DSP speaker with analog input only, measurement of signal degradation?

[ReaderZ]

So these go though D to A to D to A process, signal should be degraded compare to just D to A once. But people seems like them just fine, although also see people who has DSP speaker that take both digital and analog input say digital input is much better.

I doubt we can measure the actual dsp speakers. But set aside dsp and class d amp etc. Could we chain together 2 DAC and 1 ADC and get same type of measurement as we are measuring DAC? Would be very interesting to see the effect compare to going though DAC only once.

 

[Blumlein 88]

You can listen to music that has gone thru 8 DA/AD conversions.

https://www.audiosciencereview.com/...-choose-the-8th-generation-digital-copy.6827/

Basically, going thru once will raise the noise floor a bit maybe 3 db. Distortion will increase depending of course on the DAC and ADC used. It might for instance go from .005% to .01% neither of which is really a big concern.

 

[ReaderZ]

You can listen to music that has gone thru 8 DA/AD conversions.

https://www.audiosciencereview.com/...-choose-the-8th-generation-digital-copy.6827/

Basically, going thru once will raise the noise floor a bit maybe 3 db. Distortion will increase depending of course on the DAC and ADC used. It might for instance go from .005% to .01% neither of which is really a big concern.

Thanks for the answer and files, I am need to go out so just did a quick listen of of Ry coder one and I think B is the reference. going to go though them all and see if I can tell.

Edit: Phil wood A is reference. Jennifer Warnes B is reference, can't tell on cowboy
Bob Marley A is reference


Edit 2: Difference is small when it come to able to id them. But doable, I don't think I can do it if it's only twice.


I do think able to identify is a very low standard, if it's identifiable then the degradation is huge and completely unacceptable. But then again yours is 8 times and DSP speaker only do it twice.


I think lots of time we are mixing up audible and identifiable, just because there is an audible different doesn't mean people will be able to abx it. I think there may even be another level in between, that is sometimes you may notice a difference but can't id it reliably.


Even now that I know I can't tell the difference, I still feel put off by dps speaker without digital in... I just feel that way...

 

[andreasmaaan]

Thanks for the answer and files, I am need to go out so just did a quick listen of of Ry coder one and I think B is the reference. going to go though them all and see if I can tell.

Edit: Phil wood A is reference. Jennifer Warnes B is reference, can't tell on cowboy
Bob Marley A is reference

Edit 2: Difference is small when it come to able to id them. But doable, I don't think I can do it if it's only twice.

I do think able to identify is a very low standard, if it's identifiable then the degradation is huge and completely unacceptable. But then again yours is 8 times and DSP speaker only do it twice.

If you simply guess once which is the reference in each pair - even if you guess right - it won't be possible to know whether your guess was right because (a) you could hear a difference or (b) simply because of chance.

You need to listen to each track using an ABX comparator (e.g. this one for Foobar) and then make enough guesses for a statistically significant pattern to appear (or not, of course) in the outcomes. 15 guesses per pair is a valid minimum.

 

[Blumlein 88]

Thanks for the answer and files, I am need to go out so just did a quick listen of of Ry coder one and I think B is the reference. going to go though them all and see if I can tell.

Edit: Phil wood A is reference. Jennifer Warnes B is reference, can't tell on cowboy
Bob Marley A is reference


Edit 2: Difference is small when it come to able to id them. But doable, I don't think I can do it if it's only twice.


I do think able to identify is a very low standard, if it's identifiable then the degradation is huge and completely unacceptable. But then again yours is 8 times and DSP speaker only do it twice.


I think lots of time we are mixing up audible and identifiable, just because there is an audible different doesn't mean people will be able to abx it. I think there may even be another level in between, that is sometimes you may notice a difference but can't id it reliably.


Even now that I know I can't tell the difference, I still feel put off by dps speaker without digital in... I just feel that way...

So you got all of them wrong? (except the Cowboy Junkies which you didn't find a difference)

Yes, by definition if you can ABX it correctly it must have been audible.

And most people think what you wrote that they will notice a difference they can't ID reliably. In fact if you do some of these the reverse happens. You'll have reached the point you do not think you are hearing a difference while still correctly choosing.

This form I used with three files is known as a triangle test. Common in food and fragrance industries.

The feeling about something is okay. Lots of us feel that way. You can save yourself some grief (and usually money) if you can mostly make decisions on audio gear that aren't about feelings. Like in this case, you can see the chance the extra conversion is a degradation you can hear is very low, but you feel that isn't right. However, it is correct. The feeling is the result of being human. When knowing things influence how you feel or even hear them even when they aren't so.

