Bluesound Node Icon (Quick User Measurements)

[Sokel]

Nope.
It's purely software related.

In digital, it's also there.
Even worse.

It's a good news, since the only thing required to fix it is a software upgrade.

Seems that doing EQ is not as easy as it seems.MiniDSP struggled for a decade or more with problems down low and we see hits at other gear too.
A pristine implementation like the RME gear have is probably a hint about the knowledge (+hardware) it takes to do it right.

 

[pogo]

It's purely software related.

Are you sure about that?
A hardware chip could also have been used to realize the tone controls.

 

[Rja4000]

Are you sure about that?
A hardware chip could also have been used to realize the tone controls.

I don't think so here.
We have the exact same issue from early Node N100 to the last Icon.

And even if they were using a DSP, that's still programmable.
(I'm pretty confident they don't. You really don't need to pay a full blend DSP chip to do basic tone control when you have a real CPU at hand anyway)

 

[pogo]

This could also indicate a hardware implementation

 

[mikematrix]

Do the wiims also have this loss with their EQ?

 

[Rja4000]

Do the wiims also have this loss with their EQ?

Easy to measure.
I don't have one, so can't check...

 

[Mnyb]

Do the wiims also have this loss with their EQ?

All digital systems DSP will have some inevitable loss . For example you want a slight bass shelf +2,5 dB and some bass notch filters to tame room resonances you then add some safety margin for intesample overs and what not by for example lower the total level by -5dB to accommodate you just lost 5 dB .

But that is expected .

What's unexpected is these other behaviors . Bad programming ? Sloppy programming ? It could be that they use some DSP chip that only works with integers and set coefficient's to that and are not careful and triggers some bad behavior .

If you do this in your PC with floating point math in well made software you only get the expected losses due to lowering the levels to accommodate your DSP or tone controls.

An interesting exercise for amir would be to measure this on a digital output spdiff or coax .

I dont know how good the WiiM's internal processing are . I would also like to know ?

Also we should acknowledge that it might not be that audible ( or at all audible ) good EQ and room correction is very audible so it would be foolish to not use these features anyway .

 

[GXAlan]

Not really.
In digital, SINAD should be around 146dB.
With bass correction, you loose 31dB SINAD !

Agree ...

Definitely, there's a loss of performance -- but the idea was that if the analog stages can deliver 116 dB SINAD, then you'd expect the digital algorithm dropping to 116 dB would still be able to deliver 116 dB at the analog stage... It doesn't.

An interesting exercise for amir would be to measure this on a digital output spdiff or coax .

What’s tricky is that I tested EQ but clicking the feature on in the app without making any changes.

@Sokel returned his MiniDSP flex because it did fine in native mode but had a massive performance drop once you enabled the filters. Then others measured it and showed that the problem occurred at specific frequencies and thresholds where >0.6 dB triggered more problems.

Post in thread 'Minidsp Flex Review (Audio DSP)'
https://audiosciencereview.com/foru...idsp-flex-review-audio-dsp.30804/post-1471541

So, you really have to test it.

 

[Sokel]

Agree ...

Definitely, there's a loss of performance -- but the idea was that if the analog stages can deliver 116 dB SINAD, then you'd expect the digital algorithm dropping to 116 dB would still be able to deliver 116 dB at the analog stage... It doesn't.




What’s tricky is that I tested EQ but clicking the feature on in the app without making any changes.

@Sokel returned his MiniDSP flex because it did fine in native mode but had a massive performance drop once you enabled the filters. Then others measured it and showed that the problem occurred at specific frequencies and thresholds where >0.6 dB triggered more problems.

Post in thread 'Minidsp Flex Review (Audio DSP)'
https://audiosciencereview.com/foru...idsp-flex-review-audio-dsp.30804/post-1471541

So, you really have to test it.

To be exact,I didn't measured digital performance,only analog.

I implemented a simple L-R24dB x-over at 240Hz similar to the electronic one my speakers have,along with 3 corrections down low.
SINAD with those at place was in the 80's with a -3dB level loss (as I never correct for dips) but I also measured a long range of levels.

And that was the reason of the nice thread about the problem who led to a partial fix by miniDSP.

So...

 

[sapphari]

It's obviously a software bug.
It only exists for bass correction.
Trebble correction is done correctly.

Regarding treble correction being done correctly, doesn't post #45 show that adjusting treble by -0.5 dB yield 78.2 dB, which is far worse than no adjustment?

Post #19 shows no tone controls gives 116.6dB, and enabling tone controls (but not making any adjustments) drops to 112dB. But then post #45 shows that actually adjusting treble results in a drop to 78.2dB. Am I missing something here or does adjusting treble actually have a significant affect?

