Measurements of LJM DAC2496 (AK4393) DAC kit

[audio_tony]

Here are measurements of the LJM DAC2496 (AK4393) DAC kit (still available on Ebay)

I built this around 2018 / 2019 IIRC.

The board has had some mods (specifics not shown here).

My test setup: Asus Xonar STX sound card with a self built buffer / gain box in front of that. Host PC is an Intel i7 running Windows 10 with 32G RAM and SSD - no mechanical disks or CDROM drive present.
I also use an Altor Audio Olivine-2 ADC, however the Asus Xonar was used these measurements.
Sound card input was set to 24bit/96kHz
I used an S.M.S.L. PO100AK as a USB to coax bridge.
I used Arta for the measurements.
Output to DAC was set to 24/192kHz

My DAC assembled:

@AnalogSteph If you thought the CS4398 was bad at 0dB wait until you see this...


Board as available from Ebay +others

Completed project:

THD 0dB

THD -3dB

THD -6dB

THD -30dB

THD -60dB

THD -90dB

Frequency response (sweep):

Frequency response (white noise (smoothed)

Jitter:

Multitone:

THD vs Frequency:

 

[bunkbail]

How much is the whole build?

 

[audio_tony]

How much is the whole build?

It was that long ago that I don't remember. Now I think around £70 to £100?

 

[AnalogSteph]

THD 0dB

The heck is that?!

Something is decidedly unhappy there. Does one of the analog rail regulators start to oscillate or does analog rail loading get into Vref or what? Everything else looks fine.

 

[Palladium]

Nothing audibly wrong besides the truly awful multitone sub-200Hz.

 

[AnalogSteph]

Nothing audibly wrong besides the truly awful multitone sub-200Hz.

That's an FFT analysis artifact, the windowing function gives skirts that are too broad so the tones run into each other.

 

[pkane]

That's an FFT analysis artifact, the windowing function gives skirts that are too broad so the tones run into each other.

Increase FFT size to better resolve in the lower frequencies.

 

[audio_tony]

Updated measurements

Following the somewhat disappointing results (I was sure this had measured better previously) some further investigation found that Windows 10 was the issue here.
So I measured again using Linux.
As can be seen, the multitone and the -0.1dB results are hugely improved.

 

[sarieri]

With today’s technology, this is just waste of sand.

For comparison, this is Ivan’s 9039s dongle cost $100 measured with E1DA ADC. With APU, it measures -125 THD+N.

 

[audio_tony]

With today’s technology, this is just waste of sand.

For comparison, this is Ivan’s 9039s dongle cost $100 measured with E1DA ADC. With APU, it measures -125 THD+N.

I agree, but it performed ok for the time, along with the fact that it's a very simple kit to construct.

Remember that it's 5 years old now.

 

[restorer-john]

What is going on with the PSU? Is that a switch/buck providing HF derived AC/DC to the board?

Someone was a bit worried about reverse polarity with that 1N5408...

Just put it across the input with a series 0R fusible. No voltage drop and way more fun.

 

[audio_tony]

What is going on with the PSU? Is that a switch/buck providing HF derived AC/DC to the board?

Someone was a bit worried about reverse polarity with that 1N5408...

Just put it across the input with a series 0R fusible. No voltage drop and way more fun.

The PSU derives +/- 15v (adjustable) from 5 - 15v input.

Series voltage drop is not an issue, however I also wired the diode like that as I have some AC 'wall wart' power supplies (which are identical to the DC ones I typically use) which might accidentally be connected some day knowing me as I do...

The DC input also passes through an LM317 pre-regulator so the DAC can accept a wide range of input voltages (this and the other kit I posted here were primarily built for bench testing).