DIYINHK DAC Measurements

[Kaameelis]

Thank you. Somehow I missed this posts end.
Do you know, had ES9038Pro board same one opamp output schematic as on Figure 3 of ESS Application Note?
As I understand with this schematic 3.3V AVCC noises are going directly to the output and current output configuration (what give better parameters) cannot be used.
Can on ES9038Pro board be implemented THD compensation and other configuration writing form Arduino as on ES9016?

 

[meooms]

Thank you. Somehow I missed this posts end.
Do you know, had ES9038Pro board same one opamp output schematic as on Figure 3 of ESS Application Note?
As I understand with this schematic 3.3V AVCC noises are going directly to the output and current output configuration (what give better parameters) cannot be used.
Can on ES9038Pro board be implemented THD compensation and other configuration writing form Arduino as on ES9016?

The THD compensation can be accessed through the I2c registers using a microcontroller, like the ES9016. These are available on the DIYINHK board. It would take very low noise measuring equipment to find the optimal settings, though. The standard IV stage on the DIYINHK is very basic, but they work ok. For reaching optimal results, I would recommend using the provided pins to use a separate IV stage, like the ones from Twisted Pear audio. This works sounds great, dead silent, although I have no objective measurements on SINAD and THD.
I am in the process of trying to set the THD registers by ear, but I'm afraid that the results are to subtle to actually hear. But still working on the implementation.

 

[Kaameelis]

You can use for THD measurements Cosmos ADC.
As I understand ES9016 had no THD compensation, or at least data about it is not public.

 

[Gorgonzola]

I never cease to be amazed and envious of the DIY results knowledgeable, skilled people can achieve.

 

[mdsimon2]

You can use for THD measurements Cosmos ADC.
As I understand ES9016 had no THD compensation, or at least data about it is not public.

Yes, seems like this would be an ideal tool for the job, would also be very interested to see some measurements of the ES9038pro with an external IV stage. The hope would be it would do much better than the stock voltage mode output but I do not have tons of confidence in the DIYINHK board.

Michael

 

[mdsimon2]

I should probably add that after experimenting with these boards for a while I never finalized any projects with them. By the time you add power supplies, a case and potentially external IV stages you are easily in the same price range as a good pro audio multichannel DAC like the MOTU Ultralite Mk5. In addition I see very little reason to worry about interfacing a miniSHARC via I2S when CamillaDSP is so easy to use with a USB DAC.

Of course YMMV.

Michael

 

[Kaameelis]

I had not still get to the point of killing idea to use ES9038pro board.
Do you know how configuration to ES9038pro is written when used with DXIO board only? What is standard configuration, current or voltage ouput? Can configuration of ES9038pro be changed thru DXIO driver in this setup?
Had DXIO driver multi channel ASIO supported?

 

[mdsimon2]

I had not still get to the point of killing idea to use ES9038pro board.
Do you know how configuration to ES9038pro is written when used with DXIO board only? What is standard configuration, current or voltage ouput? Can configuration of ES9038pro be changed thru DXIO driver in this setup?
Had DXIO driver multi channel ASIO supported?

I know that you are also active in the thread on DIYAudio where I just posted some simple measurements of the ES9038, but for other others see this thread -> https://www.diyaudio.com/community/...-power-application.380293/page-3#post-6898249. That thread also discusses the rather odd I/V stage that these DIYINHK DACs are using.

As mentioned just above I do not recommend either the ES9016 or ES9038 as I think the overall cost of properly implemented DIYINHK DAC will be very near a commercial offering and I am not convinced the performance will be as good. If I had to choose a DIYINHK DAC I would choose the ES9016 over the ES9038 simply because the measured performance is better, the ES9038 has higher distortion and inconsistent performance between channels as shown in the measurements I posted at DIYAudio. Maybe this changes if external I/V stages are used but I do not know.

In terms of configuration I personally do think that the DXIO (XMOS) does anything. It does have SDA/SCL pins but the instructions clearly say these are for use with a display and do not interface with the I2C control of the DAC. I have no idea if ASIO is supported, I only used the DXIO with Mac and Linux where it is class compliant.

