Parallel TDA1541 bit error reducing formula

[noyan]

Hi,
I am trying to parallel 8 or 9 pieces of TDA1541A S1 dacs in non-os mode. Unfortunately i could not find the formula for calculation of bit error reducing by using parallel dacs (2x,4x,8x, so on) and also about reducing of the distortion. Is there any meaning of using 9 pieces instead of 8? Sorry for my bad English.
Thanks a lot!

http://stereomecmuasi.com/forum/view...p=15441#p15441

[SoNic_real_one]

Paralleling of 2 DAC's means doubling of the output current - that is 6dB.
The noise and distortion (TDH+N), because there are not statistically corellated between chips, will increase only with 3dB (some fancy math is involved here).
So doubling the output adds 6-3= 3dB to the current TDH+N ratio.
2 DAC -> 3dB
4 DAC -> 6dB
8 DAC -> 9dB
Increasing the current TDH+N means nothing if you cannot translate it into voltage with the same TDH+N. The I/V stage has a noise floor (and a distortion level), even if you are using a resistor, at room temperature will generate a noise.
So, at some point adding DAC's will improve nothing in the global S/N because you are already below I/V stage noise floor.

[ThorstenL]

Hi,

Originally Posted by noyan

I am trying to parallel 8 or 9 pieces of TDA1541A S1 dacs in non-os mode. Unfortunately i could not find the formula for calculation of bit error reducing by using parallel dacs (2x,4x,8x, so on) and also about reducing of the distortion. Is there any meaning of using 9 pieces instead of 8?

First, the TDA1541A has a Typical SNR at digital zero of 110dB(A).

Secondly, the TDA1541A is typically rated as -102dB(A) THD & N at -60dB signal.

Arguably, these measurements come from a 176.4KHz sample rate signal, but one without noise-shaping etc.

As the absolute maximum dynamic range (peak-peak) of 16 Bit audio is 93.7dB the TDA1541A is basically so close to or well past the limits for 16-Bit audio signals that no material improvement is possible by using multiple devices.

However, this performance is only attained if the DEM Clock is synchronised to the digital data and the power-supply design is of the highest quality. Rather than paralleling ton's of TDA1541A, I would suggest optimally implementing a single one as the better route to get the best possible sound quality from the TDA1541.

Of course, paralleling devices still induces changes in sound quality. I personally find the effects of paralleling DAC's in an otherwise fully optimised system not particularly to my liking.

Ciao T

[jstang]

My take on the issue is a linearity error may be a plus or minus error.... that adding more DACs should average the error rate to lower value...

For example the S chip provide a .5 LSB error rate. plus or minus... so one chip may be -.5 LSB and another may be +.5 lsb at a given value. But additively between the two of them.... no error a the same value. The more you average, the lower the error rate. Even if the spec is only a positive error rate only, the result is the same improvement in linearity averaged over the full range.

But in my mind the real value of multi TDA1541s came with implementing EC-Design's linear interpolation design....that made the whole multi-dac results over the top for me.

jk

[Nazar_lv]

that adding more DACs should average the error rate to lower value...

its trut only for not good dacs and 10-13bits

[SoNic_real_one]

Originally Posted by ThorstenL

First, the TDA1541A has a Typical SNR at digital zero of 110dB(A).
Secondly, the TDA1541A is typically rated as -102dB(A) THD & N at -60dB signal.

That's not true. Just urban legends...

[ThorstenL]

Hi,

Originally Posted by SoNic_real_one

That's not true. Just urban legends...

Hmmm, I have measured the TDA1541A (also S1 & S2) on an AP Two. Have you?

I'd disagree with "urban legend".

Ciao T

[SoNic_real_one]

No, but the manufacturer did.
Philips gave for 1541A TDH+S/N of typical 95dB, channel separation 98dB, S/N at bipolar full scale typical of 104dB.
The S kind are a different story.

[ThorstenL]

Hi,

Originally Posted by SoNic_real_one

No, but the manufacturer ... Philips gave for 1541A TDH+S/N of typical 95dB, channel separation 98dB, S/N at bipolar full scale typical of 104dB.

Yes. Note that I quoted SNR at Bipolar Zero, not full scale (page 7 in the file you link), it is 110dB.

Also note that THD & N at -60dB is typical -42dB.

All DAC's (even the latest ones) degrade considerably near full scale, this behaviour however is not indicative of the behaviour at low levels, which is where we are generally interested in.

Modern day dynamic range measurements are often taken with -60dBfs signals, with these the TDA1541A shows an excellent performance for a 16 Bit chip with good sample-to-sample consistency (we employ it in series production).

The bottom line is that even with your set of numbers the TDA1541A is sufficiently high in performance that paralleling does not improve performance with a real signal.

Ciao T

[SoNic_real_one]

OK, I agree with that, that's what I said too - paralleling more than two TDA1541A (or non-A) it is a waste. Especially because of the 16 bit limitation.