reason, don't know. However, might make more sense to pre-regulate down to somewhere in the range of 5v to 8v, then do the final regulation locally where it will be needed. At the lower end of that voltage range there will be less power dissipation at secondary regulators. At the higher end of that range they may have a little better performance characteristics.

The AK4499EQ needs +3V (or +3.3V) for TVDD and the T102 LT1963AEST-1.8 producing the +1.8V DVDD. Are you suggesting that we get the required +3V from a +5V to +8V supply then regulate it down to +3V with a regulator mounted on the DAC board?

I so, this might be possible, and it would make the design of the PSU board even simpler.
 
The AK4499EQ needs +3V (or +3.3V) for TVDD and the T102 LT1963AEST-1.8 producing the +1.8V DVDD. Are you suggesting that we get the required +3V from a +5V to +8V supply then regulate it down to +3V with a regulator mounted on the DAC board?

I so, this might be possible, and it would make the design of the PSU board even simpler.

I might have missed it, but what is your input DC power supply? Is it a switcher or a linear PSU?

Both can be made to work well, but there are different concerns for each. You might want to look into adding a common-mode choke on the input of the PSU board anyway.
 
I might have missed it, but what is your input DC power supply? Is it a switcher or a linear PSU?

Both can be made to work well, but there are different concerns for each. You might want to look into adding a common-mode choke on the input of the PSU board anyway.

For the PSU, I was thinking of building something very similar to the OSVA AAPSU01, with a 20V input coming from a USB-C port. Should I add something else?
 
For the PSU, I was thinking of building something very similar to the OSVA AAPSU01, with a 20V input coming from a USB-C port. Should I add something else?

That's probably fine, if you clone that or use it directly, then add your 3.3V regulator after the 5V output. If you put a USB-C port, you should make sure it's compliant with the USB PD spec as a sink (IMO anyway), which adds extra complexity. Frex is very experienced and knowledgeable, so you can learn a lot from his projects.

Not to dissuade you, but even the power supply alone is a lot of work :).
 
That's probably fine, if you clone that or use it directly, then add your 3.3V regulator after the 5V output. If you put a USB-C port, you should make sure it's compliant with the USB PD spec as a sink (IMO anyway), which adds extra complexity. Frex is very experienced and knowledgeable, so you can learn a lot from his projects.

Not to dissuade you, but even the power supply alone is a lot of work :).

Frex seems to be really good at what he does indeed. I've already learned a ton by following his projects, and I intend to continue doing so. As far as the power supply is concerned, there is no way that I could build one from scratch myself, therefore the closer we stick to what Frex has already designed, the better, and so far, I am just planning to make some slight adjustments to the board's dimensions.
 
Pins

We have started to put together an analysis of the input and output pins we need for the PSU, USB, and DAC boards (input only for DAC board so far). This is a very instructive exercise.

PSU
All we need are three input pins (+20V, GND, ENA). The ENA pin will be used to program or shutdown the PSU board. This is by far the simplest board from a pinout standpoint.

USB
For this one, we've taken some inspiration from the MCHStreamer and DIYINHK kits. The input section is very simple, since it is limited to USB. Because we have such severe space constraints for this board, we will probably not be able to add support for other inputs like S/PDIF.

The output section is much more sophisticated:

- 7 × isolated outputs for I²S
- 10 × isolated outputs for PDM
- 1 × isolated output for S/PDIF
- 1 × I²C interface
- 12 × GPIOs
- +5V and +3V supplies

The 18 isolated outputs are perfect for the 3 × Si866x low power six-channel digital isolators that we have included on our PCB sketch. For their parts, the 12 GPIOs are motivated by the needs of the DAC chip (more on this below). And we have yet to work on adding support for an external clock.

DAC
A quick analysis of the AK4499EQ's datasheet shows that we need:

- 12 inputs for controls
- 12 inputs for audio signals (including clocks)

The former motivate the 12 GPIOs that we are planning to provide out of the USB board. Of course, we could transmit all these control signals over a protocol like I²C, but having dedicated GPIOs should make everything a lot easier.
 
You still don't seem to understand this.

First, jitter on bit clock, word clock, and data do NOT matter one iota for AK4499 and similar parts. Your point is just plain wrong. The reason why is very simple and it's already been explained by me and Mark's quote from the datasheet. You only need to understand where the data enters the analog domain and what clock drives that action.

Second, the test by JA is done at the external interface level. The point of this test is to hamper the MCLK recovery of SPDIF receivers. It doesn't apply here.

I understand exactly what you are saying I just don't agree with it. All you've done is repeat the same thing over and over again without actually providing anything to back it up.

What Mark posted was simply repeating what I had said myself a few posts previously and the picture you linked was, again, repeating what I had said myself. That the master clock is required for the operation of the DS DAC. That is it. You won't get any argument from me on this matter, we are in agreement. But so far all you've said is that the master clock is required for the operation of the DS DAC.

Of course the DAC is going to buffer the entire 32 bit word, most likely the 64 bit stereo word, before actually processing it. Again you'll get no argument from me on this either.