Michael

 

[Kaameelis]

OK, but this is only DAC with ESS chips I had find, what had 8 non-balanced outputs and (multichannel?) DSD playback.

 

[mga2009]

I know that you are also active in the thread on DIYAudio where I just posted some simple measurements of the ES9038, but for other others see this thread -> https://www.diyaudio.com/community/...-power-application.380293/page-3#post-6898249. That thread also discusses the rather odd I/V stage that these DIYINHK DACs are using.

As mentioned just above I do not recommend either the ES9016 or ES9038 as I think the overall cost of properly implemented DIYINHK DAC will be very near a commercial offering and I am not convinced the performance will be as good. If I had to choose a DIYINHK DAC I would choose the ES9016 over the ES9038 simply because the measured performance is better, the ES9038 has higher distortion and inconsistent performance between channels as shown in the measurements I posted at DIYAudio. Maybe this changes if external I/V stages are used but I do not know.

In terms of configuration I personally do think that the DXIO (XMOS) does anything. It does have SDA/SCL pins but the instructions clearly say these are for use with a display and do not interface with the I2C control of the DAC. I have no idea if ASIO is supported, I only used the DXIO with Mac and Linux where it is class compliant.

Michael

Thanks for your findings and measurements!

I would like to know, which commercial offering do you think is comparable to the 9016. AFAIK there are no "cheap" 8 channels DACS available, as MiniDSP killed most of their lines, Okto is sold out, etc. Maybe some USB soundcard?

 

[mdsimon2]

Thanks for your findings and measurements!

I would like to know, which commercial offering do you think is comparable to the 9016. AFAIK there are no "cheap" 8 channels DACS available, as MiniDSP killed most of their lines, Okto is sold out, etc. Maybe some USB soundcard?

What is cheap to you? Focusrite 18i20 2nd gen is $340 and has 10 channel analog output, ADAT output and SPDIF output. I made some measurements of it here -> https://www.audiosciencereview.com/...-for-a-quality-stereo-setup.26753/post-964029, noise performance is definitely rather poor but in the same range as the old miniDSP U-DAC8 with much better distortion performance. My other compliant is that on Linux it does not seem to play nice when using the SPDIF input but I might try and troubleshoot further by installing the device specific driver -> https://github.com/geoffreybennett/scarlett-gen2/releases (which will be built in to the 5.15 Linux kernel).

Michael

 

[mga2009]

What is cheap to you? Focusrite 18i20 2nd gen is $340 and has 10 channel analog output, ADAT output and SPDIF output. I made some measurements of it here -> https://www.audiosciencereview.com/...-for-a-quality-stereo-setup.26753/post-964029, noise performance is definitely rather poor but in the same range as the old miniDSP U-DAC8 with much better distortion performance. My other compliant is that on Linux it does not seem to play nice when using the SPDIF input but I might try and troubleshoot further by installing the device specific driver -> https://github.com/geoffreybennett/scarlett-gen2/releases (which will be built in to the 5.15 Linux kernel).

Michael

Thanks for your recommendation!

I consider cheap around USD$400 and below, something that you can add a PC + AMPS and gives you something similar to a DIY AVR (considering all your media comes from your PC -like games, movies, music, etc.-)

Also, I noticed in your Scarlett review that you mention that the interface has a volume knob for ALL output channels, is this correct? That would be a big bonus!

The only interface I had was a Behrinfer 402HD (IIRC) and I coulnt make it output 4 channels at the same time. The Motu and Scarlett you mention, can output all 8/10 channels at the same times and these can be routed and processed using EQ APO?

 

[mdsimon2]

Yes, on the Focsurite the volume knob can be set to control analog output channels 1-2, 1-4, 1-6, 1-8 or 1-10 (all), see screenshot below from Focusrite Control.

I have not used it on Windows but on Linux I have used CamillaDSP to do channel routing and EQ on both the Focusrite 18i20 2nd and MOTU Ultralite Mk5, I see no reason why EQ APO would not work but I am not very familiar with Windows audio.

Just my two cents but I think the MOTU Ultralite Mk5 is worth the extra money over the Focusrite for much better noise performance and a smaller form factor but if you are price conscious the Focusrite is not bad as long as you understand its limitations.

