• 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 many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Measured Differences Between Software Audio Players

bibo01

Active Member
Joined
Mar 10, 2016
Messages
109
Likes
18
I was not at all interested in seeing the difference on the vertical axis. What interested me above anything else was establishing a (approximate) "value" on the X axis (Time axis) as the difference was rather macroscopic.

After all, a 14mV deviation (on vertical axis) is equivalent to a difference of about 0,12 dB! As shown in many graphs (analyzed by software programs with high discretization), the files also show differences in amplitude. Such differences, given that the files (let me repeat - several) were played by a subsequent DAC, could very well swing of 14mV (0,12db). Therefore, although your observation as some logic, in this case I do not agree with it.
With a cyclic signal my device running "empty" (sorry for bad translation) has a very low error on Y, despite its 8/12 bits of vertical resolution, which it is also found in oscilloscopes of $350,000.

I usually use a setup with noise measurement (on FFT at 1,048,576 points) of approximately -170dB, so resolution is very high...like your AP which I have also directly compared to with identical results.

For each analysis the most "appropriate" tool must be used.

Here we are talking about differences between two players that have shown clear differences during listening (in blind too). Indeed those differences were also found between the two analog signals reproduced by each player.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,385
Location
Seattle Area
I was not at all interested in seeing the difference on the vertical axis. What interested me above anything else was establishing a (approximate) "value" on the X axis (Time axis) as the difference was rather macroscopic.

After all, a 14mV deviation (on vertical axis) is equivalent to a difference of about 0,12 dB!
You say one thing but compute the opposite. Voltage differential is governed by the dynamic range of the ADC, and not accuracy of its time base. As I keep mentioning, the former which you are relying upon is only 8 bits worth and as such, subject to substantial amount of noise. The scope is only highly accurate on its time axis which is measured in units of time, not voltage. Your x-axis cursors seem arbitrary to me so not sure what it is supposed to show.

What you are stating is differential in vertical axis even though you say it is X axis. There, the first delta x = 129 mv and the second = 115 mv. Subtracting one from the other we get 14 mv which you are relying upon. So you are reading the measurements wrong I am afraid.

All you are measuring there is the scope noise+DAC noise. The former is quite high seeing how it is just an 8-bit ADC and hence using a lot of dither. And there is potential for the scope probe picking up noise from the close by computer as it runs one operating system versus another. You lack a control here.

Going with your data anyway, you say 14 mv is bad. How about the 115 mv? That is the "jitter" that you say you have measured with the "better" player. Surely if you are worried about 14 mv, you better be screaming about 115 or nearly 8X more.

Here we are talking about differences between two players that have shown clear differences during listening (in blind too).
Don't believe that at all. There is no way you can perform any kind of reasonable AB blind tests with a full system reboot between the two players involved.

Indeed if you have that proof, it is far, far more valuable than these feeble attempts at measurements. Let's see details of that. How many trials were conducted? What music selections? And what test conditions?
 

ceedee

Active Member
Joined
Jun 8, 2016
Messages
105
Likes
32
Location
DFW, TX
Don't believe that at all. There is no way you can perform any kind of reasonable AB blind tests with a full system reboot between the two players involved.
Yes I'm skeptical of this as well. It seems very difficult to test. I wonder what the best way would be … maybe recording the digital (or analog) outputs of the computer and then ABX those in software? A hardware ABX could work but then you'd need two different computers which would make the results unreliable.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,385
Location
Seattle Area
There are a lot of barriers to doing it right and hence my disbelief. For example, if you used two computers, the clocks will drift and pretty soon they won't be in sync making it easy to guess which is which in an ABX test.

Analog capture using an ADC would be one approach but if two media players sound different, not sure how we say that capture works in a neutral manner. But would be a start.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
A glitch in my internet caused me to loose a large post with lots of pictures. I'll recap here. The topic is the WTF vs Foobar testing bibo01 has been telling us about.

You can deal with drifting clocks on repetitive signals like these triangle waves. Align them as best possible and them speed one up or slow it down. By tiny percentages. You can do this in Audacity and being an Audacity cowboy that is what I did.

