DAC linearity test CD

AndrewT said:
Does the "look" correlate with the sound heard?

Originally posted by Bernhard yes
Hi,
If the sound does depend on how these very low level bits look like then your different DACs should sound different when reproducing those LSB squarewaves.
Go on, listen to them after the +40db gain stage. Tell how different they sound.
Do you need a 2bit signal to allow differentiation in the sound from the different DACs? or even more bits? Tell us what is audible.

If this same test were done with 20bit signals/DACs, how many bits would need to be used to allow differentiation in DAC performance at the LSB or slightly above level?
 
I took some photos from a Marantz DC-94 (single TDA1541A, not S1). The first picture is Track 1 - 9, except Track 3. The second is Track 3. The third is Track 10, fourth is Track 12 and the last is Track 15.
 

Attachments

  • gainblock.jpg
    gainblock.jpg
    40.6 KB · Views: 1,353
Bernhard said:
And each corrupted step ( bit ) includes exactly 16 spikes. 🙂
I am not sure about this. I will check the tracks with Waveditor, but each level has 16 identical HEX values, so the problem must be somewhere in the chip. It would be good to know more about the internals of this DAC, especially the active dividers that SoNic_real_one refers to. I will also disable the DEM clock just to see what happens.
 
I measured several TDA1541 and 1541A and they all behave differently. The strange distortion seen on Post #24 occurs at different tracks/bits on every sample of DAC chip. I have suspect that it is somehow related to the DEM clock. Now the task is to find a sample that shows the waveform seen on Post #23 at each bit...
 
Yeah, but both channels behave identically on every DAC I measured so far...

One question: I have some TDA1541As, that I think are fake. The text on them is not painted white, but engraved (I think with laser, very thin lines). The wide bar at the left is made of many parallel thin lines. Also the letter A is shifted up a bit. The code is 41073 and HSH0201 5. There surface seems sandpapered.... I have another four original from a Cambridge Audio CD-3, I will measure all of them in the weekend.
 
Hm, I found some interesting: I replaced the SAA7220P/B in my test Marantz CD-94 to a SAA7220P/A and the waveform of Post #24 moved from Track 3 to Track 2. In general, most bits up to bit 9 are nearly perfect with the B version. Bit 10 to bit 15 looks like Post #25, 26 and 27.

And this is identical with every TDA1541A chip I tried so far (maybe there are some minor level differences, difficult to read off the scope)...

I tried this also on a Marantz CD-74 modified to NOS and digital filter bypassed. Here most bits are perfect or nearly perfect, except bit 9 which shows some oscillation-like noise on the bottom level, but not serious. Another reason for NOS?