Michael

 

[Kaameelis]

The only interface I had was a Behrinfer 402HD (IIRC) and I coulnt make it output 4 channels at the same time. The Motu and Scarlett you mention, can output all 8/10 channels at the same times and these can be routed and processed using EQ APO?

I had used EQ APO with Voicemeter (to get rid of Windows OS audio handling) and ESI GIGAPORT eX, it is possible to define on EQ APO how channels are routed.

 

[mga2009]

Yes, on the Focsurite the volume knob can be set to control analog output channels 1-2, 1-4, 1-6, 1-8 or 1-10 (all), see screenshot below from Focusrite Control.

View attachment 177437

I have not used it on Windows but on Linux I have used CamillaDSP to do channel routing and EQ on both the Focusrite 18i20 2nd and MOTU Ultralite Mk5, I see no reason why EQ APO would not work but I am not very familiar with Windows audio.

Just my two cents but I think the MOTU Ultralite Mk5 is worth the extra money over the Focusrite for much better noise performance and a smaller form factor but if you are price conscious the Focusrite is not bad as long as you understand its limitations.

Michael

I had used EQ APO with Voicemeter (to get rid of Windows OS audio handling) and ESI GIGAPORT eX, it is possible to define on EQ APO how channels are routed.

Thanks for your replies!

CamillaDSP is used only as a DSP core? Like having CamillaDSP running in a tinyPC/Raspberry and use the interface's ADC (which I assume are far much better than the ones from my ADAU1701) to process the audio from the preamps from an AVR?

Windows OS audio handling is really that BAD?

 

[mdsimon2]

Thanks for your replies!

CamillaDSP is used only as a DSP core? Like having CamillaDSP running in a tinyPC/Raspberry and use the interface's ADC (which I assume are far much better than the ones from my ADAU1701) to process the audio from the preamps from an AVR?

Windows OS audio handling is really that BAD?

Yes, that is exactly the setup I use. RPi4 running Ubuntu Server 21.10 64 bit and CamillaDSP, CamillaDSP handles the channel routing and DSP. RPi4 also gives you options with displays and IR receivers which is nice and I like having everything run a small dedicated system.

I have not tried the analog inputs on the Focusrite but have done this on my MOTU M4 and Ultralite Mk5 without issue and see no reason why it would not work on the Focusrite, I'll give it a try this weekend. One thing that you need to be aware of is latency, CamillaDSP will add some latency, this largely depends on the chunk size you are using but you can expect somewhere between 10-20 ms in typical setups. I use CamillaDSP on my TV based setup without noticeable lip sync issues so to me it is not a big deal, just something to be aware of.

The link in my signature has a tutorial on how to setup CamillaDSP, it might look scary but it is not that bad if you take it slow.

To me Windows audio sounds a lot more complicated than Mac or Linux but I haven't had a Windows computer in years so I am not speaking from experience.

Michael

 

[Kaameelis]

Yes, there are some problems on Windows audio handling: https://www.audiosciencereview.com/...nding-the-windows-audio-quality-debate.19438/
It is long time know Windows own audio had problems, to overcome them is good to use ASIO drivers.

 

[mdsimon2]

Slight update on the ES9016, with the help of @meooms, @Skeptischism and Markw4 at DIYAudio I tried out an Ian Canada I/V STD MkII standard OPA I/V stage and initial results are rather promising. With the stock NJM2114D op amps I got improved THD performance from -104 dB to -106 dB at 2 V output voltage from the balanced outputs. Swapping out the stock op amps with OPA1612s got me down to -110 dB THD which I believe is limited by the MOTU M4 ADC. I have a Cosmos ADC on the way which should allow for more meaningful comparison.

I wanted 4 V from the balanced outputs so I removed half of the feedback resistors (R1/R16/R19/R33) and am now sitting at just under 4 V although the 3rd harmonic increased quite a bit.

The I/V stage also drastically lowered the turn on / off pops to the point where I could probably live with them. Thought I was done with this thing but may get a 3 more I/V stages and turn it in to a one box streamer + DSP + DAC. Still need to point out that this DAC is nowhere cost competitive especially once you add $160 in I/V stages and even more on fancier op amps but it feels like I am 75% of the way to making something useful.