What I am saying is that the timing, the reference that is used to start the conversion process and then when to process the subsequent words is tied to the bit clock, LR clock, data line, not the master clock. How can it not be?

The master clock phase isn't relevant to the conversion process at all, the data sheets say as much. The phase of the clock is exactly what would describe when the data should be processed and without it being aligned with the bit clock that information is lost.

The information within the data being fed into the DAC that contains the precise point in time of when the data should be processed isn't tied to the master clock (the rising edges of the bit clock contain this data and there's no phase relationship between the rising edges of the bit clock and master clock, so the master clock can't contain it). Yet the time of when the data should be processed is very relevant to it's conversion, ESS say as much in their white paper. You can't just start the conversion process at some arbitrary rising edge of the master clock.

And if jitter present on the data stream isn't important to the conversion process why do ESS go to such extreme lengths to remove it? And write white papers on the removal of it?

I am not saying that you are wrong mind you, a lot of what you're saying makes sense, but you aren't providing me with any technical documents to back up what you are saying.
 
DAC Board Audio Outputs

We've added the pins for our audio outputs on the pins analysis sheet. As mentioned earlier, everything downstream of the I-V conversion circuits will be placed on the XLR board, including the summing stage for mono mode. As a result, we just need 9 pins for audio outputs (8 signals and one AVSS). Now, I wonder if we should carry one AVSS ground reference per channel (4 pins instead of 1). Any recommendation?
 
Probably a mistake to decide on using only +20v power now. That would imply you would need a switching supply to produce the -15v best used for some circuitry. Eliminating all switching noise from an ultra low distortion dac output is possible, but non-trivial. So far, I am only aware of Bechmark Media products doing that successfully, and the products in question have selling prices starting above $2,000 per unit. At that price point, and with a lot of engineering effort, SMPS can work for audio. However, most manufacturers have not done as well with it and low cost products have sound quality issues as a result.

Again, I would suggest to get AK4499 eval board working at its best, then try using the prospective power supply with it. After that make a decision if you have the engineering skills to make it work for your project.

ishizeno, I am about a three-hour drive from where you are. Pretty soon, not quite yet, I would be willing to have you come here for part of a day to visit. I could let you see and hear the AK4499 eval board playing high quality DSD512 as well as other formats, and also likely be able to introduce you to Jam, a high-end audio designer friend. If you think you might be interested, we could talk about it some more by offline.
 
Last edited:
Updated USB Board Sketch

Here is an updated sketch for the USB board. Following our pin analysis, we've concluded that we had enough space to include some I²S and S/PDIF inputs as well. This will allow the board to be used for bricks that expose these inputs, even though our standard DAC bricks will not. In order to make that possible, we went from three to four Si866x low power six-channel digital isolators (marked "DI" on the sketch). Also, we've increased the count of GPIOs from 12 to 14 in order to get 2 spare ones on top of the 12 that are required by the DAC board (better be safe than sorry). And we still have quite a bit of room for the regulator(s), capacitors, and resistors that will need to be installed, even though most of them can go on the board's underside. Most importantly, we made sure to group the GPIOs and I²C interface on a single set of headers, while putting all isolated IOs on another set. This will make it easier to connect the USB board to the DAC board. On the sketch, the red line indicates the isolated area.
 

Attachments

  • USB.jpg
    USB.jpg
    638.7 KB · Views: 139
Last edited:
Probably a mistake to decide on using only +20v power now. That would imply you would need a switching supply to produce the -15v best used for some circuitry. Eliminating all switching noise from an ultra low distortion dac output is possible, but non-trivial. So far, I am only aware of Bechmark Media products doing that successfully, and the products in question have selling prices starting above $2,000 per unit. At that price point, and with a lot of engineering effort, SMPS can work for audio. However, most manufacturers have not done as well with it and low cost products have sound quality issues as a result.

I did not realize that. The post introducing the OSVA AAPSU01 mentions that the nominal input can be in the 9-20V range, while "the symmetric outputs voltages can be configured to any values between about +/-6V to +/- 15 V." Did I miss something? I must read like the total newbie that I am right now... ;-)

Again, I would suggest to get AK4499 eval board working at its best, then try using the prospective power supply with it. After that make a decision if you have the engineering skills to make it work for your project.

Agreed. I can't wait to get my hands (and ears) on one...

ishizeno, I am about a three-hour drive from where you are. Pretty soon, not quite yet, I would be willing to have you come here for part of a day to visit. I could let you see and hear the AK4499 eval board playing high quality DSD512 as well as other formats, and also likely be able to introduce you to Jam, a high-end audio designer friend. If you think you might be interested, we could talk about it some more by offline.

I would love that! Thank you so much for the offer. Please contact me at ishi at ishi dot io whenever you want.
 
Redesigned Stack

We have redesigned our stack in order to align our board-to-board interconnects with the AK4499EQ's pinout. The new design will significantly reduce the lengths of most traces and makes room for a proper power section on the DAC board.

