Generational losses

[bachatero]

I'm interested in the idea of generational losses with regards to overall audio quality. I know Audiophiles (with a capital A) love to harp on tiny differences in THD, IMD, frequency response, and so on. However, it seems like if you have a lot of components with barely imperfect performance, then those losses as a whole might audibly impact the end result.

Here's a potential chain of losses for my gym sound system:

Audio file compression
\/
Phone/player DSP
\/
Bluetooth compression
\/
BT receiver DSP
\/
BT receiver DAC
\/
Analog audio mixer
\/
Speaker ADC
\/
Speaker DSP
\/
Speaker DAC
\/
Speaker amplifier

That's a whopping 10 steps! Compare this to something like a portable wired headphone setup:

Audio file compression
\/
Phone/player DSP
\/
Phone/player DAC
\/
Phone/player amplifier

Only 4 steps! That's a huge W, and perhaps why stuff like DAC THD and lossy compression doesn't matter so much for headphones, because those losses are too small to be audible.

Note that I'm not counting stuff like cables, because those make far too little of a difference to matter unless you have a really long run of unbalanced unshielded cable in a noisy environment, which is almost never the case.

Therefore, I'd like to know the following:

1. If minimal improvements in each generational step improve the end result audibly
2. If it's possible to minimize the number of steps in "complex" setups like the 10 step one above

What do you think?

 

[NTK]

To demonstrate how much audio signal processing is a solved problem, long time member @Blumlein 88 had prepared (in 2016) test clips where you can compare a clip that had gone through 8 generations of ADC/DAC to its original. See if you can tell which is which.

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

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...

www.audiosciencereview.com

 

[fpitas]

Is this one of those, "People we have lost this year" threads?

/

 

[bachatero]

Is this one of those, "People we have lost this year" threads?

/

That's why I'm asking, because either generational loss is relevant and minute differences can add up, or they don't really matter and so even a long chain is a complete nothing burger.

 

[JustJones]

If you're streaming from a service you can add a couple more generations

 

[fpitas]

That's why I'm asking, because either generational loss is relevant and minute differences can add up, or they don't really matter and so even a long chain is a complete nothing burger.

Looks like a good thread linked up there. I very much doubt the losses are meaningful anymore.

 

[bachatero]

I just tried the 8-generations test and didn't get it right so I suppose that really means ADC/DAC conversions are too tiny to be relevant. I also compared the spectrograms and they are almost identical, which provides further evidence it doesn't matter. Done!

 

[Timcognito]

Doesn't the 1's and 0's argument kill all that?

 

[bachatero]

Doesn't the 1's and 0's argument kill all that?

It does. However, as ADCs and DACs interface to analog components, there are always going to be nonzero losses by doing those conversions. That's why it's becoming a little more common in the pro audio world to have completely digital setups to turn generational losses down to 0 as well as control everything on a network.

 

[BobbyTimmons]

Some of the losses are not like others. ADC and DAC shouldn't be audible. Bluetooth could be.

 

[dasdoing]

To demonstrate how much audio signal processing is a solved problem, long time member @Blumlein 88 had prepared (in 2016) test clips where you can compare a clip that had gone through 8 generations of ADC/DAC to its original. See if you can tell which is which.

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

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...

www.audiosciencereview.com

here are 60 generations of d/a/-a/d on almost 20 year old hardware https://www.zmix.net/Converter_Test/index.htm

 

[Waxx]

As long as you stay digital in a lossless format without processing, there are no losses, it's only in the analog part it can be the case (which also includes the analog parts of the adc or dac).

Bleutooth and other format conversions can cause lossess altough, so i try to avoid those (I only use BT for phonecalls in the car, and to stream radio in my kitchen that is noisy anyway while cooking. DSP can also cause losses, depending on how it's processed, but most modern won't do that. Idem with ADC or DAC processing. In a modern hi res sytem a few ad/da conversions should not matter at all.

And the weakest link in the chain is still your speaker, not the path towards it that is way higher in resolution and lower in noise than even the best speaker arround...

 

[bachatero]

Bleutooth and other format conversions can cause lossess altough, so i try to avoid those

I know this is in fact a controversial subject because the basic SBC codec has nearly imperceptible artifacts for most kinds of music. However, I recently did a listening test using a cheap little BT receiver box and found you could obviously hear UGLY non-harmonic artifacts with music with lots of bass. Example: Mars Needs Women (Griffin Boyce Remix)
Because that distortion is obviously audible I'm swearing off SBC forever, or at least in the gym sound system because they love that kind of bassy music.

 

[BobbyTimmons]

And the weakest link in the chain is still your speaker, not the path towards it that is way higher in resolution and lower in noise than even the best speaker arround...

The weakest chain in the link is usually not the speaker within reason, it's the source material.

 

[bachatero]

The weakest chain in the link is usually not the speaker within reason, it's the source material.

This is tricky if your sound system is the means to an end that which is the source material. Then, no matter what kind of thing you're listening to, you want to listen to that material as accurately as possible. That means reproducing the imperfections present during recording as faithfully as possible. The problem with this argument comes from those who prefer reproducing "real life" as much as possible, in which case the source material is a means to that end as well.

Personally, I'm a "recordings are the end" type of person, because I'm used to live performances if I want the real life experience, because I'm a musician first and audio person second. However, I can see where you're coming from if you want recordings to be the means to the realism end, and that's just fine because it's a personal preference.

 

[Cbdb2]

There are also files out there that compare music thru 10 opamp buffers to the original, again no difference.
The only place there might be a problem is if there is data compression (like mp3) at lower data rates.