I've attached FFTs with the OPA1612s at 2 V and 4 V for reference.

Michael

 

[Kaameelis]

Interesting. More interesting is to see ES9038 board measurement with Ian Canada I/V and OPA1612.

 

[mdsimon2]

01/30/2022 EDIT: I revised this post to show measurements made with the Cosmos in stereo mode. The first set of measurements I posted had the Cosmos in mono mode but rather than using a splitter I was taking 2 channels from the DAC similar to what Archimago did with his Topping D10 in his Cosmos ADC review. At first glance the difference between mono and stereo seemed to make sense, THD was staying the same and I was seeing an improvement in noise. However, the more I looked in to the difference between mono and stereo the more suspicious I became. In particular when measuring my Focusrite 18i20 2nd gen I was seeing a decrease in noise from -98 dB to -101 dB at 2 V output, given that the Cosmos has substantially higher SNR than the 18i20 a 3 dB increase in the Cosmos SNR as a result of switching from stereo to mono should have resulted in essentially no difference in measured noise, not a 3 dB improvement. I pulled out an XLR splitter and did NOT see the same decrease in noise when going from stereo to mono. From a THD perspective the results are worse with the XLR splitter for almost every DAC I've tested as the input impedance is halved. As a result I plan on using stereo mode as my default for all future measurements, I give up a few dBs of ADC SNR but that will really only come to play for DACs that are really pushing the envelope and I have much more confidence in measurements made this way.

My Cosmos ADC (B grade) arrived this week and I've been re-measuring a bunch of DACs including these DIYINHK DACs. See below for some updated FFTs with the Cosmos. All these DACs used the same power supply that is described in the OP. The FFTs represent the best measured THD+N with the Cosmos in stereo mode.

ES9016, stock I/V, NE5532, 1.938 V (full scale), 3.5 V ADC input setting. -109.1 dB THD, 112.4 dB N, -107.5 dB THD+N.

ES9016, Ian Canada I/V, NJM2114D, 2 V, 3.5 V ADC input setting. -106.1 dB THD, -109.6 dB N, -104.5 dB THD+N.

ES9016, Ian Canada I/V, NJM2114D, 4 V, 6.7 V ADC input setting. -105.4 dB THD, -112.4 dB N, -104.6 dB THD+N.

ES9016, Ian Canada I/V, OPA1612, 2 V, 1.7 V ADC input setting. -122.1 dB THD, -112.4 dB N, -112.0 dB THD+N.

ES9016, Ian Canada I/V, OPA1612, 4 V, 3.5 V ADC input setting. -124.4 dB THD, -115.7 dB N, -115.2 dB THD+N.

ES9038, Ian Canada I/V, NJM2114D, 2 V, 3.5 V ADC input setting. -108.1 dB THD, -108.1 dB N, -105.1 dB THD+N.

ES9038, Ian Canada I/V, NJM2114D, 4 V, 10 V ADC input setting. -99.2 dB THD, -109.9 dB N, -98.9 dB THD+N.

ES9038, Ian Canada I/V, OPA1612, 2 V, 2.7 V ADC input setting. -110.3 dB THD, -108.4 dB N, -106.3 dB THD+N.

ES9038, Ian Canada I/V, OPA1612, 4 V, 3.5 V ADC input setting. -112.0 dB THD, -113.8 dB N, -109.8 dB THD+N.

A few random thoughts on the measurements.

-Stock ES9016 does surprisingly well and is quite low noise
-Cosmos shows some power supply spurs around the fundamental much more clearly than the MOTU
-Seems like the NJM2114D may not like the low input impedance of the Cosmos
-No reason to go with the ES9038 over the ES9016, I am also using the "good" channels of the ES9038, expect inner channels to be worse
-ES9016 + Ian Canada I/V + OPA 1612 is the second best 8+ channel DAC I have measured so far, only surpassed by the Okto DAC8 pro
-Still think that once you add power supplies, I/V stages, op amps, deal with turn on / off pops and add an enclosure the cost is quite high and you are better off with a commercial offering like the MOTU Ultralite Mk5 or Okto DAC8 pro

Michael