The attacked stack shows how two USB boards and two DAC boards will share a common PSU board. As a reminder, the USB board is 1 × 1 (35mm × 35mm), while the DAC and PSU boards are both 2 × 1 (70mm × 35mm). With that in mind, the stack picture shows how the two DAC boards are laid on top of the PSU board, but rotated by 90 degrees.

The DAC picture shows the redesigned layout for the DAC board, with the power section at the top (including three 470μF capacitors and three regulators that were overlooked in the original design), plus an upcoming 5V to 3.3V regulator.

In order to make room for all this, the AK4499 is lowered a bit, allowing for a perfect alignment with two sets of 12 headers, with control inputs on the West side and audio inputs on the East side. The pin analysis that we conducted earlier today confirms that this is an appropriate layout, at the exception of the MCLK pin, which is on the "wrong" side of the chip (I'm not sure exactly why AKM engineers did this).

The new design still makes room for the LT3045-78XX and provides two sets of 6 headers for the audio outputs, one on the North side and one on the South side, directly adjacent to their related OPA1612 OpAmps.

This "crossgrain" laminate structure of the 5 boards kills two birds with one stone: it helps reduce the length of traces and it provides structural integrity for the assembly, without having to use any screws. Overall, I like it a lot better than the previous design.
 

Attachments

  • Stack.jpg
    Stack.jpg
    976.2 KB · Views: 138
  • DAC.jpg
    DAC.jpg
    953.6 KB · Views: 132
Probably a mistake to decide on using only +20v power now. That would imply you would need a switching supply to produce the -15v best used for some circuitry. Eliminating all switching noise from an ultra low distortion dac output is possible, but non-trivial. So far, I am only aware of Bechmark Media products doing that successfully, and the products in question have selling prices starting above $2,000 per unit. At that price point, and with a lot of engineering effort, SMPS can work for audio. However, most manufacturers have not done as well with it and low cost products have sound quality issues as a result.

Again, I would suggest to get AK4499 eval board working at its best, then try using the prospective power supply with it. After that make a decision if you have the engineering skills to make it work for your project.

ishizeno, I am about a three-hour drive from where you are. Pretty soon, not quite yet, I would be willing to have you come here for part of a day to visit. I could let you see and hear the AK4499 eval board playing high quality DSD512 as well as other formats, and also likely be able to introduce you to Jam, a high-end audio designer friend. If you think you might be interested, we could talk about it some more by offline.

I think you can do better with a SMPS than a linear supply in a single box system, actually. Whether everyone can do it properly is a different question. The price of the unit doesn't really have anything to do with the quality of the power supply, only the skill of the designer. Benchmark's parts are garden variety LT buck / boost converters which are only a few dollars each. So, I guess we agree to some extent. I wouldn't automatically rule it out, though.
 
I understand exactly what you are saying I just don't agree with it. All you've done is repeat the same thing over and over again without actually providing anything to back it up.

I posted a block diagram of a relatively modern delta sigma DAC which shows the switched capacitor DAC clocked directly from MCLK. If anyone's repeating the same thing over and over, it would be you :). This isn't a matter of my opinion, the DAC is just clocked by MCLK.

What I am saying is that the timing, the reference that is used to start the conversion process and then when to process the subsequent words is tied to the bit clock, LR clock, data line, not the master clock. How can it not be?

The reference to start the conversion is MCLK. The actual DAC operates at the oversampled rate that is the output rate of the DSM. I am waiting for your explanation as to how the actual converter can be clocked at Fs or 64*Fs (word and max bit clock rates) when the DAC rate is 128 or 256*Fs.

I am not sure why you can't see that the clock for the converter can be phase independent of the clocks used to clock the data into the digital filters. There are these structures called registers / buffers, you know. This is all still in the digital domain.

You are right that the MCLK absolute phase isn't important, but it still must be frequency locked. You are reading into that statement and coming up with an incorrect conclusion.

And if jitter present on the data stream isn't important to the conversion process why do ESS go to such extreme lengths to remove it? And write white papers on the removal of it?

Because ESS uses an ASRC... so there is not an input MCLK accompanying the input stream.

I am not saying that you are wrong mind you, a lot of what you're saying makes sense, but you aren't providing me with any technical documents to back up what you are saying.

This can all be boiled down to one simple yes or no question. Is the actual D->A converter block clocked by MCLK? The datasheet of this AK4499 and several other parts explicitly say Yes. Until the data reaches the point where it is clocked into the analog domain, none of the other clocks are critical and none of the jitter matters unless it is so severe that you violate setup and hold timing. So, you could use the bit clock or word clock to do whatever you want beforehand, but the only clock that matters is the one that puts the data in the analog domain.

Like I said, if this simple yes/no test doesn't do it for you, please start with explaining how the internal DAC can be clocked at a rate higher than the bit clock or word clock if it is clocked by bit clock or word clock as you maintain. If there were an internal PLL or FLL to multiply those clocks, then why would there even need to be an MCLK input?


If you don't want to believe just me:

QuantAsylum QA400 and QA401

https://www.diyaudio.com/forums/digital-line-level/228671-digital-isolator.html#post3344007

Which is the jitter-sensitive DAC input?
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.