I managed to get my hands on an oscilloscope and took a few pictures.
The measurements are done at the power amp output, the dac is a CS8420/DF1704/Balanced PCM1704K/ with passive IV conversion.
The spdif data originates from an old Pioneer PDS-802 fitted with another CS8420 set for 16bit in and 24/48 out.
The disc generating the -100dB and -90dB dithered data is an old test disc from HiFi News & Record Review produced by Martin Colloms.
I am not sure what I am seeing here, looking at the -100dB signal, are the PCM1704 switching the least (7) significant bits on and off or what....?
Any thoughts or insights are welcome!
/Michael
The measurements are done at the power amp output, the dac is a CS8420/DF1704/Balanced PCM1704K/ with passive IV conversion.
The spdif data originates from an old Pioneer PDS-802 fitted with another CS8420 set for 16bit in and 24/48 out.
The disc generating the -100dB and -90dB dithered data is an old test disc from HiFi News & Record Review produced by Martin Colloms.
I am not sure what I am seeing here, looking at the -100dB signal, are the PCM1704 switching the least (7) significant bits on and off or what....?
Any thoughts or insights are welcome!
/Michael
Attachments
I don't know how you are inferring that from the pic (when you say -100dB do you simply mean digital silence?), but yes, the LSBs will be affected due to dither. The 16-bit signal occupies the top 16 MSBs so when dither is applied it affects the remaining LSBs for 24-bit output.
-100dB refers to the -100dB test track on the disc, digital silence does not show the pattern I posted, it shows a flatline without the "spikes"
Ok, that does make sense. It is probably just a few bits turned on periodically to make the DAC generate some output.
Hi,
-100dB is very very low.
Philips used -60dB and I suggest a spectrum analyzer because the scope tells nothing.
Bernhard
-100dB is very very low.
Philips used -60dB and I suggest a spectrum analyzer because the scope tells nothing.
Bernhard
I was under the impression that the asrc would interpolate, creating intermediate values and not just re-sample the same value until the next one comes along?
It does, and I've watched AD1896 do it. Otherwise only 1 bit would change.
To see this easily, connect channel A to LRCK, channel B to SDATA (DOUT) and trigger on channel A. You can see each bit value this way, and the LSBs are clearly active.
If you have a third channel to display BCK it makes it even easier to see the transition points.
To see this easily, connect channel A to LRCK, channel B to SDATA (DOUT) and trigger on channel A. You can see each bit value this way, and the LSBs are clearly active.
If you have a third channel to display BCK it makes it even easier to see the transition points.
A 8 said:
And you won't recognize small distortions from looking at the waveform.
If it starts looking ugly, it already sounds very ugly.
I think I see what you mean. You want the sine wave to look nice and smooth? The data just isn't there. At -100dB there isn't even a signal to speak of, just some bit transitions. You're probably also at the limits of your measurement equipment with the 5mV range. Is that the lowest setting on the scope?
tiroth,
Yes, this looks as bad as my old (non asrc/16bit processing) stuff and considering the multiple spikes I was thinking the setup just resamples at 24/96 the same value until the next original 16/44 sample comes along.
The scope would not go lower then 5mV/div but as I said the measurement was done at the poweramp output at full volume which would give approx +-48 volts peak at digital 0dB
Bernhard,
The -100dB bad looking pattern still sounds like a sine but with a slight "hiss" imposed upon it.
Yes, this looks as bad as my old (non asrc/16bit processing) stuff and considering the multiple spikes I was thinking the setup just resamples at 24/96 the same value until the next original 16/44 sample comes along.
The scope would not go lower then 5mV/div but as I said the measurement was done at the poweramp output at full volume which would give approx +-48 volts peak at digital 0dB
Bernhard,
The -100dB bad looking pattern still sounds like a sine but with a slight "hiss" imposed upon it.
I would not expect it to look different with ASRC, in fact I would be surprised! Oversampling and ASRC perform the same function vis-a-vis smoothing. The difference would be in reducing jitter sidebands. This is probably occurring below -100dB, which means a very sensitive measurement setup is required. You will not be able to visualize distortion until it is extreme, orders of magnitude more than is "acceptable".
Bernhard said:
Do you mean the steps on the sinewave?
Can't this be artifacts from the digital scope's display?
What does it help to have 24 bit if the recording is only 16 bit ?
If upsampling does interpolation, detail is lost IMHO, at least it is like this with pictures.
@Bricolo
How to make a sine with a few bits ?
If upsampling does interpolation, detail is lost IMHO, at least it is like this with pictures.
@Bricolo
How to make a sine with a few bits ?
- Status
- Not open for further replies.