Hugo TT apparently has a 4 element pulse array modulator. DAVE has 20 elements and QBD76 has 16.
It seems that the element count is a substantial contributor to the quality of the DAC - the tap count isn't the whole story.
So what is the mechanism here that makes more elements better?
What is a pulse array modulator?
The pulse array modulator is the digital circuit that converts the data output (6 bits) from the noise shaper into an array of PWM signals. The output from the pulse array modulator is then fed back into the noise shaper to provide the feedback for the noise shaper. The element outputs then go to the OP flip flops, which via resistors give the analogue output that is then fed to the output stage. A flip flop and resistor is hence one element (1e). We need to use a flip flop so that the jitter from the FPGA is eliminated.
I have not talked about pulse array because its very complex - taps are a little easier to understand.
There are many benefits of pulse array over all other conversion strategies. The first is that it is a
constant switching scheme, completely independent of the digital data - the benefit of this is the constant switching energy is always the same, so the music signal has no switching related errors - so the switching errors end up being a DC value, which the digital DC servo nulls out. This is one reason why pulse array is capable of zero noise floor modulation. The second reason is that the array is fundamentally jitter immune as rising edges are cancelled by falling edges within the array. The third advantage is that the elements are all the same value, and each element has, for audio signals exactly the same signal, so this means that the tolerance of the resistors only creates a completely fixed and unvarying noise signal, with zero distortion - another reason why pulse array can have zero distortion and zero noise floor modulation. The fourth benefit is that it is a voltage switching scheme, not a current or resistor switching - which means it is several orders of magnitude faster than all other schemes, so that is why I can run the noise shaper at 104 MHz - and this high rate means that with Dave I can have the digital performance of the noise shaper exceed 350 dB as I have talked about before. The 5th benefit is that current that is going into the virtual ground node of the OP stage (assuming perfect operation and that's another story) is
completely audio signal unrelated - it can only add +1 or -1 level of current into the node - another reason for zero distortion of pulse array. All other schemes actually switch signal related currents, and this creates distortion, and as the switching is SR related it also creates anharmonic distortion too. The 6th benefit is that each element is completely independent to one another, as they are discrete, and individually decoupled. On a chip they interfere with one another, causing small signal distortion. Thus having them separate means that small signals have zero distortion, so you get much better detail resolution and depth perception - in measurements, Dave perfectly reproduces small signals with zero amplitude errors or distortion. The 7th benefit is the RF levels is very low, as the net switching noise itself is low, and it can be decoupled effectively. This means the input to the op stage is always linear, so you get zero noise floor modulation.
I told you it was complicated, that's why I don't talk about it. I have also probably forgotten something important too, but it gives you an idea as to what is going on. It may also give you some appreciation of the problems involved in creating a DAC that has zero noise floor modulation, and almost zero distortion too (and I have not talked about the OP stage distortion problems, nor the reference circuitry - the power to the flip flops - either).
The reason Dave is 20e not 16e, is that when you approach zero or 100% output, the switching activity stops, then you get distortion. With the QBD, an overhead was built in as its a fixed level DAC. With Dave I don't have that, as we want to drive the output to maximum, so the extra 4e gives the overhead to prevent distortion as the signal goes to clipping.
4e is more complicated, as compromises are involved, and in that case it sounded better to maintain noise shaper resolution (maintaining soundstage) but compromising distortion slightly. That's why Mojo is at 0.00017% but Dave is 0.000015% THD. But Mojo is still comparable with the very best DAC's available in distortion performance. But its still a constant switching scheme, so like Dave the DAC has no measurable noise floor modulation, unlike any other non Chord DAC's. And this attribute is way more important than distortion when it comes to SQ.
Rob