Modern Multi-Bit DAC vs Delta Sigma, specifically AKM's newest flagship, but also others

[Philbo King]

Sorry for veering off-topic...
This thread is sort of interesting to me. I was involved in designing and prototyping an optical D/A about 10 years ago using 10+ GHz pulsed lasers through fiber cable to directly sample RF, reaching the design goal of an ENOB of 10 bits. (Patent 8928510). It was a radically different approach to A/D. I haven't kept up to date on the progress since I retired, but one eventual goal was to implement the design on a silicon photonic chipset.

 

[KSTR]

So how does that work? Clearly the AK4191 generates multibit PWM data, not PCM. The modulator is there, not in the AK4499EX.

As explained in my previous post, the AK4499EX is a PCM flash DAC with parallel 7-bit input interface (multiplexed when used for stereo input).
Inside the DAC, the 7-bit PCM is expanded to 128-bit thermometer code with DWA.

The output from the AK4191 is also PCM with just 7 bits but a high sample rate, produced by the multibit DS-modulator.

 

[voodooless]

The output from the AK4191 is also PCM with just 7 bits but a high sample rate, produced by the multibit DS-modulator.

I think you might be wrong. An older product brief shows it's a 3rd-order modulator, which would correspond well to a 7-level output stream. Otherwise, you'd have to downsample the DS data again to then expand to 7-bit PCM. No such mechanism is described in the datasheets.

 

[DonH56]

I am not aware of any direct link to modulator order and DAC levels... A 7-bit DAC should have 127 levels, with oversampling and delta-sigma modulation pushing up the data rate and pushing the quantization noise to higher frequency to be filtered away later.

 

[KSTR]

I think you might be wrong. An older product brief shows it's a 3rd-order modulator, which would correspond well to a 7-level output stream. Otherwise, you'd have to downsample the DS data again to then expand to 7-bit PCM. No such mechanism is described in the datasheets.

Block diagram from the full AK4499EX datasheet clearly states its a "pure" DAC with a two's complement PCM parallel input (with no processing whatsoever except for the DWA and thermometer decoder), which implies that the output from the AK4191 also has the same format.

That's the whole idea of the AK4499EX: an excellent quality DAC output core which can be used in many ways, for example you can use it for a fast arbitrary waveform generator with 7-bit resolution, being fed with direct PCM.

 

[voodooless]

Mkay, seems I was wrong. Thanks guys!

 

[KSTR]

Mkay, seems I was wrong. Thanks guys!

No problem. I do admit that AKM datasheets are not always straight-forward to read/understand and often are a bit vague terminology.

Fun fact: In the above block diagram from AK4499EX datasheet there still is a string "Modulator" above the string "DWA" but it is whitened out. Looks like there have been misunderstandings what this thing actually does even inside AKM ;-)

 

[mocenigo]

For multibit DS, you’ll need more than on level to be “on” at a time. You can’t do that with 127 levels and 7-bits.

The 7 bits are not the thermometer encoding. That's a binary enoding (two's complement, in fact, as described in the data sheet).
Oh yes, the 4499EX expands 7 bits to a 128 (actually, I think 127) levels of "thermometer" encoding, but the assignment to the various resistors is scrambled in a simple way. I am not sure what order of DWA (data weighted averaging) are using, but it is probably 2 or 3.

 

[mocenigo]

I don't think it does... The datasheet for the 4499EX does not mention DSD once.
The datasheet for the previous 4499, mentions it like 70x

The 4499EX does not process DSD. But the 4191 may get a DSD input and output either all-0 and all-1 values to the 4499EX, effectively keeping it DSD, or perform some other processing (like volume control).

 

[mocenigo]

This really makes me mad. I'm not mad because the E70 Velvet sounds great, but I'm mad at the deception. If it's not a premium switched resistor stereo DAC, don't call it a switched resistor stereo DAC, call it a DS DAC with an extra bit or two...

But it is a switched resistor DAC!

 

[mike7877]

But it is a switched resistor DAC!

Technically, but it's not used as a switched resistor DAC, because you'd get only 7 bit resolution. It's a switched resistor DAC for use only with delta sigma processing if you want to use it for music.

 

[mike7877]

The 4499EX does not process DSD. But the 4191 may get a DSD input and output either all-0 and all-1 values to the 4499EX, effectively keeping it DSD, or perform some other processing (like volume control).

That was my bad - I think when I wrote that didn't know the 4499 had been split into two chips yet

 

[solderdude]

The DAC chip of the set could be used as a volume controlled DSD DAC where selecting the 7 bits would be a 127 level volume control. Needs some processing before that DAC chip.

