Can you provide the schematic showing the mods? It is always interesting to speculate the cause for a positive outcome.
Gerrit
Here you are
What do you think makes NOS sound different?
But with R1/R2 0R instead of the 5700R used as noise source.
Original version was with 1K//1n5 from output to minus input where the PCM1792 dropped it’s current.
Hans
I have changed both channels now having this hybrid or MFB filter.
Although white noise and frequency sweeps show responses flat as a ruler, just as before, but sound impression has changed.
Much to my surprise and the last thing I expected is that I have a more solid bass.
But at the HF side, it is almost as if I have removed a carpet from my room.
Sound is less warm and violins have less wood.
I have to listen to it for a longer time, but I’m not yet fully convinced what I prefer.
So much for the important role of an analog reconstruction filter, both having a flat GD beyond 20Khz.
Hans
Although white noise and frequency sweeps show responses flat as a ruler, just as before, but sound impression has changed.
Much to my surprise and the last thing I expected is that I have a more solid bass.
But at the HF side, it is almost as if I have removed a carpet from my room.
Sound is less warm and violins have less wood.
I have to listen to it for a longer time, but I’m not yet fully convinced what I prefer.
So much for the important role of an analog reconstruction filter, both having a flat GD beyond 20Khz.
Hans
Thanks for the schematic Hans,
(It can also be that the solid bass is a function of the carpet being removed). Your impressions have frequently been of my experience. It suggests of fine tuning to find some middle ground of balance as most satisfactory, that when inclined to do other things for awhile, is left to just enjoy listening to music through the system.
It is considered that there are two competing factors that both contribute to what is considered perhaps the most important factor in reproduction, being dynamic contrast. One competing factor is the obscuration of hearing details into the background, that failing all other distortion artifacts foreshortens the dynamic depth of field. The second is distortion artifacts attached to undistorted signals that punch dynamics forward, that failing obscuration artifacts magnifies or overdoes the dynamic field.
This suggests that there can exist many systems that can achieve enjoyable dynamics in some combination of obscuration and distortion (the article on temporal coherence suggests that aligned distortions would be dramatically amplified). However, this suggests that high levels of obscuration requires high levels of distortion artifacts normally achieved by turning up the volume to punch through the obscuration. This also suggests that in the absence of obscuration realistic dynamics can be achieved at the lowest levels of reproduction, representing the holy grail.
Gerrit
I think that Hans only means that figuratively. I don't read him as stating that, he litterally removed a floor carpet from his room. Just that the sound became subjectively brighter or more airy, despite measuring as flat as before. I find it interesting that in addition to a perceived increase in brightness, he perceived a stronger bass. Which subjectively means, of course, that the midrange would therefore sound relatively suppressed.
Last edited:
Yes Ken,
It was just a virtual carpet.
But to be honest, I like the sound lesser and lesser because of being almost harsh, analytical and fatiqueing.
The only thing that I can think of as a cause is that the 1792/1794 does not like voltage excursions at it’s current output.
If true, it would mean that those implementations with a passive I/V conversion for the PCM1792/1794 may also experience a more analytical sound .
And since quite a number of NOS Dac’s are equiped with passive I/V, could this be part of the typical sound difference ?
It’s a pitty that so far nobody with a NOS has done the up/down sampling test.
That could bring us a step furher.
Hans
It was just a virtual carpet.
But to be honest, I like the sound lesser and lesser because of being almost harsh, analytical and fatiqueing.
The only thing that I can think of as a cause is that the 1792/1794 does not like voltage excursions at it’s current output.
If true, it would mean that those implementations with a passive I/V conversion for the PCM1792/1794 may also experience a more analytical sound .
And since quite a number of NOS Dac’s are equiped with passive I/V, could this be part of the typical sound difference ?
It’s a pitty that so far nobody with a NOS has done the up/down sampling test.
That could bring us a step furher.
Hans
You always have voltage excursions at the current output. The input of your original transimpedance amplifier must have behaved similar to an LR parallel network, what you have now behaves similar to an LRC parallel network and the passive solution is typically an RC parallel network. They're all different and they all have a nonzero impedance.
Voltage excursion on the opamp’s input with the original version is larger and less smooth as with the used MFB circuit, which is good to prevent oversteering as shown in previous sims.
However the voltage excursion at the Dac’s current output is much larger now with the MFB version because of the 22R/680pF in between.
And although these voltage excursions with MFB are very smooth when simulating the Dac as a perfect current source, resulting in LTSpice with less IMD, reality is not supporting this.
Harsh sound is mostly caused by IMD, opposite to what LTSpice predicts.
Only the Dac properties in the chain are unkown, most likely causing the differences between the sim and reality. All other active components are still the same.
Ergo, it must be the Dac that doesn’t behave like a true current source, which will then most likely also be the case when terminated with a passive RC network.
Hans
However the voltage excursion at the Dac’s current output is much larger now with the MFB version because of the 22R/680pF in between.
And although these voltage excursions with MFB are very smooth when simulating the Dac as a perfect current source, resulting in LTSpice with less IMD, reality is not supporting this.
Harsh sound is mostly caused by IMD, opposite to what LTSpice predicts.
Only the Dac properties in the chain are unkown, most likely causing the differences between the sim and reality. All other active components are still the same.
Ergo, it must be the Dac that doesn’t behave like a true current source, which will then most likely also be the case when terminated with a passive RC network.
Hans
With reference to the circuit of post #512, when all components are ideal, the input impedance is that of an LRC parallel network with L = R1 R2 C1, R = R1//R2, C = C2. You can't make R2 too low because the op-amp output will then see too much capacitive load. sqrt(L/C) should be larger than R to prevent high-Q resonances. I guess the only thing you can do is to change both C1 and C2 to shift the impedance peak to a frequency range where it does less harm, if any.
Marcel,
Thanks for the suggestion, I will give it a try.
Anyhow, with my PCM1792 DAC, generating an output every 5usec., the biggest current step it will make is some 2mA from the +/4mA it can produce.
So I offered a square wave of that size from the current sources .
In the image below on the left side in blue the somewhat messy voltage excursion of the combined Dac output and Virtual opamp input with the used LT1486, however it stays within +/-5mV.
In teal the diff. input of the OPA1632, nothing spectacular here.
The right part of the image shows the voltage excursions for the MFB version.
Now in red the Virtual input of the LT1486 opamp, that is also within +/-5mV, but much smoother as before.
Also in teal the diff input of the OPA1632, that's again fully o.k.
But in blue the Dac's voltage output divided by 10 just for this image, to enable comparing the voltage excursions optimally.
So the Dac is making a voltage excursion of almost +/-40mV, where I think that the real DAC most likely reacts in a different way as in the sim, because not being a real current source with a high and constant output impedance.
Hans
.
Thanks for the suggestion, I will give it a try.
Anyhow, with my PCM1792 DAC, generating an output every 5usec., the biggest current step it will make is some 2mA from the +/4mA it can produce.
So I offered a square wave of that size from the current sources .
In the image below on the left side in blue the somewhat messy voltage excursion of the combined Dac output and Virtual opamp input with the used LT1486, however it stays within +/-5mV.
In teal the diff. input of the OPA1632, nothing spectacular here.
The right part of the image shows the voltage excursions for the MFB version.
Now in red the Virtual input of the LT1486 opamp, that is also within +/-5mV, but much smoother as before.
Also in teal the diff input of the OPA1632, that's again fully o.k.
But in blue the Dac's voltage output divided by 10 just for this image, to enable comparing the voltage excursions optimally.
So the Dac is making a voltage excursion of almost +/-40mV, where I think that the real DAC most likely reacts in a different way as in the sim, because not being a real current source with a high and constant output impedance.
Hans
.
Attachments
Information that I have for the PCM1792/1794 is 200nsec.
But 100nsec or 200nsec makes no difference in the sim.
Hans