|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
In Reply to: Thanks Christine posted by Eric LeRouge on June 15, 2004 at 11:25:37:
Well, not really, but hot pics of the SACD waveforms at least :)Christine,
I can't post things in a very professional way on a web page, but here's part of the data I got with the Diana Krall album. I tried to record the entire album at 32/192 but my PC/software refuses to let me write files bigger than 4Gb, so I could only record the first nine tracks in one file (about 45mn), track 10 is in a separate file.
I recorded using my Lynx soundcard, using CEP and checked with Sound Forge. The results were identical but I preferred doing everything in CE to use the same statistics, and not have to do file saving (saving a 4Gb file took 10mn!!) or any copy/paste.
Anyway, as you know the volume level is very low on the SACD as on all SACDs, so I normalized the entire file to 99.8dB. I won't go into too many aspects, but here's my impression:
- the SACD isn't clipped when you focus on the highest peaks (not like the bad cases on the DVD-A that you showed), but it has a very distinct signature similar to using some form of limiting. This is very clear in track 4 and maybe track 7, but in other instances as well. The only type of music where I get this type of pattern is either music that has badly clipped at the recording or at the mastering (eg Pink Floyd) or electronic music when the synths are set to a constant volume, but never on acoustic music when the levels and dynamics are never exactly the same. In short, it looks like some sort of limiting has been used, but I can't say how and when.
This is easily checked on the most obvious track #4, I drew a red line on the limit that has been imposed on the track (again, this could be at the recording or at the mixing stage). The same thing can be found on other tracks, including track 7 and 9
- regarding frequency distribution and analysis, I do not find the missing frequencies as described, but I do see a small peak at -80 or -82dB, but I think that's a problem with my soundcard since I installed additional hard disks, i think the noise floor was lower before (not sure, but I'll check that). When I applied the Bessel filter as per your description, that little "bump" didn't disappear, but it was lower, towards -90dB, so maybe the filter helps in some way.- To conclude, when I did try to use a 6th bessel filter, first at 25760Hz, then at 15760, and it didn't change the frequency distribution, but it changed (slightly) the statistics, which are as follows
32192 no filters (norm 99.8dB) Same, filtered Bessel @ 25670HzLeft Right Left Right
Min Sample Value: -31076.61 -32168.22 -30810.97 -31084.36
Max Sample Value: 31428.37 32702.85 30772.38 31140.45
Peak Amplitude: -.36 dB -.02 dB -.53 dB -.44 dB
Possibly Clipped: 0 0 0 0
DC Offset: 0 0 0 0
Minimum RMS Power: -132.77 dB -127.44 dB -138.04 dB -128.78 dB
Maximum RMS Power: -5.16 dB -5.97 dB -5.16 dB -5.97 dB
Average RMS Power: -19.14 dB -18.76 dB -19.15 dB -18.76 dB
Total RMS Power: -17.93 dB -17.65 dB -17.93 dB -17.66 dB
Actual Bit Depth: 32 Bits 32 Bits 32 Bits 32 Bits(Sorry about the layout).
There are some notable differences with your own figures, but this may be due to the fact that the statistics are computed on the whole sample of the first 9 tracks, not just one song, so maybe there are much louder tracks in the sample.
I think the only significant factor is that the filter lowered the minimum RMS value (I'm not sure how that makes sense, since it is supposed to delete samples above certain threshold, in this case 50kHz). The difference between minima/maxima as per our previous discussion (which to me means "dynamics") would then increase. But it is unclear to me whether this is an artifact due to the filter or whether the "real" dynamics appear after the filter has been engaged, maybe you and Frank have a better idea. Average and total RMS power is unchanged, which is strange.Special bonus:
The following charts are really more ornamental than anything, but they show how much ultrasonic noise is really floating in the SACD tracks. On the "linear" view, I highlighted the frequencies where I have peaks, because it seems that on all my recordings I get the same peaks on those frequencies. The largest peak at around 60kHz seems higher than usual, but I'm not sure.
Obviously, this was done before the Bessel filter was applied.
They also show some interesting differences between left and right channel, but I have no idea where that comes from, it looks like the two channels didn't receive exactly the same treatment. I'll double check with other recordings whether that is the case or if that could come from my soundcard.I hope this helps in your analysis
Best
Follow Ups:
Nice work Eric,Regarding the hf peaks in the spectrum plot:
You should check if the peak at 60kHz correlates with your monitor frequency. Simply adjust the videocards scanrate. If the peaks shifts to another frequency you can be certain that it's interference from your videocard.***I think the only significant factor is that the filter lowered the minimum RMS value (I'm not sure how that makes sense, since it is supposed to delete samples above certain threshold, in this case 50kHz). The difference between minima/maxima as per our previous discussion (which to me means "dynamics") would then increase. But it is unclear to me whether this is an artifact due to the filter or whether the "real" dynamics appear after the filter has been engaged, maybe you and Frank have a better idea. Average and total RMS power is unchanged, which is strange.***
The filters lowers the wideband noise contribution. It's normal that the minimum RMS value will be lower after applying a filter.
This is true because the noise energy and the signal energy ratio is small.It's also logical that the avarage and total RMS power changes very little. The signals energy is mostly concentrated below the filters cutoff frequency. Also the signals energy is much stronger than the noise contribution.
Thanks Frank,Very smart idea! I'll check my monitor and video board, because I think something's not right (I'm sure they don't share any IRQ though, but I'll try to see how I can check that.
(more work!)
Cheers
yes - you've definitely found signs of peak limiting - no doubt about it.i'll try and record the whole album this weekend on both DVD-A and SACD and see if i can replicate your results.
This post is made possible by the generous support of people like you and our sponsors: