Question regarding what measurements would need to be done?

[andyr]

I have active spkrs - with a miniDSP 'nanoDIGI' providing the XOs between subs & mid/bass drivers ... and mid/bass drivers & tweeters.  Also delay of the 'mains' ... as the subs are further away from my ears.

 

A recent upgrade to my system occurred when Roon 'Convolution' was used to add what I call a "FIR filter overlay" to the IIR filters which the nanoDIGI provides.

 

It is easy in my system to compare two situations:

1. Roon convolution enabled - and all PEQ in the nanoDIGI disabled (only keeping the XOs, the delays and the relative driver dB levels).
2. Roon convolution disabled - so using the PEQ in the nanoDIGI, which I developed during a long session with REW.

 

I - and everyone who has come to listen and heard the comparison - prefer #1.  The simple explanation I have for why it sounds better ... is that, as in #1, the inherent phase issues of IIR filters have been negated ... so the sound from the spkrs has more resolution.

 

But, as there a number of folk here who believe that only measurements can show whether what you are hearing one time is actually different to what you are hearing another time ... I would be most interested to know how REW graphs - or any other measurements - could be used back up the fact that #1 sounds better.

 

BTW, I've heard the same beneficial effect of Roon convolution in 2 other systems - where the spkrs are passive (with no miniDSP in sight).

 

 

Edited June 13 by andyr

[Mpr_65]

Hi Andy,

I'm no expert here but are you saying that the REW FR measurements are identical for each of the options 1 & 2?  

[andyr]

5 minutes ago, Mpr_65 said:

Hi Andy,

I'm no expert here but are you saying that the REW FR measurements are identical for each of the options 1 & 2?  

 

No Marty - I'm not saying that.  

 

All I'm saying is that I (and others) prefer the sound of option #2.  I have a simple explanation of why this should be so ... but I'd be interested in seeing measurements which will show why #2 really is better.  (BTW - I suspect the FR graphs of the two options will be the same ... although the phase graph should be different.)

 

[Mpr_65]

Thanks Andy, 

Don't you say in you original post you prefer option #1? 

 

On 13/06/2023 at 2:53 PM, andyr said:

 

I - and everyone who has come to listen and heard the comparison - prefer #1

 

 

[andyr]

30 minutes ago, Mpr_65 said:

Thanks Andy, 

Don't you say in you original post you prefer option #1? 

 

Correct!!  Sorry for the confusion?  I prefer the Roon Convolution.

 

[Mpr_65]

I guess in this case it might be better to post the REW graphs for options #1 and #2.  If there is a difference others (with more knowledge than me) might be able to interpret them.  If there is NO difference it would cause me to wonder.

As to what "sounds better" might be more difficult as it's quite a subjective thing and relative to your sample size.  I don't know how many people listened etc.   To extend this further we risk going down the DBT rabbit hole.  

Having said this I don't doubt your experience.

 

Or am I missing the point entirely?

[Keith_W]

Ahh, I see. You are trying to compare the results of convolution done in Roon, vs. convolution done by your NanoDigi. 

 

The method is simple - measure the output at the DAC. To do this, you need an interface like a Focusrite 2i2 (I have one that you can borrow). Plug the output from the DAC into the input at the interface, and compare the two curves. 

[BugPowderDust]

15 hours ago, Keith_W said:

The method is simple - measure the output at the DAC. To do this, you need an interface like a Focusrite 2i2 (I have one that you can borrow). Plug the output from the DAC into the input at the interface, and compare the two curves. 

In addition to this, record a loop of the same audio on both via the sound card, then open up Deltawave and compare. It's also worth grabbing a copy of Multitone to further diagnose what's changed.

[Grizaudio]

I'm confused, why measure at the DAC, how does this show one filter is superior to another, other than the outputs being different? 

You need to measure the system response in room.

 

Unless the two filters use the same coefficients, you can expect different in room measurements, which will in turn show in analysis graphs.

 

I would look at the important stuff, Impluse response (I would suspect the signal should settle faster using FIR), Spectrogram (the convolution 'should' provide a smoother response. I.e. less choppy), waterfall/bass decay (I would expect to see improvements in bass decay with either IIR or FIR), clarity (You should see a rising improvement in clarity with DSP done right). Obviously you can compare phase on the SPL/Phase response. 

 

How did you generate your convolution filter?

Did you use convolution with inversion? or is the convolution a set of PEQ's (I.e. different or the same as your subjective IIR DSP)

Is the FIR correction - correcting to a curve? or otherwise

 

It sounds like your IIR correction is purely subjective tweaking, so understanding what the FIR is correcting for is important. 

Edited June 18 by Grizaudio

[andyr]

Hi Griz,

 

On 18/06/2023 at 9:11 PM, Grizaudio said:

I'm confused, why measure at the DAC, how does this show one filter is superior to another, other than the outputs being different? 

You need to measure the system response in room.

 

I think Keith's thinking is that one gets a clearer picture of what effect the different filters have, if the effects of the room are taken out of the picture.

