Listening with your eyes? How to read graphs.

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#1
Listen with your ears, not your eyes!
This may seem funny, but for the past 2-3 years I spend hours replying to emails from angry customers that have downloaded files from HDtracks and use inferior tools to “look” at the FFT and spectrogram and determine it’s an upsampled file.
What I’m going to try to do today is to teach everyone how to read these files and ways to determine if it’s really hi-rez or not!
So many times I’ve seen reviews of SACDs touting its sonority until someone determines that it is just an upsampled file and all of a sudden it doesn’t sound good anymore. What a crock!
O’kay… let’s start off with an easy one. We all know that Redbook CD’s are 16/44.1. A few major labels thought they could pull a fast one and reissue their complete catalog into this new format. Easy money, right? They just take these 24/44.1 files and convert them to DSD and produce SACDs of the same material. Unfortunately they send the SACD’s to me to get the DSD information off of them for download and look what I find. A definitive, easy to detect, upsampled SACD.
 

Attachments

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#2
Now on to a harder one! As you are well aware, DSD files contain high frequency noise if filters are not applied in the workstation. Some music contains very little content above 20k. DSD noise starts to creep in around 22k. What if the performance was a quiet one? The DSD noise will overpower the content above 20k. So how do you determine if it’s hi-rez? The below photo shows that phenomenon.
 

Attachments

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#4
This is a picture of the whole album. But what if a label doesn’t have a high rez file of a particular song on the album? What you need to do then is a spectral analysis of the whole album. You can see below that tracks 4, 9, 12, 14, 17 and 19 are NOT hi-rez.
 

Attachments

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#5
Are you still with me?
We can throw more monkey wrenches into the mix. Let me put these scenarios out to you for discussion. Theoretically 24bit files have a dynamic range of 144dB. That means you need a software program that can go that low into the darkness. We use a program that can go down to -150dB.
What if they upsampled a 16bit file? What happens to a redbook file that is upsampled to DSD and run through an analog console? How much noise do dither algorithms produce?
I’ll post more about this rant later after some discussion!
 

garylkoh

WBF Technical Expert (Speakers & Audio Equipment)
#6
Nice! Thanks for taking the time to do this, Bruce. It's not only the SACDs that can just be up-sampled Redbook, there are many DVD-A that are the same.

For those of us who are just curious, the free software Audacity will also be able to do this spectrum analysis, but obviously more limited than the resources that Bruce can bring to bear.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#7
Nice! Thanks for taking the time to do this, Bruce. It's not only the SACDs that can just be up-sampled Redbook, there are many DVD-A that are the same.

For those of us who are just curious, the free software Audacity will also be able to do this spectrum analysis, but obviously more limited than the resources that Bruce can bring to bear.
Unfortunately Gary, Audacity have been proven over at the CA forum to be unreliable and give false readings. That's what I mean about inferior tools in the hands of inexperienced people.

Which brings me to yes, we have found quite a few DVD-A AND a handful of Blu-ray discs with upsampled files.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#8
Bruce, brilliant idea and post. Sorry, but, no I am not with you. For posts 4, I think you might tell us specifically what you see on the graph or spectrum,

Post 4 did not readily pop out at me at first glance!

Tom
Post #2 dips down to 23k at approx -123dB. Now if this is a 24bit file, I would say this started out as a 48k file, since we know if it was a Redbook file, it wouldn't contain anything over 22.5k.

Post #3 shows a dip at about 22.5k, but only at about -119dB before the DSD high-frequency content takes over. A way to look into this further is to make a considerable gain change in the file for the high frequency content to overcome the DSD artifact.

Post #4 is a one hour timeline going across with the freq. content going vertical. Over on the left side you can follow 22.5k across the timeline and see where some tracks cut off at the frequency. Those are upsampled tracks within the file.
 

garylkoh

WBF Technical Expert (Speakers & Audio Equipment)
#9
Unfortunately Gary, Audacity have been proven over at the CA forum to be unreliable and give false readings. That's what I mean about inferior tools in the hands of inexperienced people.

Which brings me to yes, we have found quite a few DVD-A AND a handful of Blu-ray discs with upsampled files.
Oh no. Does it also make bad recordings??
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#12
Here are 2 more graphs. I've put a rectangle around spikes at approx 28k. This is more likely due to tape bias.

I've also drawn a black line and arrows across the spectrogram to show the tracks that are just upsampled.
 

Attachments

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#18
Guess this needs clarifying as such.

The colored spectrogram is based on on this bar graph.

The more amplitude you have, the more red the picture is.... the less amplitude.. the more blue.

This is better than a FFT graph because it gives you amplitude over a timeline, whereas an FFT doesn't. A FFT graph averages over a period of time.. Just like I showed you above. If it were not for the spectrogram, you would not be able to pick out the tracks of an album that were not hi-rez
 

Attachments

rbbert

Active Member
Dec 12, 2010
3,423
0
36
Reno, NV
#19
I'd like you (Bruce B.) to post your graph of (any track from) Steely Dan's "Gaucho" album (HDTracks version), and compare it to any track from the "Aja" SACD. Gaucho has a dive in frequency response at about 20-21 kHz, and then a slowly rising (but very low) amplitude level at frequencies above that. Aja, OTOH, shows a smooth and gradual decrease in level all the way up to about 26-28 kHz without that rise afterwards. Since the original music is very similar in content, my interpretation is that there is an artificial low-pass filter used in the mastering of the HD (and/or DVD-A) version of Gaucho, and that it's really more like a 24/44.1 recording than a true 24/88 or 24/96 (as an aside, it looks almost identical to the spectral analysis from the Citizen Steely Dan box - all albums - except for some brickwalling of the DVD-A/HDTracks version). If there is a reasonable alternative explanation for that high frequency cut-off, I'd like to know what it is.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,558
2
38
Seattle, WA
www.pugetsoundstudios.com
#20
Here are the 2 graphs.

Steely Dan - Gaucho, taken from HDtracks downloads at 24/96
Steely Dan - Aja, tape transfer downsampled to 24/96
 

Attachments

Last edited: