Do you only mean that you still don't see any intermodulation products at audio frequencies or that you also don't see any increase of the tones around half the sample rate at the summing nodes?
I don't understand the question so I have no idea if this answers it, but it's a very soft tone that gets frequency-modulated by the desired audio signal. It can look like peaks on a DFT plot, like in the measurements bohrok2610 did, but it is not simply a distortion due to a slightly curved input-to-output characteristic. It is dependent on the signal, but its carrier frequency doesn't need to be commensurable with the signal frequency.
That was it. PSS as used in RF measurements is a continuous time signal. Seems to me that there is no simple word to describe what makes for a spur (short for spurious/unwanted spectral line) on an FFT or DFT. Any signal, continuous or not, that puts energy in a bin (which is to say, a signal that has some correlation with a bin frequency) will to some extent appear as a spectral line. The reason I would tend to refer to it as PSS in the context of a DFT/FFT is simply to emphasize the nature of a what we are looking at (and because ESS chose to use it that way, IMHO). Again: Some energy in a bin, as opposed to being spread over multiple bins. And therefore some correlation with a bin frequency.
The problem I am trying to address is that people used to using to FFT measurements for audio frequency amplifiers don't always have a good understanding of what a spectral line really is, what it shows. Its not necessarily directly and only HD/IMD in continuous time sense. There is some leap of faith and lack of understanding involved in that model.
The problem I am trying to address is that people used to using to FFT measurements for audio frequency amplifiers don't always have a good understanding of what a spectral line really is, what it shows. Its not necessarily directly and only HD/IMD in continuous time sense. There is some leap of faith and lack of understanding involved in that model.
Whether or not the spectral line is HD/IMD or something else it is still a real unwanted tone. How the tone is formed does not make it less audible. Claiming that clock line source series resistance change from 15 to 20 ohms is audible while these spurs are not does not sound credible either.
I've recently wired up your latest active filter using OPA1632's and been running it in , BTW thank you for sharing the circuit posted to try with Marcels dac , would any changes need to be made to the active filter with the latest FW for the PCM2DSD ? Cestrian kindly sent me one of his PCM2DSD with latest FW to compare against mine with previous FW . It provides easy listening comparisons
The levels are quite low, but if they weren't, I'd say plain old harmonic distortion or intermodulation distortion between audio frequency signals would be preferable. Those occur all over the place anyway, including in your ears and loudspeakers, but tones that get frequency modulated by the desired audio signal are rather unusual.
So are you trying to say that changes made to the PCM2DSD firmware to reduce the spurs were not needed?
We are writing about quite small spurs, some 20 dB smaller than on the DSC 2.5.2 that the PCM2DSD was designed for. For anyone considering those spurs as acceptable, 20 dB smaller spurs should also be acceptable. In that sense, the configuration file update was unnecessary.
Nonetheless, the configuration file changes are an improvement in my opinion, especially the updated dither.
Nonetheless, the configuration file changes are an improvement in my opinion, especially the updated dither.
Thank you , I'm liking what I hear with the OPA1632 stage so far theres a definite difference in sound compared to Marcels original active filter . Both are good but I personally prefer the 1632 so far
AFAIK increasing data length is what mitigated the splitting of the peaks. Hard to see why that would not be an improvement as well. Adding DC offset sounds like a kludge but I'm no expert.
You know there is a difference in meaning between "especially" and "only", don't you?
Instead of just shifting the spurs in frequency, the improved dithering made them smaller. I think that's the most important improvement, because any signal or offset in the programme to be played through the DAC will modulate the momentary frequency anyway.
Instead of just shifting the spurs in frequency, the improved dithering made them smaller. I think that's the most important improvement, because any signal or offset in the programme to be played through the DAC will modulate the momentary frequency anyway.
I don't want to interrupt the technical discussions but thought I'd add another build to the list. This one is currently on test "lash-up mode" to prove out the principles and test various power arrangements. It is a build for a friend on another forum who bought the pcbs originally but hadn't had time to get around to assembling. The power supplies are nearly all ldovr LT3045/3094 modules although these will ultimately be going into my 2 RTZ Dacs to replace the LM317 ones I have currently. I have made an interface pcb to mount the JLSounds I2SoverUSB and provide the I2S/DSD signals to the PCM2DSD.
The actual power supplies that he will use have not been finalized yet.
I am currently listening to it via the positive phase outputs only but in due course will "borrow" the Lundahl 1588s from one of my DACs to try them on this unit and direct A/B comparison with then be closer.
ps. It is sounding very nice indeed.
The actual power supplies that he will use have not been finalized yet.
I am currently listening to it via the positive phase outputs only but in due course will "borrow" the Lundahl 1588s from one of my DACs to try them on this unit and direct A/B comparison with then be closer.
ps. It is sounding very nice indeed.
I only looked up to 30 Khz after the filter but if any modulation product would have been there, they were invisable below the noise at -130dB
But I’ll redo the sim and compare the tones at the original fs/2 at 2.8Mhz to the new fs/2 at 5.6Mhz for the original DSD128 file, both at the summing points as you mentioned.
And what loading should be there at the summing point ?
1) No loading at all, 2) a passive LC filter or 3) the original OPA2210 filter.
A sim takes many hours to finish, so better to know in advance what you would like to see.
Hans
@Cestrian
Quite Impressive with all those neat separate supplies.
My question is, have you ever considered to do the PCM to .dsf conversion in advance like with JRiver.
Once converted it’s there forever.
That would make the PCM2DSD converter obsolete and be replaced by the supposedly superior conversion that JRiver performs.
Hans
Quite Impressive with all those neat separate supplies.
My question is, have you ever considered to do the PCM to .dsf conversion in advance like with JRiver.
Once converted it’s there forever.
That would make the PCM2DSD converter obsolete and be replaced by the supposedly superior conversion that JRiver performs.
Hans
@Hans Polak
Pcm2dsd has a bypass mode when you send DSD stream. So you can use other ways to send the DSD stream to dac.
Pcm2dsd has a bypass mode when you send DSD stream. So you can use other ways to send the DSD stream to dac.
I like more the real time conversion PCM-DSD that does Daphile than use the PCM2DSD.
The idea is to repeat each sample, so the tones near 2.8 MHz should remain near 2.8 MHz when the clock frequency is doubled. They should be very big, so you can probably see them in a relatively short simulation run.
I would be most interested in case 3.