48bit parallel DAC schematics

[Comzone]

Do anyone has some schematics about 2 parallel DAC acts as 48bit DAC? (Not a just stack it)
Thanks!

[weissi]

Ha, have you ever calculated the LSB voltage levels, even of a 24Bit DAC? I guess they would be in the nV range for a 48bit DAC... I don´t think that CMOS (like SigmaDelta Converters) or even Bipolar designs can handle this voltages at the required speed...
Of course, the technical possibility is there, but don´t ask me about the costs of an REAL 48bit system......

[AuroraB]

1 Vout in 48 bits gives LSB voltage of 3.55 E-15............
that is 3.55 femtoVolts... ...............

[rfbrw]

There is nothing to stop one building a multibit dac where the the current sources are weighted in ten's of mA except common sense. The downstream electronics will have to all valve though.

[OliverD]

Two 24 bit parallel DACs make one 25 bit DAC. The famous Burr Brown PCM1704 is implemented as two 23bit DACs on one chip.

If you "stack" them so that the upper chip delivers bit 47 to bit 24 and the lower chip bit 23 to bit 0, you would need a way to set the reference of the two chips individually. But then, even if the upper had a reference voltage of 10kV, the other one would need 600µV (for MSB, the LSB being 35 pV), both of which is definitely out of range for a small silicon chip.

Paralleling DACs improves signal-to-noise ratio and linearity and is implemented in some high end gear (both DIY and commercial). 24bit resolution represents a dynamic range of 144dB. With the additional use of dither, this is well below the noise floor of current analog circuitry. I once read that Chesky records are using custom-made 27bit-ADCs, but they are connecting tube microphones directly to them, without further amplification. 27bit is about as high you can go with today's audio technology.

[Pjotr]

Besides that AMT, no 24 bit ADC or DAC passes the 20bit boundary in reality. The lower bits just produce noise. You only get that 24 bit out of it if you average many 100’dreds of times which is only possible with harmonic steady signals like pure sines. But not with momentarily varying signals like music signals. And even then, no 24 bit ADC or DAC is monotonic over the full dynamic span, only around the zero crossing.

[Steven]

I once heard a Crystal engineer say at an AES exhibition about a new DAC (or was it ADC?) they presented that the new device was 20-bit, consisting of 18 technical bits and 2 commercial bits.

Steven

[Comzone]

That's mean 48 bit parallel DAC is not worth to make?

Is it the marketing strategy of the Hi-end DAC factory?

[Steven]

Originally posted by Comzone
That's mean 48 bit parallel DAC is not worth to make?

Is it the marketing strategy of the Hi-end DAC factory?

Well, I doubt it is worth to make, but you certainly cannot make. For every bit you add the accuracy and the complexity of the converter doubles. Manufacturers spend a lot of effort in development of converters with more significant bits, but at best they can reach something like one bit extra every year or every couple of years (something similar as Moore's law). No way you can make a technoly jump of more than decades, if not centuries, and make your own 48-bit converter from normal components, currently available.

Part of the more bits for the buck is certainly pure marketing. You have to read the specs carefully to know how many bits are really significant.

Steven