AK4499 DAC Design

Hello,
I wanted to design an open source and open hardware DAC around the AK4499 DAC Chip that other members are free to copy, reuse, share, and remix. I previously started a thread for doing this around the ES9038, which piqued my interest at first, but have found that the restrictions that ESS places on their documents go against the values of this project. I am moving forward with the AK4499 As the DAC chip, either using another AK chip or the WM8805. Planning on using either an Op-amp or JFET based I-V stage, and separate LDO's for each of the power supplies on the chip. Unfortunately haven't seen the Eval Board schematics, would appreciate if somebody could post them.
 
Got my hands on the eval board schematics, moving forward with this design. A couple choices right off the bat:

I will be using two AK4499 Chips in dual Mono, starting out with an SE Class A output stage, and maybe going balanced in the future. I will be using software control mode, using a microcontroller to control both chips. I think that I will use an XMOS chip, and crystek oscillators.
 
Dual AK4499 is totally not needed, unless perhaps 8-channels is a must have. Its hard enough to do a good job with one AK4499. Due to the complexity of AK4499 output powering (dac chip and I/V), cost would be expected to go up quite a bit for extra channels.

Also, opamp I/V conversion is probably the only realistic way to go. Could be followed with class A after that. Otherwise, you would probably be increasing distortion by way of trying to reduce it.

Not clear that Crystek clocks are actually better than NDK SDA series, if the implementation is done properly. Some people prefer the sound of NDK.

Regarding SPDIF receivers with master clock recovery, SPDIF often tends to be jittery. Master clock recovery can work well under near-ideal conditions, but ASRC to a local crystal clock reference may be a more reliable choice. (That said, most if not all ASRC chips are limited to 2-channel processing.) AKM's ASRC solution, AK4137, also has the interesting ability to convert incoming PCM audio to DSD, up to and including DSD256. Of course a SPDIF receiver is still needed in some form. TI's SRC4392 is another good stereo ASRC with built-in SPDIF transceiver, but no DSD conversion.
 
Last edited:
Adam,
Please let me know if I seem too critical or not tuned into what you want to do. I would be happy to keep quiet and let you have at it if that's what you want. Reason I brought up a lot of things in the last post was to stimulate thinking about design choices that I would want to think about before finalizing them, that's all.

Wishing the best for your project's success! :)

-Mark
 

samoloko

Member
2006-06-23 10:45 am
you will produce pcb?
can you post shematic
please use true balanced output per channel using 4 chip outputs to corespond opamps (7 single opamps) not just 2 chip outputs like most do

Got my hands on the eval board schematics, moving forward with this design. A couple choices right off the bat:

I will be using two AK4499 Chips in dual Mono, starting out with an SE Class A output stage, and maybe going balanced in the future. I will be using software control mode, using a microcontroller to control both chips. I think that I will use an XMOS chip, and crystek oscillators.
 
Last edited:
Thanks for all the feedback guys, I will be posting schematics here as they come along, definitely want to go for as modular of a design as possible, I'm leaning towards a backplane type system with a well-defined standard so that people can design their own modules, and hopefully have a completely open-source DAC ecosystem. I think I'm going to start with a single DAC chip per your comments @Markw4. I currently have an all-SE system, but plan on upgrading to balanced in the future and want this DAC to be capable of that. I'm still in the very early stages of this project, driven only by the desire to have a fully open-source state of the art future-proof dac, and here are the hurdles I expect to have to deal with:
- USB Input Stage
- Other Inputs ( Definitely SPDIF, maybe AES/EBU + Coax SPDIF )
- Clocking and clock distribution
- Power Supply
- I/V stage
- i2c control of the DAC chip

This is just at the schematic level, haven't even begun to approach layout yet. Let me know if you think I'm missing anything.
 
They are not listed at Mouser and Digikey holds 0 samples.

Digikey is the authorized evaluation board and dac chip retailer for this region. Don't know when they will have chips in stock, supposed to have been a few days ago. They have evaluation boards in small quantities from time to time. Guess AKM may be having chip production problems, don't know.
AKM's wholesale chip distributors also vary by region and are listed on AKM's website. They might be able to find out more than Digikey is saying.
 