 

In my system, the DACs are the last things in the signal chain before the power amps.  My miniDSP 'nanoDIGI' sits in front of the DACs - and the PC running Roon Convolution is before the nanoDIGI.

 

On 18/06/2023 at 9:11 PM, Grizaudio said:

Unless the two filters use the same coefficients, you can expect different in room measurements, which will in turn show in analysis graphs.

 

The two filters are different; hopefully the following better explains how my system works:

• the nanoDIGI (coax digital in and out) provides 3-way XOs for the spkrs (2-way 'mains' plus a sub each side).
• these are standard 'IIR' filters - so exhibit a phase difference across the XO region.
• in Config #1 in the nanoDIGI, I also have the EQ I developed through REW measurements, plus delay for the mains relative to the subs (which are located further away from my ears).
• I also have Config #2; this does away with the EQ - but keeps the XOs and the delay.
• the Roon Convolution provides a 'linear-phase overlay' for the IIR filters - plus EQ to deliver a gradually falling FR, according to the specs provided by Con.

 

My interest is in finding measurement differences that can 'show' why Config #2 plus Roon Convolution ... sounds better than Config #1.

 

On 18/06/2023 at 9:11 PM, Grizaudio said:

I would look at the important stuff, Impulse response (I would suspect the signal should settle faster using FIR), Spectrogram (the convolution 'should' provide a smoother response. I.e. less choppy), waterfall/bass decay (I would expect to see improvements in bass decay with either IIR or FIR), clarity (You should see a rising improvement in clarity with DSP done right). Obviously, you can compare phase on the SPL/Phase response. 

 

Yes, I saw a difference in the phase graphs - but nothing that leapt out at me as being associated with "sounding better"!  

 

Likewise the waterfall graphs weren't very helpful.  

 

On 18/06/2023 at 9:11 PM, Grizaudio said:

How did you generate your convolution filter?

 

I'm afraid a friend did it - so I have noooo idea on how it was done.  

 

On 18/06/2023 at 9:11 PM, Grizaudio said:

Did you use convolution with inversion? or is the convolution a set of PEQ's (I.e. different or the same as your subjective IIR DSP)

 

Sorry - can't answer you.  

 

On 18/06/2023 at 9:11 PM, Grizaudio said:

Is the FIR correction - correcting to a curve? or otherwise

 

AIUI ... the Roon Convolution is:

1. "linear phasing" the IIR XOs at 110Hz and 2800Hz
2. and providing EQ to correct to a curve.

 

On 18/06/2023 at 9:11 PM, Grizaudio said:

It sounds like your IIR correction is purely subjective tweaking, so understanding what the FIR is correcting for is important. 

 

Hopefully, what I've said above shows you it's not just "subjective tweaking"!    We did an REW sweep with Config #2 (just the XOs and delay) loaded into the nanoDIGI ... and this provided input into Con's Convolution process.

 

[Keith_W]

On 18/06/2023 at 9:11 PM, Grizaudio said:

I'm confused, why measure at the DAC, how does this show one filter is superior to another, other than the outputs being different? 

You need to measure the system response in room.

 

Yeah, an in room measurement is required as well. Measuring output at the DAC tells you the overall electrical effect of the interventions. Measuring the result in the room tells you what the in-room result will be. I had assumed that Andy had already taken measurements of his in-room response. @andyr if you haven't, please accept my apologies. My bad. You also need to measure your in-room response. 

[Grizaudio]

Just now, Keith_W said:

 

Yeah, an in room measurement is required as well. Measuring output at the DAC tells you the overall electrical effect of the interventions. Measuring the result in the room tells you what the in-room result will be. I had assumed that Andy had already taken measurements of his in-room response. @andyr if you haven't, please accept my apologies. My bad. You also need to measure your in-room response. 

 

Yep that's what matters. 

[Grizaudio]

15 minutes ago, andyr said:

In my system, the DACs are the last things in the signal chain before the power amps.  My miniDSP 'nanoDIGI' sits in front of the DACs - and the PC running Roon Convolution is before the nanoDIGI.

Hi Andy, definitely understand. That's pretty typical of an active system. 

 

 

[andyr]

Yes, Keith - Con took an in-room measurement of Config #2 (no EQ - only XOs and delay) which he used as input to his Convolution program.

 

I also have REW measurements of Config #1 (with my EQ and with no Roon).

 

Edited June 23 by andyr

[Keith_W]

3 minutes ago, Grizaudio said:

Yep that's what matters. 

 

Measuring the in-room response is the first thing I do after I generate correction filters. This is a necessary step to verify that your corrections have indeed corrected what you intended. I do it as part of my workflow, and it is so ingrained in me that I assume that everyone else also does it  

[Keith_W]

1 minute ago, andyr said:

Yes, Keith - Con took an in-room measurement of Config #2 (no EQ - only XOs and delay) which he used as input to his Convolution program.

 

I also have REQ measurements of Config #1 (with my EQ and with no Roon).

 

 

Do you have the result to post? 

[andyr]

8 minutes ago, Keith_W said:

Do you have the result to post? 

 

Which REW graphs?

