• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

PGGB upsampler: DeltaWave null analysis

TNT

Active Member
Joined
Jun 24, 2020
Messages
237
Likes
155
Are you skeptikal because you have an adamant belief in your testing script's ability to express audible differences?
LOL... -200 dB

Thats like an atom bomb exploding standing 3 meter away.... get a grip...

//
 

tuga

Major Contributor
Joined
Feb 5, 2020
Messages
3,984
Likes
4,274
Location
Oxford, England
LOL... -200 dB

Thats like an atom bomb exploding standing 3 meter away.... get a grip...

//
Sorry, I know nothing about bombs. Which parameter does -200dB refer to? Is it another oversimplistic mish-mash like SINAD?
 

fatoldgit

Active Member
Joined
Feb 29, 2020
Messages
263
Likes
310
Practically?
Code:
sox INPUT OUTPUT rate 705600
Probably doesn't make sense to aim for better noise rejection than 16 bit.

But I guess we don't care about practicality :) Then one of those:
Code:
sox INPUT -b32 OUTPUT rate 705600
sox INPUT -b32 OUTPUT rate -v 705600
sox INPUT -b32 OUTPUT rate -s 705600
depending which tradeoffs one prefers:

Note that the latest version of sox has "-u" option which is one level above "-v"...not done any testing to see if it makes any diff.

The quality-level options are as follows. If none is chosen, the default is "-h" (high):

-q = quick
-l = low
-m = medium
-g = medium-high
-h = high
-e = extra
-v = very high
-u = ultra

1677747233436.png


On Linux at least you typically need to download the source code and compile as most distro's dont use the latest version.

Peter
 

theREALdotnet

Major Contributor
Joined
Mar 11, 2022
Messages
1,164
Likes
1,933
The A-weighted difference between the two files in the time domain is -204dB.

That means there is still a 0.000000006% difference between them. You call that inaudible?
 

danadam

Addicted to Fun and Learning
Joined
Jan 20, 2017
Messages
932
Likes
1,473
Note that the latest version of sox has "-u" option which is one level above "-v"...not done any testing to see if it makes any diff.
It looks like it gets rid of that ripple in the middle and at the end:
full.png

zoom.png
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,505
Likes
10,012
Location
North-East
That means there is still a 0.000000006% difference between them. You call that inaudible?

Painfully audible, but only to Rob Watts from Chord, who claims to hear noise floor modulation down at -300dB :eek:
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,505
Likes
10,012
Location
North-East
Note that the latest version of sox has "-u" option which is one level above "-v"...not done any testing to see if it makes any diff.

The quality-level options are as follows. If none is chosen, the default is "-h" (high):

-q = quick
-l = low
-m = medium
-g = medium-high
-h = high
-e = extra
-v = very high
-u = ultra

View attachment 268723

On Linux at least you typically need to download the source code and compile as most distro's dont use the latest version.

Peter

When I get a few minutes, I'll give the -u option a try to see if it makes a substantial difference.
 

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,745
Likes
5,796
So, DeltaWave upsample result is closer to PGGB than SoX with -v option, but overall, the difference is still completely inaudible and well below 23 bits.
What you mean to say is
“My ASR-tested, ultra-precision $500 upsampler software based upon the field-proven, instrumentation-grade DeltaWave technology provides results nearly identical to the twice as expensive PGGB but at world-class speed and convenience. My hand-selected algorithms provide superior matching to PGGB’s maximum quality setting than even the open source SoX upsampling tool.”
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,505
Likes
10,012
Location
North-East
What you mean to say is
“My ASR-tested, ultra-precision $500 upsampler software based upon the field-proven, instrumentation-grade DeltaWave technology provides results nearly identical to the twice as expensive PGGB but at world-class speed and convenience. My hand-selected algorithms provide superior matching to PGGB’s maximum quality setting than even the open source SoX upsampling tool.”

I was going to charge $900 for it, actually ;)
 

GXAlan

Major Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
3,745
Likes
5,796
I was going to charge $900 for it, actually ;)

$900 with $400 coupon code :)

Then you can say “performs as well as software twice the price”
 
OP
pkane

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,505
Likes
10,012
Location
North-East
Oooh, I may have to discount DeltaWave upsampler after all! :)

SoX -u flag produced an even better match to PGGB. The difference between PGGB and SoX is now well below -200dB, unweighted (rms of -207dBFS and -209dBA):

1677784788595.png
 

fatoldgit

Active Member
Joined
Feb 29, 2020
Messages
263
Likes
310
Ouch, the steep (-s) conversion clipped and that's why the noise floor was elevated. After fixing that, the noise floor is now comparable to the non-steep version (and the same applies if you use it together with -v or -u):

I may be wrong... but "-s" isnt a quality setting... so without a quality setting specified your quality will default to "-h" (and the related combo settings it drags in, as per my original post above)

"-s" relates to Bandwidth (Audio Passband Percentage):

-s = -3 dB 99% (steep filter)
-b <74 .. 99.7> = -3 dB percentage
-B <53 .. 99.5> = 0 dB percentage

I would imagine you need to use a quality setting and "-s" in combo?. I theory "-u -s" gets you the good stuff from "-u" (as per my original post above) plus a steeper filter.

Maybe play around with that?

Also in theory "-b 99.7" is an even steeper passband!!!

Peter
 

fatoldgit

Active Member
Joined
Feb 29, 2020
Messages
263
Likes
310
Yes, that's what I meant by "use it (the -s) together with -v or -u."

cool.

Hate to think you spend time doing the analysis to find the wrong params were in play.

I am really enjoying this thread... like many I use sox over other options (although I have play around with camilla) so its great to have empirical evidence that sox is very very good....and its free!!!.

Peter

PS. might want to change the graphic to read "-u -s"... or are you saying that "-s" buggers up anything it touches (maybe overrides all other options when used)?
 

danadam

Addicted to Fun and Learning
Joined
Jan 20, 2017
Messages
932
Likes
1,473
PS. might want to change the graphic to read "-u -s"... or are you saying that "-s" buggers up anything it touches (maybe overrides all other options when used)?
I'm not sure I understand. The gray graph is the result of "rate -s" (implied "-h") conversion. Now that I fixed the clipping in the generated files, its noise floor is more or less the same as the noise floor of the magenta graph, i.e. "rate" (implied "-h") conversion. So as the documentation says, "-s" only affects the passband.

If I plotted the result of "rate -u -s" conversion, it would be more or less like the green graph, only with wider passband.
 

manisandher

Addicted to Fun and Learning
Joined
Nov 6, 2016
Messages
648
Likes
598
Location
Royal Leamington Spa, UK
SoX -u flag produced an even better match to PGGB. The difference between PGGB and SoX is now well below -200dB, unweighted (rms of -207dBFS and -209dBA)

Is it clear that PGGB and SoX are more accurate than DeltaWave?

(Not that it matters in the least bit for all practical purposes. Purely out of interest...)

Mani.
 
Top Bottom