AK4493 DAC

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

First of all I'd like to wish you all a happy new year! To start the new year properly I want to build a good DAC.

I want to use the AK4493 DAC. The reference voltage source I have found to be quite high spec'd is the LTC6655 from Linear technology. The problem I have come across is that the AK4493 requires a really high value capacitor (470uF minimum, value dependant on quality of power supply). The LTC6655(5V version) should not be used with capacitive loads larger than 100uF. At 100uF there already is a noise peak, the ideal capacitor should be around 10uF.

My question now is, how do I optimize the VREF capacitor? Should I just go for the 10uF and put the LTC6655 as close as possible?

Thanks in advance!

- Hidde
 
Does anyone perhaps have recommendations for the output filter?
I can understand that the Q should be ±0.707 and THD as low as possible, but what about Fc and order of filter?

I thought the Fc should be 20kHz, but then the phase response becomes really messed up within the audio band.

In the datasheet it states that you can select the oversampling ratio. Since I am using a 11.2896 MHz oscillator. Would that mean that I should use it in normal mode (256x oversampling) and use 11.2896 MHz as Fs?

Also what kind of order is recommended? I've been reading about 4th order, but 4th order messed up the phase response a lot (though I was tinkering with it at 20kHz).


Thanks in advance :D!
 
How did you progress with this one?

I have been just experimenting with AK4490 and did a search on AK4493. I want to do similar tests on this one at some point as well although will move forward using AK4490 in the current project.

At least on AK4490 I have noticed that it really depends on the voltage source for VREF if big caps are required or not. When using superb LDO, in my case LT3042, big caps did not bring any benefit. And in fact I have seen small deterioration in performance when trying very large 1000+ uF caps. It may be because big caps are not recommended with LT3042 either.

Probably too late but regarding output filters, you do not need to aim fc at 20 kHz - that is the beauty of oversampling DACs. As you see often in datasheets the corner frequencies are somewhere around 100-200 kHz. Then use a filter type you prefer, for example Butterworth and Bessel are common, and orders anywhere from one to four. Two can be a good compromise.
 
How did you progress with this one?

At least on AK4490 I have noticed that it really depends on the voltage source for VREF if big caps are required or not. When using superb LDO, in my case LT3042, big caps did not bring any benefit. And in fact I have seen small deterioration in performance when trying very large 1000+ uF caps. It may be because big caps are not recommended with LT3042 either.

Probably too late but regarding output filters, you do not need to aim fc at 20 kHz - that is the beauty of oversampling DACs. As you see often in datasheets the corner frequencies are somewhere around 100-200 kHz. Then use a filter type you prefer, for example Butterworth and Bessel are common, and orders anywhere from one to four. Two can be a good compromise.

My project is currently on hold. I can only work on it in the weekends as I am working on my bachelors graduation project.
I learned about the filters a while ago yes, but thanks for the effort! I am going to look into Bruno Putzeys way of low distortion lowpass filters.

As for Vref caps, I was going to use Nichicon UKZ 1000uF 50V. They seem like nice caps and I only need 2. A local c0g np0 cap will be used near the pin for the high frequencies.

Unfortunately I do not have the luxury of distortion measurement equipment (this project was part of a DIY audio interface for measurements). I was planning on getting it to work and ask a company I used to work for to verify performance once.
I don't think they appreciate me sitting in their lab and use their equipment for a few days for a hobby project. I'd love to do the research though.
Keep me updated on your project though! I'd love to see the performance.
 
My project is currently on hold. I can only work on it in the weekends as I am working on my bachelors graduation project.
I learned about the filters a while ago yes, but thanks for the effort! I am going to look into Bruno Putzeys way of low distortion lowpass filters.

As for Vref caps, I was going to use Nichicon UKZ 1000uF 50V. They seem like nice caps and I only need 2. A local c0g np0 cap will be used near the pin for the high frequencies.

Unfortunately I do not have the luxury of distortion measurement equipment (this project was part of a DIY audio interface for measurements). I was planning on getting it to work and ask a company I used to work for to verify performance once.
I don't think they appreciate me sitting in their lab and use their equipment for a few days for a hobby project. I'd love to do the research though.
Keep me updated on your project though! I'd love to see the performance.

I will post something when I try to measure it, and test if it behaves similarly to AK4490. I need to first study the differences of AK4490 and AK4493. If I can use the same boards, I may post something soonish, otherwise some time later. Now mostly focusing on finishing the current project. Also working on evenings and weekends and although I can use quite a lot of free time on that, the progress is still quite slow. Things just take time.

I know not all have access to good measurement kit, then it is more like guessing the technical performance. It will be good if the basics are in order. When you want to get it to the top range and squeeze out the last dBs, then it requires more careful considerations and measurements. Even if I have access to the measurement kit after work, it is so time consuming that I do not want to test too many combinations and spend all nights in the office.
 
I will post something when I try to measure it, and test if it behaves similarly to AK4490. I need to first study the differences of AK4490 and AK4493. If I can use the same boards, I may post something soonish, otherwise some time later. Now mostly focusing on finishing the current project. Also working on evenings and weekends and although I can use quite a lot of free time on that, the progress is still quite slow. Things just take time.

I know not all have access to good measurement kit, then it is more like guessing the technical performance. It will be good if the basics are in order. When you want to get it to the top range and squeeze out the last dBs, then it requires more careful considerations and measurements. Even if I have access to the measurement kit after work, it is so time consuming that I do not want to test too many combinations and spend all nights in the office.