 

[mike7877]

The DAC chip of the set could be used as a volume controlled DSD DAC where selecting the 7 bits would be a 127 level volume control. Needs some processing before that DAC chip.

That's an excellent idea!

edit: if you had two, could you put a resistor in front of one to make its largest resistor the next sstep down from the least significant bit of the other and make a 14 bit DAC?

 

[mocenigo]

Technically, but it's not used as a switched resistor DAC, because you'd get only 7 bit resolution. It's a switched resistor DAC for use only with delta sigma processing if you want to use it for music.

Well, it IS a DAC based on resistors. That are switched. In parallel (not in a R2R chain). So it is a switched resistor DAC. The manufacturer calls it a switched resistor DAC. The resolution is not an issue there, one can make a low depth DAC, it is not forbidden. In particular, it is a segmented DAC. A segmented switched resistor DAC. Now I am getting quite repetitive by repeating repeated things so I should finish by stopping, and conclude here.

 

[solderdude]

That's an excellent idea!

edit: if you had two, could you put a resistor in front of one to make its largest resistor the next sstep down from the least significant bit of the other and make a 14 bit DAC?

I have no idea, would require splitting the signal and there would be no benefit.
It is already possible to get >14 bit resolution with cheap DAC chips.

 

[mocenigo]

That's an excellent idea!

edit: if you had two, could you put a resistor in front of one to make its largest resistor the next sstep down from the least significant bit of the other and make a 14 bit DAC?

This is, in a sense, already done. Some DACs are segmented using a thermometer encoding for the most significant bits in (I.e. with 2^t - 1 resistors for the top t bits) and then the lower, say, 16-t, bits are handled with an R2R network.

 

[mike7877]

This is, in a sense, already done. The most significant bits in some DACs are segmented (I.e. with 2^t - 1 resistors for the top t bits) and then the lower, say, 16-t, bits are handled with an R2R network.

Ah. Are those the 4,5,6, sometimes 7 bit "resistor" DACs, or the higher 8, 10, 12 sometimes 14-16, for things like o-scopes?

I thought the lower bit, especially 4,5, maybe 6, resistor DACs weren't too difficult to pull off with just individual, exact value resistors

 

[mocenigo]

Ah. Are those the 4,5,6, sometimes 7 bit "resistor" DACs, or the higher 8, 10, 12 sometimes 14-16, for things like o-scopes?

I thought the lower bit, especially 4,5, maybe 6, resistor DACs weren't too difficult to pull off with just individual, exact value resistors

Analog Devices had the AD1866 which used a thermometer coding, 7-element resistor array for the 3 MSBs and a R2R network for the 13 LSBs.

The Burr Brown PCM1791A used a more sophisticated approach whereby the most significant 6 bits to a thermometer encoding and converts the least significant bits using 4-level delta sigma modulation, but the two values are added together to be converted by a segmented 67-element resistor array.

Having different exact value resistors is probably not precise enough, but maybe they could be fine for a few LSSb, if not that since they are at a much lower level than the MSBs, they would not need that level of precision, and thus a less expensive R2R network or a limited oversampling sigma delta modulator would both be perfect. Higher precision is needed for the MSBs.

 

[mike7877]

Analog Devices had the AD1866 which used a thermometer coding, 7-element resistor array for the 3 MSBs and a R2R network for the 13 LSBs.

The Burr Brown PCM1791A used a more sophisticated approach whereby the most significant 6 bits to a thermometer encoding and converts the least significant bits using 4-level delta sigma modulation, but the two values are added together to be converted by a segmented 67-element resistor array.

Having different exact value resistors is probably not precise enough, but maybe they could be fine for a few LSSb, if not that since they are at a much lower level than the MSBs, they would not need that level of precision, and thus a less expensive R2R network or a limited oversampling sigma delta modulator would both be perfect. Higher precision is needed for the MSBs.

What does 4 level delta sigma modulation for the least significant bits mean?
I just looked through the 1791A's datasheet and didn't find any info alluding to its nature - how did you find out about this? And the resistor array?

You seem to know about the inner workings of DAC chips - have you ever come across a designs that used a charge pump to increase the source voltage for the most significant bits to increase accuracy?
Or one that uses extremely HF PWM to regulate the current of the least significant bits, which use the next LSB that's not PWM'd as reference to operate (set as a fraction of the reference)?
If you've never seen either thing implemented, I think they'd both be pretty cool to see in a design - as long as there's nothing inherently wrong with either lol