 

[Grizaudio]

@andyr I would be pretty certain you should see improvements in clarity (should be rising) post EQ, waterfall smoothness, improvements (reduction) in bass decay times and specifically time domain charts like impluse - if the correction is done well.  

 

For instance here is my system impluse pre EQ. 

 

Post FIR convolution (inversion) 

 

Edited June 23 by Grizaudio

[Grizaudio]

Do you see differences in impluse between IIR and FIR? 

[Keith_W]

23 minutes ago, andyr said:

 

Which REW graphs?

 

 

The REW graphs that you want to compare. Roon + FIR overlay + Nanodigi IIR vs. Nanodigi IIR alone. Frequency response, phase, impulse, and step. 

[davewantsmoore]

On 15/6/2023 at 12:54 PM, andyr said:

No Marty - I'm not saying that.  

Then THAT is what you are hearing.  The difference in frequency response.

 

On 15/6/2023 at 12:54 PM, andyr said:

BTW - I suspect the FR graphs of the two options will be the same

So... you're not saying they're the same... but you suspect they will be the same?

 

What you need to do (to get the answer you are seeking) .... is check.

 

 

[davewantsmoore]

On 18/6/2023 at 9:11 PM, Grizaudio said:

I'm confused, why measure at the DAC, how does this show one filter is superior to another, other than the outputs being different? 

You need to measure the system response in room.

Yes.   Put the mic in a position.   Record a sweep of the speaker playing (with #1 and then with #2, without touching/changing anything in between).

[andyr]

1 hour ago, Grizaudio said:

 

@andyr I would be pretty certain you should see improvements in clarity (should be rising) post EQ, waterfall smoothness, improvements (reduction) in bass decay times and specifically time domain charts like impulse - if the correction is done well.  

 

For instance, here is my system impluse pre EQ. 

 

Post FIR convolution (inversion) 

 

 

Sure, I can see the difference between the red 'fuzz' (pre) and purple fuzz (post - better!), Griz.  But what is the light blue line - which seems to be the same in both?

 

In contrast here are my Impulse graphs (I think it's the same graphs):

 

Just my nanoDIGI - IIR filters and REW-EQ:

 

 

 

nanoDIGI providing IIR filters and no EQ; Roon Convolution providing FIR correction and EQ:

 

 

 

That certainly looks a lot better!  

 

Does it explain why I think the spkrs sound clearer (with Roon Convolution 'on')?

 

 

 

[davewantsmoore]

On 18/6/2023 at 9:11 PM, Grizaudio said:

I would look at the important stuff, Impluse response (I would suspect the signal should settle faster using FIR)

... but you can see all of this information in the phase vs frequency chart (and it's a million times easier to read/compare).

 

On 18/6/2023 at 9:11 PM, Grizaudio said:

, Spectrogram (the convolution 'should' provide a smoother response. I.e. less choppy)

Only to the degree that the amplitude vs frequency and phase vs frequency charts are not identical.... so again... you are better to look at the amplitude and phase charts.

 

On 18/6/2023 at 9:11 PM, Grizaudio said:

waterfall/bass decay (I would expect to see improvements in bass decay with either IIR or FIR)

As above.

 

On 18/6/2023 at 9:11 PM, Grizaudio said:

, clarity (You should see a rising improvement in clarity with DSP done right)

As above.

 

On 18/6/2023 at 9:11 PM, Grizaudio said:

so understanding what the FIR is correcting for is important. 

Very much

 

If the FIR filter is used to manipulate the phase... the phase will be different between the two charts.... but should be inaudible, unless very very big (in which case the measurement / data is likely flawed)

If the FIR filter corrects the amplitude to a different target than the IIR .... then this should be quite audible.

 

[davewantsmoore]

1 hour ago, andyr said:

Does it explain why I think the spkrs sound clearer (with Roon Convolution 'on')?

 

Not really.

What it shows you is that the way you have your delay setup for your subwoofer is wrong.   I believe that we have discussed this, at least tangentially, in another thread where I said I suspect that would be the case.

 

By then adding a "phase flattening correction" ..... all of that time error disappears.

 

BUT .... what will matter for how it sounds like, is not what it shows like on this impulse (or phase) chart .... but how they correspond in the amplitude vs frequency.

 

For example.... it is possible to have a crossover between mains and subs which is entirely timed wrong ... that is EQed so that the amplitude is flat.   It should sound "ok" (ie. not as terrible as the phase / impulse / whatever, time domain chart might "look")..... and it is possible to have a crossover between mains and sub, which has the correct delay/phase .... but has a terrible frequency response.  This will sound bad.

 

3 hours ago, Grizaudio said:

Do you see differences in impluse between IIR and FIR? 

But this will only be audible to the degree that this fuzz translates into the amplitude response.

 

 

4 hours ago, andyr said:

the EQ I developed through REW measurements

 

vs

 

a gradually falling FR, according to the specs provided by Con.

 

This is what you are hearing the difference between.

 

If you post a (SPL and phase vs frequency) chart of each (your EQ on, convolution of f..... vs your EQ off, convolution on) .... you will see.    I will bet the difference is enormous.