Something stirs in the Undergrowth
#2611 Re: Something stirs in the Undergrowth
So a million taps, at 192k two channel, Isn't that 100000 * 2 * 192000 multiplications and additions a second? or 48 multiplications and 48 additions for each clock tick of a 8Ghz processor?
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
#2612 Re: Something stirs in the Undergrowth
That's quite a few taps. My Powersoft amps have 2048.vinylnvalves wrote: ↑Sun May 24, 2020 2:08 pm I am limited to 9600 taps at 48khz with the Nadja.. so no FIR capability, except at the top end of the frequency range, where I don’t need it.
Playing with RePhase, 9600 taps should be enough to make some adjustments down to 100Hz-ish. Have a play around.
Nick, good question, and I'm not sure what the answer is. All I can tell you is that the CPU demands of running the media player far exceed the demand of the FIR processing, even at a million taps. I haven't delved much further than that - just a happy end-user.
The processor in this laptop is 2.4GHz nominally, quad core, with hyperthreading. Even so, EQ APO's suggestion of a mere 2.1% CPU use seems very low. I'll do some reading around and see what I can find.
Chris
- IslandPink
- Amstrad Tower of Power
- Posts: 10041
- Joined: Tue May 29, 2007 7:01 pm
- Location: Denbigh, N.Wales
#2613 Re: Something stirs in the Undergrowth
OK, here's a basic question that would have serious implications : when you run one part of a loudspeaker system using FIR, is there a inherent delay associated with it ?
"Once you find out ... the Circumstances ; then you can go out"
#2614 Re: Something stirs in the Undergrowth
Yes.
If I went into my Powersoft amps and used an FIR filter on one channel but not the other, the FIR'd channel would have some delay.
How much delay there is will depend on a few factors.
Another approach would be to use IIR crossovers (analogue or digital, line-level or speaker-level) etc, and then apply FIR filtering to the overall system to unwrap the system's overall phase shift.
That's pretty much what I've done here. Stereo source, stereo amp, pair of passive 2-way speakers. I've applied FIR processing at the source (laptop) which corrects the passive speakers.
Chris
If I went into my Powersoft amps and used an FIR filter on one channel but not the other, the FIR'd channel would have some delay.
How much delay there is will depend on a few factors.
Another approach would be to use IIR crossovers (analogue or digital, line-level or speaker-level) etc, and then apply FIR filtering to the overall system to unwrap the system's overall phase shift.
That's pretty much what I've done here. Stereo source, stereo amp, pair of passive 2-way speakers. I've applied FIR processing at the source (laptop) which corrects the passive speakers.
Chris
-
- Old Hand
- Posts: 1081
- Joined: Sat Jan 26, 2013 12:08 pm
#2615 Re: Something stirs in the Undergrowth
That’s total not per channel . As you know the number of taps for the convolutes is inversely proportional to the frequency. I cannot remember the actual figures off the top of my head but it wasn’t many milliseconds delay before they were all used up . I did a crude xover at 48k but didn’t progress much further. Maybe I need some guidance on how to make more efficient ones.That's quite a few taps. My Powersoft amps have 2048.
Playing with RePhase, 9600 taps should be enough to make some adjustments down to 100Hz-ish. Have a play around.
#2616 Re: Something stirs in the Undergrowth
Ah, okay. I can see how 9600 total would be more limiting, but IIRC I used 2048 taps on my Powersoft amps and managed to flatten down to 100Hz.
I'd recommend trying RePhase to make and simulate FIR filters. It's really handy.
Chris
I'd recommend trying RePhase to make and simulate FIR filters. It's really handy.
Chris
#2617 Re: Something stirs in the Undergrowth
When you say FIR Filtering, I suspect you just mean a delay line, in which case there will be next to no processing involved as its just a circular buffer. That may explain the 2% of CPU used.Another approach would be to use IIR crossovers (analogue or digital, line-level or speaker-level) etc, and then apply FIR filtering to the overall system to unwrap the system's overall phase shift.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
#2618 Re: Something stirs in the Undergrowth
Double-post, but hey.