Have you considered using a high current (±50mA) as a supply for the AK4493?
You can use one high performance voltage reference and a low noise high current opamp.
Something like the AD8599 would be a good choice. Quite the expensive opamp though. I haven't played with discrete buffers yet.
 
Have you considered using a high current (±50mA) as a supply for the AK4493?
You can use one high performance voltage reference and a low noise high current opamp.
Something like the AD8599 would be a good choice. Quite the expensive opamp though. I haven't played with discrete buffers yet.
In DLCP board they installed NE5532 opamp and 3x red LEDs as voltage reference. It is very unusual. Do you mean the same?
 
Have you considered using a high current (±50mA) as a supply for the AK4493?
You can use one high performance voltage reference and a low noise high current opamp.
Something like the AD8599 would be a good choice. Quite the expensive opamp though. I haven't played with discrete buffers yet.

High current opamp I presume you meant. I did try LM4562 (just because I had it at hand) to see how opamp buffer behaves. That was for VREFs that don't draw much current. There was something odd happening, noise was first very high (although reference was good and low noise) but after adding a large cap at the output again, things got very good. Not as good as LT3042 though. But I assume there was maybe something not quite right in that experiment, and it may need a bit more external components. For example, AK4490 EVB uses opamps buffered with discrete transistors. I'm sure one can make excellent opamp buffer with some more work.

Anyway, although LT3042 is an expensive LDO, using a high performance opamp with voltage reference, external components and large capacitor is not cheap either, and takes significantly more PCB real estate (which is one limiting factor in my tiny DAC). Therefore, I did not see it necessary to look for other more complex solutions.
 
High current opamp I presume you meant. I did try LM4562 (just because I had it at hand) to see how opamp buffer behaves. That was for VREFs that don't draw much current. There was something odd happening, noise was first very high (although reference was good and low noise) but after adding a large cap at the output again, things got very good. Not as good as LT3042 though. But I assume there was maybe something not quite right in that experiment, and it may need a bit more external components. For example, AK4490 EVB uses opamps buffered with discrete transistors. I'm sure one can make excellent opamp buffer with some more work.

Anyway, although LT3042 is an expensive LDO, using a high performance opamp with voltage reference, external components and large capacitor is not cheap either, and takes significantly more PCB real estate (which is one limiting factor in my tiny DAC). Therefore, I did not see it necessary to look for other more complex solutions.

I indeed meant op amps!
Well yes, the VREF doesn't draw much current, but the analog supply needs 50mA MAX according to the data sheet. I figured if I use a two channel op amp I can use one channel for the voltage reference and one channel for the analog supply. That way I have a super low noise analog power supply as well!

Also do you have measurements of the 4562 buffer as a reference? Also what kind of reference did you use? Diodes like Hypex did?
If you don't have measurements, what kind of differences are we talking about? 1-2 dB or 10-20dB?
 
I indeed meant op amps!
Well yes, the VREF doesn't draw much current, but the analog supply needs 50mA MAX according to the data sheet. I figured if I use a two channel op amp I can use one channel for the voltage reference and one channel for the analog supply. That way I have a super low noise analog power supply as well!

Also do you have measurements of the 4562 buffer as a reference? Also what kind of reference did you use? Diodes like Hypex did?
If you don't have measurements, what kind of differences are we talking about? 1-2 dB or 10-20dB?

I have some measurements although as said I didn't dig this deeper.

So LM4562 was sourcing VREFR (ch2 in graphs), the positive input had 10u + 1kohm + 10u filter and the opamp as voltage follower. The reference came from ADP7118 regulator that wasn't sourcing anything else at the moment. VREFL (ch1) was sourced by LT3042.

I have attached THD+N and THD graphs. As you see, THD+N increases at higher levels and is -104 dB at 0 dBFS, compared to -111 dB at Left channel. THD graph reveals it is increased noise floor rather than distortion. Adding large 1000u cap on VREFR pushed it to -108 dB. It is probably pushing the noise down, I don't have graphs from that.

I didn't do more tests on that.

edit: But at some point I had a noise floor of around -90 dBV instead of usual -111 dBV when connected just slightly differently on my test board, but in that case 1000u cap pushed it low again. So there was something going on at some point I didn't quite understand or analysed further.
 

Attachments

  • THDN.png
    THDN.png
    111.3 KB · Views: 724
  • THD.png
    THD.png
    119.6 KB · Views: 701
Last edited:
It is forbidden to put a capacitor from the output of a voltage amplifier to ground if you do not want problems.

Good point! I forgot to mention that. It doesn't really seem right to put a massive 1000u cap directly at opamp output even if it seems to work. You do see circuits posted where there is a large cap but it may be different type of amplifier then. I'm not expert in that matter but it is better to test and analyse a bit more if doing so.
 
With a very large cap the current limitation of the op-amp will limit the voltage excursion to a level that can't generate oscillation but the problem remain and the noise problem that is mentioned above is a good example.
A L-C filter will have an effect in a much wider band than any shunt or LDO regulator.

It depends, there needs to be a stability analysis. A lot of op-amps used in this way will be unstable with a small/moderate capacitive load but will again enter a region of stability with a huge amount.

If you know something more please do a demonstration (mathematically).
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.