TOPIC:

I am wondering if anyone else has compared the sound quality of the Pwr-ice with comparable filters, between the FIR firmware and the regular IIR.

I am using my Pwr-ice 250s to power some homemade hybrid electrostatic floorstanders and I definitely can hear a difference. It seems that the transient response is noticeable worse with the FIR, and overall the sound is somewhat muddied (especially within the FR of the Esl, as it is accurate enough to hear it).

Is this maybe due to the lower sampling freq? Does anyone have any other ideas?

I imagine that this would be inaudible with anything other than ESL's or high end tweeters but it is noticeable. As a result I will be switching back to the IIR firmware.

Hi, are you sure there isn't something else you haven't accounted for? Are the measured magnitude responses the same in both cases?

'Agree with John. I think you probably have some other variable(s) sneaking into your evaluation.
It's absolutely essential to make sure you're making a valid comparison. Even the sampling rate should be the same.

Driver relative polarities can catch you.......among many other things.

Dave.

Well I have simplified it down as much as possible at this point just be sure that I'm not crazy.

At this point I am comparing without any PEQ. All other settings are the same. I am crossing at 300 Hz using 48db Butterworth filters. One difference is that I don't have a perfect FIR filter shape due to the number of taps available, but within reasonable magnitudes it is close enough.

Regarding the sampling rate I cannot control that, unfortunately. The IIR plug in runs at 96 while the IIR plug in runs at 48.

So, just to eliminate as many variables as possible:

I have verified:
polarity
Driver delay
Driver gain

FYI, I my crossover point is low enough that I don't think that the crossover should be an issue. It makes me wonder if it is due to something related to sampling, mostly because there isn't much left to consider.

I'm surprised that you can do a 300 Hz crossover... I will have to look again later.

Still, have you measured the speaker?

What is surprising? That the plug in can do it or the speaker itself?

These Esl panels and transformers can easily go to 150 but it gets a bit too close to the panel resonance. Also the amplifier seems happier with a 300 Hz cutoff.

I have measured and of course the FR is comparable between the two. I can't make anything of the IR, though, due to normal room reflections + dipole reflections. It is quite messy.

Channel 2 is only 300 taps... (Or did something change?) rephase doesn't look all that happy (LR8 high pass) -



What design program are you using?

I am using Rephase. But I used the 2300 taps for the HPF and 300 taps for the LPF. The LPF shape is a bit off but not so much that cannot be fixed with PEQ.

The 2300 taps is intended for the LPF and the 300 for the HPF. That said, a crossover frequency of 300 Hz is too low for the 300 taps available. The stop band just can't achieve enough rejection and the transition and passband can have audible differences from the equivalent IIR.

Here's a FIR attempting a minimum phase 300 Hz LR8 HPF in FIR Designer.


With 300 taps, best to use a crossover frequency of 600 Hz or higher. Here's a FIR attempting a minimum phase 600 Hz LR8.


If your intention, in using FIR, is less phase wrap than a LR8, it's possible to do a mixed IIR/FIR design where the IIR does the LR8 and the FIR undoes the phase wrap, but even with this, 300 taps isn't quite enough to undo all the wrap from about 200 Hz upwards. I just simulated this in FIR Designer and there's enough mag and phase error on the windowed phase-correcting FIR, that magnitude differences will probably still be audible when compared to the IIR LR8.

Kind regards,
Michael

Well your plots don't look familiar to me. I never would have used the FIR filters if the magnitude response looked like that.

I must have done something wrong in Rephase. I will look into it later and see what happened.

Anyways, I agree that with those response curves the FIR filters don't make sense for my application.

And yes, I was trying to improve phase response a bit though it really was just for experimentation. I didn't have any problems with my results using IIR filters.

You are implementing minimum phase filters with FIR?

Hello

if you deal with FIR filter you must have respect the delay of it.

Calaulated:
Delay in msec =(N-1)/(2*FS)
N= Taps
FS= Samplingfrequency in kHz

= (300-1)/(2*48kHz)=3,11msec
=(2300-1)/(2*48kHz9=23,95msec
23,95-3,11=20,84msec

i would recomend to delay the LPF with 20,84msec.



this could fix the unusual sound

Best regards
from a very bad German

Actimel makes a good point. The filters need to be correctly time aligned. His numbers assume linear-phase filters on both channels, with their respective peaks at half their tap lengths.

More generally the peak in a FIR can be anywhere in the impulse response - particularly when doing mixed phase designs that are asymmetric - and so it can be necessary to tweak the channel delays to time align the channels. If you're confident in your FIR filter designs, the delays can be calculated, but it's always good to verify with measurements.

Kind regards,
Michael

OK, now things are starting to make more sense. I didn't even consider the time delay when setting up the FIR...

I know that the 2300 taps is supposedly for LPF and 300 for HPF but I don't understand why I cannot reverse it, especially if I get these results (which aren't absolutely terrible).

I will have to go back and look more closely at the IR for my REW measurements. Honestly I had almost entirely disregarded the IR because it is so messy with the dipole effect and reflections. But I can at least look at the IR peak from the woofer and the first IR peak from the ESL to compare.

For the time being I have gone back to IIR filters and everything sounds just fantastic. I didn't really have a good for reason for switching to FIR except that I have the unending desire to tinker with things.