Why not trust your own ears and try to find the right answer that way.
Hans
Hans
What I do here is use one modest linear +5v power supply which is shared by the clean side of I2SoverUSB and by PCM2DSD. They both have their own secondary voltage regulators and they are both going to be on the same ground anyway. An LM7805 or something like that would be good enough. That said, I would isolate and reclock after PCM2DSD using external clocking, so the shared power supply should have little or no effect on final SQ anyway.A question for those who have experimented with PSs. Which of the following is more likely to have a noticeable impact on the sound quality of this DAC; using a low noise power supply on the Reclock side of the i2soverUSB or on the PCM2DSD ? Replying “both” will not answer the question I’ve asked 😉. Any thoughts?
I suggest you also include a HQPlayer quasi-multibit modulator (e.g. AMSDM7 512+fs) as a reference to your listening evaluation as such modulator has been shown to work well with this dac.I have the free version of HQplayer but unfortunately it doesn’t work with Amazon , it’s no problem to download a file though and use HQplayer .
The plot in post #2696 shows that a +/- 1 ppm change of the bit weight depending on the previous bit is enough to produce about the same intermodulation products in the audio band as the present DAC and filter do.
And according to a post in TI's EE forum they expect CMOS ICs inter-gate skew to be less than 250 ps although this is not guaranteed.
In any case, the idea of lifting the spectral lines by sampling the input data twice with a clock at twice the original BCK, made the spectral lines 0.5fs and 1.5fs raise by 60 to 70 dB, as visible in the attachments 3 and 5.
But now looking at the filtered spectrum, the filter with op-amps gave crazy results maybe because of overloading its input, that's why I used a 4th order LC filter instead.
With this filter no signs of modulation products are visible, see sixth attachment.
It will be considerably better within one batch, but only the people doing production testing know how much better. Chips at the edge of a wafer usually differ more from their colleagues than those that are not near the edge.
What matters are pulse width changes depending on other data bits.
The 4th order LC Filter was used as the refence to prove that the whole setup worked properly.An LC filter made of ideal components is linear and time invariant, so it cannot create intermodulation components. That's as expected.
I'm curious what your crazy spectrum at the output of the op-amp filter looks like. With 60 dB larger spectral lines around fs/2 and its odd multiples at the summing points, second order intermodulation products at the op-amp outputs should increase by 120 dB if nothing clips (to be more precise, if the second-order term still dominates). It is very well possible that the distortion gets larger than the intended signal.
By the way, using the normal clocking but changing the value of one FIRDAC resistor a bit (such that it doesn't match the others anymore) could be a way to increase the stuff around fs/2 by less than 50 dB ... 70 dB.
I have been taking a brief look on some of the 74 types you mentioned in #2984 and #2982 and one of the differences I notice is that the inductances of the pins of the various packages vary quite a lot.
And here your 74ALVC164245 IC appears to have a comparatively very balanced pin inductance which I reckon would be a feasible quality in this FIRDAC context ...
And then, regarding clock adjustment I noticed that you would prefer to use a 1/100 division probe or an active probe. Would you happen to have a suggestion for an active probe that is accessible price-wise and works with BNC connectors?
Would something like this work:
And, to this end, would you say that a 200 MHz oscilloscope would be sufficient for such a measurement?
Also, since you appear to be well-versed in various IC packages maybe I can ask you for a bit of advice on this: I am for comparison considering a basic R2R DAC. Would you happen to have a suggestion for an IC version that would be suitable for such a DAC?
And then, just for clarification and FYI ... I feel like mentioning that we may have different tastes in music as I am mostly into classical music ...
The other op-amps either don't converge or don't distort at all; is distortion even modelled for the latter group?
Of course. I think TSSOP or TVSOP are acceptable in practice. BGA seems unnecessary difficult to process, even in commercial volume production.
When I needed one, I made my own, differential. Validated on 5GHz RF generator/analyser.
Regarding an R2R DAC as I mentioned I am thinking about a basic R2R DAC - the idea is to just get an idea about the R2R DAC sound relative to the DSD sound.
With this in mind I would not venture into anything costly or complex or involving an IC R2R DAC solution.
Regarding R2R dacs, typically they are segmented with the first few most significant bits being thermometer coded, and with the lower order bits being R2R. As it happens Andrea Mori makes such a dac. Have one here, in fact. Of course comparisons between Andrea's R2R and DSD dacs have been done here and in Italy. So, I can tell you what subjective differences have been reported from the two groups, here and in Italy. However if nothing will do but to build your own, then guess I can't help.