Last edited:
For those that have the evaluation board manual, Adobe Reader may not properly display text fonts for the dac chip daughterboard schematic. However, opening the PDF file in Google Chrome does work to display component values.

Unfortunately, the schematic in that area is still not without a few errors. Actual silkscreen labels and components installed on the underside of the daughterboard can be seen in the pics below.
 

Attachments

  • AKevm2a.jpg
    AKevm2a.jpg
    939.1 KB · Views: 1,261
  • AKevm3.jpg
    AKevm3.jpg
    873 KB · Views: 1,044
Last edited:
Even when there were some eval boards in stock, people balked at the price. Thing is, AK4499 can sound marvelous or it could sound mediocre. Depends on implementation of course, and of particular note it depends on how many and what kind of voltage regulators are used. The eval board has banana jacks and jumpers to experiment with the regulators that have the most effect on sound quality. To arrive at a marvelous sounding result, one should either use the eval board or make the equivalent from scratch so that it is relatively easy to experiment.

One thing the eval board design leaves out is a way to easily run the I/V opamps from a dedicated set of voltage regulators. That is probably because of the risk of damage to the dac chip if proper power sequencing is not observed (see AK4499 data sheet, page 88). However, if suitable precautions are taken it does turn out that there is some improvement in sound quality with dedicated I/V stage voltage regulators. One way of implementing a mod for that can be seen in the pic below (although the forum software has compressed the image quality a lot).
 

Attachments

  • AK4499 Eval Board.jpg
    AK4499 Eval Board.jpg
    741.4 KB · Views: 1,071
Last edited:
Regarding the pic of the whole board, a brief description I wrote up for someone previously:

At the left top of the pic you can see an external 5v and 3.3v power supply board for the (mostly) digital circuitry. Also, at top-center are 3 wires (colored yellow, orange, and green) that is the I2C bus connection. An Arduino not shown in the pic programs the dac registers.

In the middle of the pic offset to the right is a blue PCB which is JLSounds I2SoverUSB (better than various other USB boards I have tried, I2SoverUSB has less jitter to my ears).

Towards the lower left of the pic you can see a brass bar from left to right with some little wires coming out of the board to a 3-pin connector. That is a patch to power the I/V opamps directly for 2 of the dac channels. I left the other 2 channels closer to stock in order to compare differences in sound quality.

The trick to making this thing sound good is to power everything with dedicated voltage regulators and use the right type of regulator for each load. LDOs for the digital and RF stuff. The I/V opamps and the Reference Voltage (similar to AVCC, but +5v) need to have regulators that sound good for analog audio (which is kind of an interesting topic, but probably best left for another time).

There are also some extra electrolytic caps that can be seen soldered on at the top and bottom of the dac chip daughter board. Those reduce LF distortion (per the eval board manual, page 28), but would flatten the dynamics if it were not for a trick circuit I added to prevent the flattening. Trick is confidential, the IP belonging to Jam, so I can't say more about that. Presumably, the dynamics flattening problem is why AKM didn't add the caps themselves.

There are some added little blue trim pots and resistors on the dac daughter board which are for trimming the I/V stage output DC offset to 0v (as per AK4499 data sheet, section 10.4.1, Voff = 1.9V). Same thing some people are doing with Sabre dac I/V stages. Seems to help lower distortion a little.

With AK4499 configured as shown in the pic, it sounds clearly better than any Sabre dac I have heard, including a Benchmark DAC-3 we have here for comparison. And Jam said it may be the best dac he has ever heard, which would include some that are quite expensive >$10k.
 
Last edited:
With AK4499 configured as shown in the pic, it sounds clearly better than any Sabre dac I have heard, including a Benchmark DAC-3 we have here for comparison. And Jam said it may be the best dac he has ever heard, which would include some that are quite expensive >$10k.
How was the listening comparison set up? How was the visual bias compensated?