Some R2R DAC designs appear to achieve "24bit" processing using two 16bit chips segmented into MSB and LSB paths. One DAC is sent the upper 12bits and the other the lower 12bits.
Is that feasible with the TDA1387?
Will it lower the noise floor and increase dynamic range?
Could it achieve 20bit linearity?
Can the digital data segmentation be processed on available PC software, so a PC with a USB-I2S multichannel device can do the work and save on IC's?
Is that feasible with the TDA1387?
Will it lower the noise floor and increase dynamic range?
Could it achieve 20bit linearity?
Can the digital data segmentation be processed on available PC software, so a PC with a USB-I2S multichannel device can do the work and save on IC's?
Some improvement in noise can be achieved by simply paralleling chips. With some DSP support an array of chips probably can do better than 16bit resolution. I think its highly unlikely the improvement would get anywhere near 24bits. But then no real-world DAC is 24bits in terms of noise, the best of current tech is around 22bits. 20bit linearity might well be possible.
To use TDA1387 in a 'segmented' approach you've described I think won't work as the multiplying factor needs to be very precise and the stability of gain in TDA1387 isn't up to that, nor is its noise. But paralleling outputs and feeding different data to different chips holds promise.
I think its a good idea to use a multichannel USB->I2S interface to provide feeds to an array of chips.
To use TDA1387 in a 'segmented' approach you've described I think won't work as the multiplying factor needs to be very precise and the stability of gain in TDA1387 isn't up to that, nor is its noise. But paralleling outputs and feeding different data to different chips holds promise.
I think its a good idea to use a multichannel USB->I2S interface to provide feeds to an array of chips.
Thanks for clarifying that so clearly. Experts make tricky stuff seem so simple. Backing up the issues you raised with " bit segmentation" design issues is the Denafrips Terminator ladder segmentation design which might kick in at around 6bits? Hardly seems bothering with for just 8dB better SNAID: