Using 20 bit data with a pair of old 16 Bit DAC

[stefanobilliani]

Hello All .

Is it possible to let a pair of TDA1543 play the 20 bit data from a DVD player . Can somebody help with a schematic ?

Thanks

 

[rfbrw]

So long as the data is in the right format for the TDA1543, you do not have to do anything. The 4 LSB's will be ignored.

 

[Glassman]

..and cause an ugly truncation distortion due to undithered requantization

 

[stefanobilliani]

..and cause an ugly truncation distortion due to undithered requantization

INFACT The distorsion goes away as the recording level goes down

The idea is to split the 20 bit data word in 2 words of say 10 bit , one for each 16 bit dac ....

but I need some help to realize it in practice ....

 

[rfbrw]

..and cause an ugly truncation distortion due to undithered requantization

Thats only 4 bits lopped off. There are those that send 24/96 data to the TDA1543 and swear it is the greatest thing ever.

 

[Glassman]

doesn't matter at all how many bits it is, the damage is exactly the same whether you start with 20, 24 or even 32 bits.. luckily this gentleman does not attempt anything of this sort..

btw. good luck calibrating the two DACs to within half the 20th bit level..

 

[rfbrw]

Digital filters often perform straight truncation. There is no way the the sonic effects of a 2 bit truncation are as bad as a 4 or 8 bit truncation.

 

[kevinkr]

Why not just get a modern 24 bit pcm dac like the Ti PCM1794 or PCM1798 or the Analog Devices 24 bit AD1852 sigma delta dac?

The results are liable to be way better with any one of these than with two ancient 16 bit dacs used with a 20 bit word.

The point about calibrating the two is also valid, and I think this will cause a huge linearity issue..

The two chips are unlikely to have identical drift so it is unlikely that the two will hold their calibration over time and temperature either. Such things are regularly done in dc ATE applications, but are subject to frequent calibration and checking. AC I think would open a whole new can of worms.

Edit: fix typo

 

[stefanobilliani]

Digital filters often perform straight truncation. There is no way the the sonic effects of a 2 bit truncation are as bad as a 4 or 8 bit truncation.

Yes of course , that is true if I can use a 24 bit dac and start truncating bits to hear the effects .
The problem is that 18, 20 or 24 ( or more ) bits into a 16 bit dac simply doesn work properly .

Why not just get a modern 24 bit pcm dac like the Ti PCM1794 or PCM1798 or the Analog Devices 24 bit AD1852 sigma delta dac?


Edit: fix typo

Yes Ok thats right . But I am getting rid of the AK 4365 (20 bit sigma/ delta) for a multibit DAC for personal taste reasons . I should try to find an old 20bit AD1862 , but I think it comes down to space matters inside the little DVD player .

EDIT :

.... more over ( and correcting myself ) the dvd player outputs I2S in Philips format , while both PCM and AD think want Sony format . SO ...

 

[Glassman]

Digital filters often perform straight truncation. There is no way the the sonic effects of a 2 bit truncation are as bad as a 4 or 8 bit truncation.

such filters should be prosecuted by the law.. and yes, 2, 4 or 8 bits of truncation - no difference, it only depends on the level at which the truncation happens, not on the number of bits you're loosing..

 

[rfbrw]

but I think it comes down to space matters inside the little DVD player

Unless you plan to make a PCB using chipscale devices and the like, space will always be an issue. You might as well opt for a 20 bit dac and the logic to split the I2S signal.

 

[rfbrw]

such filters should be prosecuted by the law.. and yes, 2, 4 or 8 bits of truncation - no difference, it only depends on the level at which the truncation happens, not on the number of bits you're loosing..

In this context level is irrelevant, it is a straight hardware truncation and 8 bits will be worse than 2.

 

[Glassman]

In this context level is irrelevant, it is a straight hardware truncation and 8 bits will be worse than 2.

the amplitudes of harmonics due to truncation are the same when going from 24b to 16b and from 20b to 16b..
in terms of amount of information lost, 8b is of course more than 2b, the distortion products resulting from truncation are nevertheless equivalent.. prove me wrong, I'd like to know where I'm wrong in my reasoning..

 

[rfbrw]

If you really believe a 2 bit truncation is the same as an 8 bit one there isn't much more to be said.

 

[Glassman]

the white spectrum is 24b truncated to 16b, the red one is the same 24b source dithered to 20b and then truncated to 16b.. to me this doesn't look all that different and it's 4b vs. 8b truncation..

 

[JohnW]

Anybody considered that even "lopping off" one lowest bit you loose dithering information recorded in the source... not to mention 4 or 8 Bits.

And, Glassmans basically correct...

John