Hi,

We know that NRZ DACs are less jitter sensitive than RZ DACs for CT delta-sigma ADCs.
However, the dynamic range, or the lowest input signal for which you get good SNR (say 0 dB SNR)
ought to be the same for both cases RZ/NRZ.

At large signal levels, the modulator has very low activity factor and so there are very few transitions in a NRZ-type modulator which improves jitter-limited SNR.

But as one goes on reducing the signal level, the activity increases. For a 1-bit quantizer, we will have more or less a 1 0 1 0 1 0 kind of pattern out of the quantizer (modified a bit by the modulator). But basically, there will be quite a bit more activity which would mean that the jitter sensitivity becomes almost the same as for a RZ type DAC.

Since one of the main reasons for a  high dynamic range (DR) requirement is the ability to resolve small signals, this seems to be equally good (or poor) no matter whether one uses RZ or NRZ DACs in CT delta-sigma modulators.

Or am I missing something ?

Regards
Vivek

Hi Stefan,

All that is fine, when speaking about a real world application. However, what I am trying to say is that a manufacturer/designer does not really get any advantage in terms of the dynamic range (DR) measured in the lab and hence its value in the datasheet when using NRZ, atleast in terms of jitter. There are other advantages such as slew etc. in design.

But if I were to make 2 CT DSMs (one with RZ and one with NRZ) and apply nice, clean, single tone signals to them in the lab, then I should see the jitter-limited DR to be the same, because as I begin to reduce the signal strength, there are more and more 1-0-1-0 kind of transitions, and even the NRZ DAC output looks more and more like an RZ DAC output. So, same minimum noise floor results in either case.

This is all just with a single-bit quantizer mind you. A multibit quantizer may show the difference between RZ and NRZ, but the difference will depend strongly on the design of the modulator. If the modulator is very stable, the quantizer outputs may still swing back and forth between just 2-3 levels and one may still see no difference in terms of jitter-limited sensitivity. What do you say to that?

Regards
Vivek

vrl wrote on Jul 17^th, 2008, 7:01pm:


Why would that be so? If we take a 1-bit DAC. Say the reference voltage is +1 V for quantizer output = +1 and -1 V for quantizer output = 0.

Now NRZ 1,0,1,0 gives +1,-1,+1,-1
RZ has  +1, 0, -1, 0, +1, 0, -1, 0.

The RZ case has the same total transition step but the difference is that it first switches to 0 before going down to -1.

Regards
Vivek

Hi Vijay,

Thanks for pointing out the fine detail. But I think this is still a fine detail. If the pulse in a RZ waveform is 99% as wide as in a NRZ waveform, then one would need to make it 1/0.99 times larger in amplitude. So, a small correction factor. I expect that this factor is not > 1/0.9.

I think this correction factor may make a difference of < 0.5 dB. But the situation remains quite comparable for low signal levels between RZ and NRZ, as far as jitter is concerned.

Any other things that might make it worse. I am just trying to see whether there really is such a great problem using RZ with a 1-bit quantizers as long as I am interested in SNR at low signal amplitudes.

But perhaps I will anyway take the suggestions posted elsewhere in this discussion and use a multibit DAC, with NRZ of course, to save on slew and jitter sensitivity.

Regards
Vivek

Hi All,

Thanks for the responses. I now have sufficient insight. So looks like I have to go with NRZ, and multibit, and make sure that ISI is not significant (some suppression due to differential scheme).

Regards
Vivek