I recreated what I did for the HiFi speakers, because I didn't save the settings first time around. See before/after graphs (link in the post above) for how this worked when I measured the speakers with processing in place.
The number of taps (and therefore overall delay) was a compromise between what I'd programmed (blue) and what I got (red) - the bit around 500Hz wasn't quite as flat as I'd like, but hey, it's not bad.
In terms of programming, though, it really is as easy as importing a measurement, and hammering away until you've got a good resultant curve with an acceptable delay.
The dips lower down are down to mic position etc resulting in some cancellations. Can't have everything.
Chris
I recreated what I did for the HiFi speakers, because I didn't save the settings first time around. See before/after graphs (link in the post above) for how this worked when I measured the speakers with processing in place.
The number of taps (and therefore overall delay) was a compromise between what I'd programmed (blue) and what I got (red) - the bit around 500Hz wasn't quite as flat as I'd like, but hey, it's not bad.
In terms of programming, though, it really is as easy as importing a measurement, and hammering away until you've got a good resultant curve with an acceptable delay.
The dips lower down are down to mic position etc resulting in some cancellations. Can't have everything.
Chris
#2619 Re: Something stirs in the Undergrowth
It would be interesting to see if you can output the actual filter coefficients. Maybe see how many of the 1million are non zero.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
- IslandPink
- Amstrad Tower of Power
- Posts: 10041
- Joined: Tue May 29, 2007 7:01 pm
- Location: Denbigh, N.Wales
#2620 Re: Something stirs in the Undergrowth
I wonder if the FIR discussion deserves its own thread ?
"Once you find out ... the Circumstances ; then you can go out"
#2621 Re: Something stirs in the Undergrowth
I think that would be useful as it seemingly has much wider applications than just your speaker journey and it would create a resource for information on the topic - that would be useful to people like me who don't have a full grasp of the theory.IslandPink wrote: ↑Mon May 25, 2020 12:19 am I wonder if the FIR discussion deserves its own thread ?
As a user of HQPlayer for my playback I have the option of incorporating 'correction' as one of the playback optons (using a .wav file as described by Chris) and my Soekris R2R DAC, being FPGA-based, has the ability to use custom filters; IIRC, it is possible to use use two stereo Soekris boards to build a DAC/crossover - must look into that.
Last edited by Ray P on Mon May 25, 2020 7:48 am, edited 1 time in total.
Sorry, I couldn't resist!
#2622 Re: Something stirs in the Undergrowth
From the DIY Audio topic on the Soekris DAM1121 DAC;
soekris wrote:It have always be possible to connect multiple slaves to one master, with the digital inputs going to the master. That is mainly for doing digital crossovers using the biquads in the multiple dam1121's, or for balanced or parallel operation.
Sorry, I couldn't resist!
#2623 Re: Something stirs in the Undergrowth
FIR Stuff split as suggested. Wasn't sure where the best place to split was so this is where it was.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
- IslandPink
- Amstrad Tower of Power
- Posts: 10041
- Joined: Tue May 29, 2007 7:01 pm
- Location: Denbigh, N.Wales
#2624 Re: Something stirs in the Undergrowth
Thanks Nick - though I was thinking a little earlier, perhaps from where Chris said "I'm curious, Mark, as to where you're reading those reports", it's nearly all FIR discussion from that point on.
Took me a while to find the new thread .. !
Took me a while to find the new thread .. !
"Once you find out ... the Circumstances ; then you can go out"
#2625 Re: Something stirs in the Undergrowth
Just to dampen my enthusiasm, though on reflection perhaps not surprising given recent posts here, it seems the Soekris FPGA doesn't have the grunt for FIR filters. you can, however, still use the two (or more) board concept for implementing DAC based crossovers.
Sorry, I couldn't resist!