La Scala DSP Settings - Gain & Delay?

[Chris A]

17 minutes ago, Robbie010 said:

This looks like you need to add ~1.5-2 ms delay to the tweeter.  Look at the excess group delay plot to see where the curve is plateauing below the ~6 kHz crossover point.

[Robbie010]

8 minutes ago, Chris A said:

This looks like you need to add ~1.5-2 ms delay to the tweeter.  Look at the excess group delay plot to see where the curve is plateauing below the ~6 kHz crossover point.

 

Its going to take me a while to get to grips with all these graphs and how to read them...... here is the group delay plot:

 

[Robbie010]

Ah Ha! 😃

 

Here is a new graph with a 1.5ms delay added to the tweeter......... it looks to like a further 0.5ms may bring that further in line, which I assume is what I'm aiming for!?

 

[Chris A]

You can magnify into the group delay vertical scale so that the most minor divisions are 0.5 milliseconds (500 microseconds), then you'll be at about the right level to see group delay clearly.  Right now you're zoomed out to the 50,000 foot level.  The control is on the upper left hand of the plot itself when you place the cursor in the plot area, and you will see the scale controls light up when you do. 

 

Yes, the tweeter delay is much closer, but the excess group delay curve --when zoomed in to see the 500 microsecond tick marks--will tell you what you need to know to fine tune the delay settings.  The phase curve in the SPL & Phase plot will also get much better, but probably will still extend out of the boundaries of the vertical scale, even when zoomed all the way out.

 

Don't be afraid to "split hairs" trying to get the delay just right, as I've found it does have an effect (though very small) on sound quality.  I try to get the midrange-bass bin delays good to the nearest ~0.05 ms, and the tweeter to midrange crossover within ~0.005 ms (or less depending on the frequency at crossover--which can be a significant phase shift if not right). 

 

And yes, the ear can hear this stuff over extended listening times...believe it or not.  If I can hear it, then I'm sure others can too (i.e., I don't really have golden ears). 

 

Chris

[Robbie010]

18 minutes ago, Chris A said:

You can magnify into the group delay vertical scale so that the most minor divisions are 0.5 milliseconds (500 microseconds), then you'll be at about the right level to see group delay clearly.  Right now you're zoomed out to the 50,000 foot level.  The control is on the upper left hand of the plot itself when you place the cursor in the plot area, and you will see the scale controls light up when you do. 

 

Yes, the tweeter delay is much closer, but the excess group delay curve --when zoomed in to see the 500 microsecond tick marks--will tell you what you need to know to fine tune the delay settings.  The phase curve in the SPL & Phase plot will also get much better, but probably will still extend out of the boundaries of the vertical scale, even when zoomed all the way out.

 

Don't be afraid to "split hairs" trying to get the delay just right, as I've found it does have an effect (though very small) on sound quality.  I try to get the midrange-bass bin delays good to the nearest ~0.05 ms, and the tweeter to midrange crossover within ~0.005 ms (or less depending on the frequency at crossover--which can be a significant phase shift if not right). 

 

And yes, the ear can hear this stuff over extended listening times...believe it or not.  If I can hear it, then I'm sure others can too (i.e., I don't really have golden ears). 

 

Chris

 

There must be some differences between my version of REW and yours, I can't get my scales to match up to yours at all. What version of REW do you use?

 

 

 

 

[Chris A]

V5.20 Beta 61 presently.  The beta version of REW really isn't what I'd really call beta--its more mature than that, and many of the better features reside in that version, leading the "released" version by a year or more. 

[Robbie010]

2 hours ago, Chris A said:

V5.20 Beta 61 presently.  The beta version of REW really isn't what I'd really call beta--its more mature than that, and many of the better features reside in that version, leading the "released" version by a year or more. 

 

Thanks, Chris. I've downloaded the beta.

 

I've just re-read your basic guide (below) and I think I may have gone wrong at the outset.......

 

I started from scratch this morning and the process I went through was as follows:

 

First, I loaded the basic crossover points for each driver, first order 400Hz low pass (inverted) for the woofer, first order 400Hz high pass for the midrange and first order 4,500Hz high pass for the tweeter. All good.

 

I then set the umik mic at 1m from the speaker, at my listening position ear height for the woofer and took a sweep within a limited frequency range of 20Hz -  2,000Hz and flattened its response using REW. I then did the same thing for the midrange, mic at 1m from the midrange horn, sweep between 200Hz - 8,000Hz and flattened the response and again for the tweeter, mic at 1m from the tweeter, sweep from 1,000Hz - 20,000Hz.

 

However, I struggled with the levels. Prior to making each individual sweep, I tried to set the levels for each driver at 75db using the pink noise test tone in REW and I used the volume control on the plate amplifiers to level match as far as practical. However, whenever I used my digital sound meter at the mic location to confirm the levels, they were all over the place.

 

What is the correct way to complete these first steps? Should I to set the level with all drivers playing and then simply mute two of the drivers and run a sweep on one driver at a time, without any further level changes, or was my method the right way to do it?

 

🥴

[Robbie010]

I had a play around today with the spectograph and delays in the new version of REW and managed to get to here:

 

 

I'm not sure where to go from here, even small changes to the timings seem to make the spectograph worse from here........

