8th generation ADC/DAC loop vs the original.....can you hear it?

[Blumlein 88]

Download this zip file. It will unzip into a folder with three 30 second snippets of a Fiona Apple song. Two are the original digital file. One is an 8th generation loop thru DAC/ADC. Pick the odd file out on the poll.

https://www.dropbox.com/s/vod0d289ncr5lck/Fiona Apple Gen 8.zip?dl=0


Or alternatively download this zip of Hyperion Knight piano playing Mussorgsky. Recorded by Wilson Audio with all Spectral electronics.

https://www.dropbox.com/s/m6wpv0umbesgpvb/Hyperion Knight Gen 8.zip?dl=0

 

[fas42]

Not having much luck, 88, eh ... ?

 

[Blumlein 88]

Not so much.

 

[amirm]

I just tested the second track with foobar ABX:

foo_abx 1.3.4 report
foobar2000 v1.3.2
2016/05/01 18:13:04

File A: C:\Users\Amir\Documents\Test Music\Blum Test\Hyperion Knight Gen 8\Hyperion Knight HHHH.wav
File B: C:\Users\Amir\Documents\Test Music\Blum Test\Hyperion Knight Gen 8\Hyperion Knight IIIIIIII.wav

18:13:04 : Test started.
18:13:58 : 01/01 50.0%
18:14:05 : 02/02 25.0%
18:14:13 : 03/03 12.5%
18:14:18 : 04/04 6.3%
18:14:25 : 05/05 3.1%
18:14:31 : 06/06 1.6%
18:14:37 : 07/07 0.8%
18:14:46 : 08/08 0.4%
18:15:08 : 09/09 0.2%
18:15:15 : 10/10 0.1%
18:15:24 : 11/11 0.0%
18:15:31 : Test finished.

----------
Total: 11/11 (0.0%)

I renamed the files in the report so that people can't tell from my result which is which. As you can see, I could obviously tell them apart with extreme certainty. Why? Because there is a very obvious "tell" in this test .

Without it, with my highly plugged ears, I have not yet been able to tell the files apart.

 

[Blumlein 88]

So, with a tell you managed it, but I take it with just normal listening back and forth you didn't hear them as different?

BTW, thanks for doing this, discussion like this is what I hoped to elicit.

One more question, I'll send in a conversation.

 

[TBone]

So, with a tell you managed it, but I take it with just normal listening back and forth you didn't hear them as different?

BTW, thanks for doing this, discussion like this is what I hoped to elicit.

One more question, I'll send in a conversation.

I did the same with the Fiona Apple track, "thought" I heard slightly less clarity / transparent in one particular track played thru my computer, so I cheated and I found the "tell" aligning with the track I chose. Listening to them again confirmed what I first heard, but then I knew the answer so that finding was neither here nor there. That said, even tho I know, I should give em a listen on my main rig just to hear if these differences are that much more appreciable.

 

[amirm]

So, with a tell you managed it, but I take it with just normal listening back and forth you didn't hear them as different?

Thats right. This is peak allergy season for us now and despite taking two kinds of allergy medication, my ears are heavily plugged. The difference between blowing my nose and not is huge. Going to an ear-nose-throat doctor tomorrow hoping he may have a solution to this. Failing that, critical listening needs to wait till later in the year .

For now, these are hard tests to pass.

 

[Blumlein 88]

Alright the poll is closed. If anyone downloaded without voting,

Spoiler

file Z was the 8th generation copy. It did have a reasonably obvious tale that was left in by mistake. Still not so easy.

 

[DonH56]

Amir, AVS has a "spoiler" feature that would allow Blumlein 88 to "hide" the answer until you clicked on it. Anything like that available here? New folk could try without bias, and old folk like me who never seem to have free time could delay the inevitable (still working tonight at nearly 10 pm, found out today my vacation day tomorrow I need to work, as well as Saturday; sometimes Work really wreaks havoc on Life!)

 

[Blumlein 88]

Amir, AVS has a "spoiler" feature that would allow Blumlein 88 to "hide" the answer until you clicked on it. Anything like that available here? New folk could try without bias, and old folk like me who never seem to have free time could delay the inevitable (still working tonight at nearly 10 pm, found out today my vacation day tomorrow I need to work, as well as Saturday; sometimes Work really wreaks havoc on Life!)

Working to live or living to work?

I am so glad I am retired. Life is much better that way.

If you really want to try it. I can rescramble the names and make the files available to you. Let me know.