 

[ReaderZ]

If you simply guess once which is the reference in each pair - even if you guess right - it won't be possible to know whether your guess was right because (a) you could hear a difference or (b) simply because of chance.

You need to listen to each track using an ABX comparator (e.g. this one for Foobar) and then make enough guesses for a statistically significant pattern to appear (or not, of course) in the outcomes. 15 guesses per pair is a valid minimum.

Going to try that.

15 out of how many? If 15 out of 15 that's well beyond unacceptable level of difference. IMHO if I can get it right 8 out of 10 times it's already completely unacceptable level of difference. Otherwise I would buy a sub 100 DAC and listen to music on youtube.

 

[Blumlein 88]

If you simply guess once which is the reference in each pair - even if you guess right - it won't be possible to know whether your guess was right because (a) you could hear a difference or (b) simply because of chance.

You need to listen to each track using an ABX comparator (e.g. this one for Foobar) and then make enough guesses for a statistically significant pattern to appear (or not, of course) in the outcomes. 15 guesses per pair is a valid minimum.

Yes and no. As I said this was a triangle test. It wasn't enough choices to be statistically valid alone. It was more of a chance for people to see that an 8th generation copy isn't so degraded as they might imagine. And these few would be statistically significant if enough people took the choices under similar enough conditions. This was just for fun as much as anything however.

 

[ReaderZ]

So you got all of them wrong? (except the Cowboy Junkies which you didn't find a difference)

Yes, by definition if you can ABX it correctly it must have been audible.

And most people think what you wrote that they will notice a difference they can't ID reliably. In fact if you do some of these the reverse happens. You'll have reached the point you do not think you are hearing a difference while still correctly choosing.

This form I used with three files is known as a triangle test. Common in food and fragrance industries.

The feeling about something is okay. Lots of us feel that way. You can save yourself some grief (and usually money) if you can mostly make decisions on audio gear that aren't about feelings. Like in this case, you can see the chance the extra conversion is a degradation you can hear is very low, but you feel that isn't right. However, it is correct. The feeling is the result of being human. When knowing things influence how you feel or even hear them even when they aren't so.

Ahh, I got them all right save that one I can't tell at all, brain farted wrote they are reference, meant they are different.

But going to try the foobar thing to see if I just got lucky or not.

 

[Blumlein 88]

Going to try that.

15 out of how many? If 15 out of 15 that's well beyond unacceptable level of difference. IMHO if I can get it right 8 out of 10 times it's already completely unacceptable level of difference. Otherwise I would buy a sub 100 DAC and listen to music on youtube.

The ABX comparator in Foobar will tell you the chance you are guessing. Generally 95% or less than 5% chance of guessing is the hurdle to get across. You need 15 or 20 for the statistics to be reliable at a minimum. My personal preference is even higher than that, but that is just me.

 

[andreasmaaan]

Yes and no. As I said this was a triangle test. It wasn't enough choices to be statistically valid alone. It was more of a chance for people to see that an 8th generation copy isn't so degraded as they might imagine. And these few would be statistically significant if enough people took the choices under similar enough conditions. This was just for fun as much as anything however.

Yes, I should have been more clear. What I was suggesting was that an ABX test needs to be performed using these files to establish that @ReaderZ's belief that A or B is the reference is not (or technically, is unlikely to be) attributable to chance.

 

[ReaderZ]

Yes, I should have been more clear. What I was suggesting was that an ABX test needs to be performed using these files to establish that @ReaderZ's belief that A or B is the reference is not (or technically, is unlikely to be) attributable to chance.

OK, I can't get it work now so I just wen to do normal AB again and even knowing my own choice the first time, I would give a different answer on Ry coder , even I know it's the wrong answer... So I don't need foobar's help to know I can't really tell...


So I should have waited for the E30 and buy the footprint 01, same budget... More bass to go boom and dsp to play with... Or a genelec SAM.

 

[andreasmaaan]

IMHO if I can get it right 8 out of 10 times it's already completely unacceptable level of difference.

Actually, 8 out of 10 would not even be proof that there is any difference discerned, let alone an indication of the degree of alleged difference.

In fact, there is a >5% chance of getting 8 out of 10 correct even if no difference is discernible.

A significant result is generally taken to mean 95%+ confidence. This requires that the number of correct responses exceed:

For 10 trials, this would mean giving >8 (i.e. 9) correct responses.