 

 

[Chris A]

17 hours ago, Robbie010 said:

What is the correct way to complete these first steps? Should I to set the level with all drivers playing and then simply mute two of the drivers and run a sweep on one driver at a time, without any further level changes, or was my method the right way to do it?

With the Hypex FusionAmps, I would first dial-in the crossover filter slopes and the delays, then start to work on flattening the SPL response.  When you run a measurement sweep with each individual channel (all the PEQs reset to zero_, then you can use REW to find the suggested PEQs to add to flatten the response. You can use the thread that I linked on the first page of this thread for that portion). 

 

With the Hypex FusionAmps, they apparently have no input channel PEQs, so you'll have to run individual sweeps of the woofer, midrange, and tweeter and produce a nominal -3 dB of each driver just at the center crossover frequencies (nominally 400 Hz and 4.9 kHz, or whatever you actually chose). 

 

Once you get the individual drivers/horns flattened and carefully attenuated to about -3 dB at the crossover frequencies, then run a sweep with all three drivers playing, then you can set the individual channel gains to bring all three driver channels to the same SPL level, and set any remaining PEQs to further flatten the response. 

 

With the exception of attenuating room modes, it is possible to achieve ±1.5 to ±2 dB flatness (using REW's psychoacoustic smoothing) from about 70 Hz to 16-18 kHz.  Once that is done, you've got the 99% solution in hand.

 

23 minutes ago, Robbie010 said:

I'm not sure where to go from here, even small changes to the timings seem to make the spectograph worse from here...

 

If you've flattened the SPL response after setting the delays and crossover filters (the SPL response will change), then it's time to listen.  If you want any further fine tuning, you can PM me for my email address to send the measurement files (*.mdat). 

 

I think you'll find that they sound a lot different now--and better--with the time alignment much closer and the SPL response flat on-axis.  You should have a very large soundstage and balanced timbre for your best quality recordings.

 

Chris

[Robbie010]

17 hours ago, Chris A said:

With the Hypex FusionAmps, I would first dial-in the crossover filter slopes and the delays, then start to work on flattening the SPL response.  When you run a measurement sweep with each individual channel (all the PEQs reset to zero_, then you can use REW to find the suggested PEQs to add to flatten the response. You can use the thread that I linked on the first page of this thread for that portion). 

 

With the Hypex FusionAmps, they apparently have no input channel PEQs, so you'll have to run individual sweeps of the woofer, midrange, and tweeter and produce a nominal -3 dB of each driver just at the center crossover frequencies (nominally 400 Hz and 4.9 kHz, or whatever you actually chose). 

 

Once you get the individual drivers/horns flattened and carefully attenuated to about -3 dB at the crossover frequencies, then run a sweep with all three drivers playing, then you can set the individual channel gains to bring all three driver channels to the same SPL level, and set any remaining PEQs to further flatten the response. 

 

With the exception of attenuating room modes, it is possible to achieve ±1.5 to ±2 dB flatness (using REW's psychoacoustic smoothing) from about 70 Hz to 16-18 kHz.  Once that is done, you've got the 99% solution in hand.

 

 

If you've flattened the SPL response after setting the delays and crossover filters (the SPL response will change), then it's time to listen.  If you want any further fine tuning, you can PM me for my email address to send the measurement files (*.mdat). 

 

I think you'll find that they sound a lot different now--and better--with the time alignment much closer and the SPL response flat on-axis.  You should have a very large soundstage and balanced timbre for your best quality recordings.

 

Chris

Thanks, Chris. I’ll drop you a line for the email address, it will be interesting to get an opinion on the filters. 
 

When it come to dealing with the delays, is better to use a full frequency sweep done at 1m or at the LP?

[Chris A]

One metre.  The only time you might take a measurement at the listening position is when measuring well below ~100 Hz (subwoofers mainly) but not above that frequency.  The issue in that case is being in the extreme near field of the loudspeaker due to the very long wavelengths at those low frequencies will affect the measurements.  In general,  I take measurements almost always at one metre.

 

Chris

[Robbie010]

2 hours ago, Chris A said:

One metre.  The only time you might take a measurement at the listening position is when measuring well below ~100 Hz (subwoofers mainly) but not above that frequency.  The issue in that case is being in the extreme near field of the loudspeaker due to the very long wavelengths at those low frequencies will affect the measurements.  In general,  I take measurements almost always at one metre.

 

Chris

 

Hi Chris,

 

See the new spectograph below, this has been done with only the basic crossover filters applied but I'm struggling to smooth out the kink at 3,000Hz and there is the small dogs leg between 200Hz & 400Hz. Is there any insight you can gain from the graph??

 

 

 

[Chris A]

I'd have to look at that .mdat measurement file and your FusionAmp project file to understand what is occurring.  It looks like there is a doubled output of the midrange and tweeter above 3 kHz.  PM me and I'll send you my email address.

 

Chris

[Robbie010]

53 minutes ago, Chris A said:

I'd have to look at that .mdat measurement file and your FusionAmp project file to understand what is occurring.  It looks like there is a doubled output of the midrange and tweeter above 3 kHz.  PM me and I'll send you my email address.

 

Chris

 

PM sent, Chris. Thanks.

 

A little more playing around with the gains and delays has got me to this point:

 

 

 

[Chris A]

That looks better.

 

Chris

[The Dude]

Need to add this to favorites

Sent from my SM-G986U using Tapatalk