 

[Mnyb]

@Sokel returned his MiniDSP flex because it did fine in native mode but had a massive performance drop once you enabled the filters. Then others measured it and showed that the problem occurred at specific frequencies and thresholds where >0.6 dB triggered more problems.


So, you really have to test it.

Wow , no understatement there you really have to test , the manufacturer should have at least ? This requires automation to test ? small db and frequency increments ?

 

[Rja4000]

Regarding treble correction being done correctly, doesn't post #45 show that adjusting treble by -0.5 dB yield 78.2 dB, which is far worse than no adjustment?

Post #19 shows no tone controls gives 116.6dB, and enabling tone controls (but not making any adjustments) drops to 112dB. But then post #45 shows that actually adjusting treble results in a drop to 78.2dB. Am I missing something here or does adjusting treble actually have a significant affect?

In all my measurements, trebble correction gives no issue at all.

 

[GXAlan]

Post #19 shows no tone controls gives 116.6dB, and enabling tone controls (but not making any adjustments) drops to 112dB.

But then post #45 shows that actually adjusting treble results in a drop to 78.2dB. Am I missing something here or does adjusting treble actually have a significant affect?

That’s what I measured.

If you look at the graph alone, it doesn’t visually look the way the bass tweaks do (with that huge difference in noise to the left of 1 kHz) but if you look at the actual SINAD numbers, it drops a lot.

Wow , no understatement there you really have to test , the manufacturer should have at least ? This requires automation to test ? small db and frequency increments ?

Exactly. Cannot just turn it on and leave it at +0 dB. It’s tempting to think you found the impact.

 

[sapphari]

That’s what I measured.

If you look at the graph alone, it doesn’t visually look the way the bass tweaks do (with that huge difference in noise to the left of 1 kHz) but if you look at the actual SINAD numbers, it drops a lot.

Looking at the noise floor for tone controls on (with no tone adjustments) vs a treble adjustment, the floor raises 40dB. The same happens when bass is adjusted, but it also adds another 15dB to the noise floor in the 20-200hz region. The highest amplitude of a 1kHz harmonic also increases >20dB if either bass or treble is adjusted (relative to tone controls on with no adjustment). So it seems clear that adjusting either bass or treble has significant affects to performance, with bass being worse than treble.

@Rja4000 Would you mind posting or linking to your measurements? It would be good to try to understand the discrepancy.

 

[Sokel]

Wow , no understatement there you really have to test , the manufacturer should have at least ? This requires automation to test ? small db and frequency increments ?

You can imagine that a device who handles multiple levels has it's very own gain structure.

Now,combine that with a half-baked (for years,miniDSP forum is full of complains about performance down low) filter implementation plus the intersample overs (fixed as they moved the VC in a prior stage after the thread) plus the analog sins and one can never be sure without testing it as it sits in real world conditions.

It's my habit to do some sanity measurements before putting anything in the chain,first the basics with a DMM and then through a simple loopback.
Flex is the only thing I returned in my whole life.Didn't make any sense as I don't need delays or elaborate steep filters and as long as the stock el. crossover did more than 20dB better in use.

Best buy ever at this hobby is something who let you see the performance,doesn't need to be elaborate like an AP,just a decent interface who can detect some basic problems.

 

[bluefuzz]

So, you really have to test it.

On my NAD C658, which also uses BluOS, you can point a browser at <IP address of your device>/diag?print=1 and see the software diagnostics. It gives an idea of what's going on.

I think BluOS just uses the linux built in Alsa mixer and FFMPEG for the internal tone controls ...

 

[Rja4000]

On my NAD C658, which also uses BluOS, you can point a browser at <IP address of your device>/diag?print=1 and see the software diagnostics. It gives an idea of what's going on.

I think BluOS just uses the linux built in Alsa mixer and FFMPEG for the internal tone controls ...

Could you copy it here ?

 

[Mnyb]

Ok it's then clear to me that any product with DSP must be tested with DSP in active use , the RAW DAC performance does not seem to be the limiting factor it's the "other" stuff that makes or breaks these things .

How does an AVR fare ? They may actually be better ?

 

[rdytro]

Aside from further exposing general sloppiness of this product / BlueSound's software, I think these tests have been invaluable in shedding some light on avenues of enquiry which would be useful for Amir to further explore.

 

[bluefuzz]

Could you copy it here ?

You mean the diag output? Not really relevant for this thread as it's a different device and there's pages and pages of it. It's basically the output of dmesg and /var/log/messages. It's no secret that bluOS is based on linux/busybox. As others have pointed out before bluOS is in principle open source and Lenbrook are legally obliged to supply the source to anyone who asks. I'm pretty sure there used to be a Github repo for bluOS but it seems to have disappeared ...