 

[amirm]

Amir, AVS has a "spoiler" feature that would allow Blumlein 88 to "hide" the answer until you clicked on it. Anything like that available here? New folk could try without bias, and old folk like me who never seem to have free time could delay the inevitable (still working tonight at nearly 10 pm, found out today my vacation day tomorrow I need to work, as well as Saturday; sometimes Work really wreaks havoc on Life!)

Yes, just manually use the "spoiler" tag with brackets around it. Here it is:

Spoiler

this is a test

And how I did it:

[spoiler]this is a test[/spoiler]

 

[amirm]

I edited the results to be a spoiler for the future people who may want to test.

I say without the "tell" that I found, this test was very difficult. It should be an eye opener for anyone thinking there is something fundamentally wrong with digital. Blumlein, can you leave the files available for download for a while?

 

[Phelonious Ponk]

I wonder if a lot of people, like myself, listened but didn't respond because they couldn't differentiate the files.

Tim

 

[amirm]

I wonder if a lot of people, like myself, listened but didn't respond because they couldn't differentiate the files.

Tim

I certainly was one of them. On multiple occasions I tried to find a difference but could not. Now, my ears are very plugged but with respect to mid to low frequencies, in ABX testing, it was darn near impossible.

Note that in ABX testing I could often get three answers right in a row but then fail to do so from there on. So it is very easy in a simple one-trial AB test to think you can tell them difference but one requires more iterations to know for sure.

I should say this kind of test is made harder than it should be though . We should provide the original and then ask which file is the same and which is 8th generation away.

 

[RayDunzl]

I wonder if a lot of people, like myself, listened but didn't respond because they couldn't differentiate the files.

There wasn't a poll choice for that.

Being deaf, I just looked at them.

It was pretty close visually, both levels and wave... ignoring what I assume was the "tell".

The 8th generation file is very very slightly shorter than the original. The generations created a very slight decrease in the length of the file. For example, at 22.084 seconds in the original, the 8th generation is at 22.070 seconds (after alignment at the beginning, which precludes performing a strong 'null' on them (by inverting one and combining).

There seems to be a little phase slippage (my guess), the levels ended up very close. Phase - maybe the change is similar to sitting another inch or three from the speaker, maybe it's all there, but relocated.

Here's my best effort at nulling the difference in one channel near the start after best-guess time alignment within this segment.

 

[Blumlein 88]

Yes I can leave the files up. I also could remove the tale.

I removed a bit over 100 bits of one original file and 200 of another vs the 8th gen. This to prevent people from easily seeing which two were the same. Other than the chopping of samples they are the same for the two originals.

Now on another forum, I had some claim my removing those samples altered the entire sound and they indeed sounded different. More different than the 8th gen file which is certainly interesting as a Rorschach test.

I too was surprised 8 generations of conversion were this close. And with two pieces of gear costing about $400 each. So if we get that with non-exotic bog simple low priced gear, well what is it we are getting with more expensive gear? Yes it is measurably better, but does it do anything we can really hear in a one generation playback? Or put another way, who needs MQA, just quit mucking up the bits between recording and what we get to purchase.

 

[Blumlein 88]

I certainly was one of them. On multiple occasions I tried to find a difference but could not. Now, my ears are very plugged but with respect to mid to low frequencies, in ABX testing, it was darn near impossible.

Note that in ABX testing I could often get three answers right in a row but then fail to do so from there on. So it is very easy in a simple one-trial AB test to think you can tell them difference but one requires more iterations to know for sure.

I should say this kind of test is made harder than it should be though . We should provide the original and then ask which file is the same and which is 8th generation away.

Amir, this kind of test I did was a triangle test. If you provide an original, and ask for people to choose between another original and the 8th generation that would be a duo-trio test. Either can be made to work and might be used for various reasons. Triangle is considered more discriminating when looking for non-specific differences. Duo-trio might be used with a specified aspect to listen for like "pick which file has the deepest bass".

I might try one with a tetrad format next. You are given two original files, and two 8th generation files and are asked to pair them up. The advantage here is if you can't pair things up you know you haven't heard it yet. I am thinking maybe try a 32nd gen vs originals. Assuming I don't go bonkers repeating this 32 times.

 

[RayDunzl]

Assuming I don't go bonkers repeating this 32 times.

Borrow a 32 channel digital mixer and patch each analog monitor out to the next analog in.

 

[Blumlein 88]

Borrow a 32 channel digital mixer and patch each analog monitor out to the next analog in.

That is a great idea. I have an 8 channel one. So doing stereo, I could do 4 trips at once. That wouldn't be too hard.

 

[RayDunzl]

I wonder how much of the degradation (what little there is) is from the cable.