ES9038PRO VS AK4497 FIR Filter

[diyinhk]

I just do a quick test on the default FIR filter using the latest ES9028PRO and AK4497, i.e. the default sound output look like.
The ideal output should be a square wave.
The test file is Left channel 1KHz Right channel 2KHz square wave PCM 48kHz sampling rate

AK449X DAC series can output ideal square wave like those NOS R2R DAC when SSLOW=1, (although there is a small overshoot when comparing to ad1865 nos)

es9028pro010

ak4497sslow0sd1slow0

ak4497sslow1sd0slow1

ad1865nos

 

[JensH]

What output level did you use? Was it at or close to maximum?
It looks like the filters are clipping the output, especially with the ES9028PRO.
The title says ES9038PRO, but the measurements seems to be on the ES9028PRO.

 

[Zoran]

It is not strange thing,
1. actually if You emulate passive n times RLC netwotk filter (like Zanden style for the ex.) You will get the similar or same results.
2. IF You plot the (Sin(x))/x function You will get the same or sim. results.
.
In these chips that is done by the software. Other Question is what is more friendly to the ear? Personally I prefer passive way, with no-dsp in the dac-s.
.
DiyHk is right older dacs (r2r) have not present this kind of "phenomenon". For instance I measured almost perfect square @ 20KHz (Yes You reading right 20KHz ) with NOS TDA1540 philips 14bit R2R dac...

 

[Zoran]

Another Q is that almost 100% music "recorded" captured, recently, is from the same sound signature hand-writened by the contemporary ADCs, DSPs, "remastering" etc. etc.
When You play - the output result is squared to worse... .

 

[Zoran]

Diyinhk, could You please post the square of 10KHz od 20KHz from the same dacs?
Thanks

 

[themystical]

These tests are really interesting especially the AK4497 ones. Was the difference in the different FIR filter settings audible in your tests?


Sent from my iPhone using Tapatalk

 

[scott wurcer]

It is not strange thing,

True, a plus to minus full scale transition in one clock period violates Nyquist.

 

[TNT]

True, a plus to minus full scale transition in one clock period violates Nyquist.

Isn't it > one full scale transition but not =< ?

//

 

[chris719]

It's a step function. You couldn't have sampled it in a band-limited system. People love to see that "clean" square wave, but if you have a proper reconstruction filter you will never get it. As Jens mentions, it does look like the Gibbs ringing is clipped.

What do you get from the other filter settings on the ES9028PRO?