Firstly the 44 khz files.
I found with basic alignment you could only get nulls of the fundamental of 47-48 db. I could then alter speed of one file and get places with nulls of 70 db. You know there is still clock misalignment and real nulls are deeper than that. It looks like this. The small thin segment is where the timing of samples gets closest to perfect. This particular one comparing W2-002 and F2-009. I got similar results all the way around. I took the W2 track combined it with the inverted F2 track after adjusting the speed to get this result in the screenshot.


Going with locked clocks you could get better than 100 db nulls which presumably is the basic noise floor of the gear. This with Foobar or WTF files. The fundamental is barely above the noise floor and all harmonics are buried in the noise as you would expect of locked clocked samples.

There is some problem with the locked 192 khz files. The timing varies with null depths being different from the first couple to the last couple seconds. Sometimes more than 20 db. And the null depths possible are not very good at all. Any results from locked 192 khz files need throwing out altogether. Is this the set of files used for the oscope traces?

The unlocked 192 khz files were fine and provided better results than the locked ones. Comparisons between files with speed changes to gain alignment resulted in 80 db deep nulls. And knowing if perfect alignment were done this would be better than that in fact.

So I see little reason to think these two softwares are acting differently. At least from this limited info to go upon. The problems with the locked 192khz files were worse with Foobar than WTF, but there is some issue there and those results should be ignored as neither set was acting as locked clocks should.

Maybe it is time to describe how the blind testing was carried out?
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,385
Location
Seattle Area
Thanks for the analysis. :) I gave up on Audiodiffmaker long time ago after it repeatedly crashed on me. Had not thought about using audio workstation software and only work on small segments. That's clever! :)
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,385
Location
Seattle Area
Holy Batman! I went to download WTF to kick its tires and was surprised to see that it only runs on Linux. But the shocking aspect was this in their feature list:
  • Fully command line driven. No graphical mode. No network either. The whole system is optimised for sound playback from local disks.
What kind of audio masochist would in this day and age use a command line tool to play one track at a time? No network share either?

I said there were costs to running alternate players but had no idea the cost was this high.
 
Last edited:

Phelonious Ponk

Addicted to Fun and Learning
Joined
Feb 26, 2016
Messages
859
Likes
215
Holy Batman! I went to download WTF to kick its tires and was surprised to see that it only runs on Linux. But the shocking aspect was this in their feature list:
  • Fully command line driven. No graphical mode. No network either. The whole system is optimised for sound playback from local disks.
What kind of audio masochist would in this day and age use a command line tool to play one track at a time? No network share either?

I said there were costs to running alternate players but had no idea the cost was this high.

The kind with Kool-Aid on his breath.

Tim
 

TBone

Major Contributor
Joined
Mar 16, 2016
Messages
1,191
Likes
348
my mistake, already aware of wtfplay for Linux, thought I read Windows.
 

Phelonious Ponk

Addicted to Fun and Learning
Joined
Feb 26, 2016
Messages
859
Likes
215
Amir we need to make the forum run in dos 6.0, make it super basic and charge for the privilege.. Change its name to... Winkershunt or GFY.

Will have millions of new members and make a killing :D

GFY? Don't know what that stands for but it would be hard to find an acronym more appropriate than WTF.

Tim
 

Sal1950

Grand Contributor
The Chicago Crusher
Forum Donor
Joined
Mar 1, 2016
Messages
14,068
Likes
16,598
Location
Central Fl
What kind of audio masochist would in this day and age use a command line tool to play one track at a time? No network share either?

I said there were costs to running alternate players but had no idea the cost was this high.
Didn't they teach you command line at Microsoft. LOL ;)
 

Sal1950

Grand Contributor
The Chicago Crusher
Forum Donor
Joined
Mar 1, 2016
Messages
14,068
Likes
16,598
Location
Central Fl
Go f yourself.. That's basically what your doing by using command line interface to play music:D
Nah, it removes the noise created by the GUI from modulating the data stream. :p
 
Top Bottom