I actually finished a USB DAC just like that a few months ago. I took the digital board, analog section and power supply from an old Nakamichi OMS-7 CD Player (laser stopped working) which has a pair of TDA1540D. I use a PCM2706 as a USB to I2S to the translator board. I got the translator board from cdream5.com. I measured it too and dynamic range and distortion are very good for 16-bit. I know the TDA1540 is 14-bit only, but the SAA7030 does some magic to bring the resolution to 16-bit. I'll open up the DAC and post some pictures this weekend. It is not difficult, the power supply in the CD Player provides all the voltages I needed.
The SAA7030 just upsamples ? (and add a filter as well). As far I know you can not ad bits that don't exist, but up/over-sampling may help to push away som unwanting problems to go back in the audio pass band.
IIrc this 13+1 bits dac chip was developed just before it was fixed with Sony up to 16 bits ? (I'm not sure about that, it could be a myth as well)
IIrc this 13+1 bits dac chip was developed just before it was fixed with Sony up to 16 bits ? (I'm not sure about that, it could be a myth as well)
Looks like its making the dac into a simple version of a modern sigma-delta noise shaping architecture.
This 13+1 dac chip is transformed in a 15+1 bits dac chip or it's more the pre filter that manage a bits truncation of the 16 bits material ?
This story is not clear to me indeed. How does it work ?
This story is not clear to me indeed. How does it work ?
I have no idea how you get to 13 + 1 and 15 +1. I could try to figure it out but I fear my brain will explode.
that's what the tda1540 and tda1541 are according papers, not real full 14 bits or 16 bits either.
Certainly changes nothing for your ears though... dunno indeed for the brain 😀 !
Certainly changes nothing for your ears though... dunno indeed for the brain 😀 !
well I don't read bits, not those at least. But please explain about this SAA chip and 16 bits material for the TDA1540; we are all curious t understand here.
I am not sure what there is to explain. There is the general case of digital filters increasing word length which is a function of binary arithmetic. And there is the specific case of the SAA7030 which receives 16/44.1 and sends 14/176.4 to the TDA1540. Phillips simply state that by running at 176.4 their 14 bit dac yields the same performance as a 16bit dac running at 44.1. It is all in the datasheet.
Last edited:
Thanks for that but I really don't understand, maybe the maths, how this can occurs : "There is the general case of digital filters increasing word length which a function of binary arithmetic"
I think I have not the knowledge enough here. So the filter decrease the bit depth of the 16 bits depth cd disc by upsampling it to 176.xx ? Or is it just an upsampling to clean the audio pass band (interpolation?) while still ignoring 2 bits ? (dynamic loss?)
It's above my heads, but when I listen to it no big issues, whatever this strange phrase in the datasheet that is not so clear but for the braves with maths 😕
I think I have not the knowledge enough here. So the filter decrease the bit depth of the 16 bits depth cd disc by upsampling it to 176.xx ? Or is it just an upsampling to clean the audio pass band (interpolation?) while still ignoring 2 bits ? (dynamic loss?)
It's above my heads, but when I listen to it no big issues, whatever this strange phrase in the datasheet that is not so clear but for the braves with maths 😕
If a 16-bit number is multiplied by another 16-bit number, the result is a 32-bit number. That is what happens in a digital filter.
Modern sigma-delta dacs may only have, say, a 5-bit dac in them. But if you operate that 5-bit dac very fast, the average (filtered) voltage output at audio frequencies might be good to 20-bits or maybe 21-bits.
Also maybe helpful to consider, DSD is a 1-bit dac. Its average output after being filtered can be good for much more than 1-bit of accuracy. Right?
Modern sigma-delta dacs may only have, say, a 5-bit dac in them. But if you operate that 5-bit dac very fast, the average (filtered) voltage output at audio frequencies might be good to 20-bits or maybe 21-bits.
Also maybe helpful to consider, DSD is a 1-bit dac. Its average output after being filtered can be good for much more than 1-bit of accuracy. Right?
It seems exactly this happens. 4x oversampling will push down the quantization noise, but the DAC gets 14 bit program material nevertheless. BTW the SAA7030 has 16-bit mode, too.
So following this speed logic ans I assume wecare not talking samples but bits then with a crystal with higher speeds my 16 bits tda could read for example a 18 or a 20 bits recorded material and playback it without bit depht with the adhoc filter chip before ?
I am furious Pedja Rogic and Thorsten Loesch had not said that before. Well I knew from them a tda 1540 or tda 1541 could play higher samples...till 388 K hz in sim mode for the tda1541...but not they could fil with 1untill zero the bits above what their datasheet shows.
Ahaha I m glad my ad1862 can also be 24 bits...so cool.
I am furious Pedja Rogic and Thorsten Loesch had not said that before. Well I knew from them a tda 1540 or tda 1541 could play higher samples...till 388 K hz in sim mode for the tda1541...but not they could fil with 1untill zero the bits above what their datasheet shows.
Ahaha I m glad my ad1862 can also be 24 bits...so cool.
No, dear Iggy.. I have to defend aforementioned friends.. Thorsten, for example, had spent loads of time describing how he would prefer applying oversampling and proper dithering /noise shaping to a good multibit dac, reaching crazy high levels of resolution..
I am too old and lazy, but all this is there waiting to be discovered in the tons of precious material accumulated in diyhifi.org..
I am too old and lazy, but all this is there waiting to be discovered in the tons of precious material accumulated in diyhifi.org..