|
Linear Phase without FIR |
Post Reply 
|
Page <123> |
| Author | ||
LeruSound 
Registered User 
Joined: 27 August 2015 Status: Offline Points: 171 |
 Post Options
 Thanks(0)
 Quote Reply
 Posted: 13 September 2017 at 11:30am |
|
|
Ok, so this I miss, wouldn't i have the same phase response if I use delay compensation in smaart, both from 3 meters and 40m?
|
||
 |
||
|
Jo bg
Young Croc 
Joined: 08 March 2017 Status: Offline Points: 552 |
Post Options
Thanks(0)
Quote Reply
Posted: 13 September 2017 at 12:03pm |
|
|
ciao Leru!
i Am afraid you can't do linear phase filters with xilica xp (i also use the same now, xp4080), you can do it with the XD series processors, but even there with limited taps/capability, you can't use FIR for subwoofer crossover, not enough processing and the delay would be huge anyway. i think that to get linear phase crossovers you have to buy high end xta,lake/lab,xilica xd; linea resaerch has got linear phase filters too, don t know if it uses FIR, Timo Beckman has got some videos on it check it out; anyway, at least 2k € new. or amps with strong dsp... Powersoft i think, lab and linea again... It should be normal to have a phase wrap at every crossover using normal filters, dpn't worry too much about it, a lot of systems don't use FIR. i don't think you can use xilica all pass to get linear phase crossover, but you can use them to correct phase without touching magnitude to flatten individual driver responses to have a better phase allignement with nearby bands. i would look at adressing driver integration at crossover, dispersion and allignement issues if something sounds strange in your midtops. |
||
|
||
|
fatfreddiescat
Young Croc
Joined: 15 October 2010 Location: N.E.Wales Status: Offline Points: 1083 |
Post Options
Thanks(0)
Quote Reply
Posted: 13 September 2017 at 12:08pm |
|
|
If you use delay compensation then that will remove the excess delay and will just show the speakers / filters phase response, looking at what you have done so far with 5 filters the measurement looks very respectable.
For example: With a delay of 1 second you would get 360 degrees @ 1Hz, 720 degrees @ 2Hz, 3600 degrees @ 10Hz etc. 360 degrees for each full cycle, it take 1 second for 1Hz to go through a full cycle. LR24 will give you a 360 degree phase shift at the xo point.
|
||
|
||
|
LeruSound
Registered User
Joined: 27 August 2015 Status: Offline Points: 171 |
Post Options
Thanks(0)
Quote Reply
Posted: 13 September 2017 at 6:08pm |
|
|
Thank you guys. overall performance are good... Everything is well aligned, passive crossover have been made on misurated prediction and alinged perfectly. Just some guy i met noticed me that i have lot of wraps, and told me it can cause some blurry sound, losing focus and detailed dynamicVSfreq. It makes alsi sense to me... Is not that good to have 4th order filters, but i have 2 xover point in very small band..
Dont you think theres any other way to achieve linear phase without FIR? As fatfreddie said that graph should only represent phase shift caused by filters... i'll try... |
||
|
||
|
corell
Young Croc
Joined: 19 August 2013 Location: Berlin Status: Offline Points: 1161 |
Post Options
Thanks(0)
Quote Reply
Posted: 13 September 2017 at 7:56pm |
|
True. Group Delay is the derivation of Phase Slope. So a linear Phase means a constant GD. You then just need a "normal" delay (as opposed to a frequency depending delay, what a FIR is able to) to get all frequencies arriving at the same time! |
||
|
||
|
LeruSound
Registered User
Joined: 27 August 2015 Status: Offline Points: 171 |
Post Options
Thanks(0)
Quote Reply
Posted: 13 September 2017 at 9:07pm |
|
|
So not sure, I know phase shift and group delay are directly related, but I wanted to obtain a fir response using analog filter with allpass.
Is it possible? Further more, at this point, tell me what you see.. does not look like there's any bad gr.dly causing bad shifting in phase resp.. isnit? |
||
|
||
|
corell
Young Croc
Joined: 19 August 2013 Location: Berlin Status: Offline Points: 1161 |
Post Options
Thanks(0)
Quote Reply
Posted: 13 September 2017 at 11:55pm |
|
|
The phase slope you posted is very nice imo. there is a 360° shift for the two crossovers (probably 24db LR ?) and one 360° for the acoustical highpass the enclosure introduces.
And no, i dont think it is possible to make a unsmooth phase smooth (linear or constant, depending on your pre-delay) without using FIR. An unsmooth phase means you have "spikes" in your GD (group delay, simply put, is the time you need to transform an electrical input to an acoustical output through your system), so some frequencies arrive later than others. IIR filters are only a digital implementation of analog filters, so you would need to add some kind of objects to your signal path with some frequency dependend time behaviour but no impact on the response. Maybe a sufficient amount of all pass flters could do that? im not sure tbh. |
||
|
||
|
toastyghost
The 10,000 Points Club 
Joined: 09 January 2007 Location: Manchester Status: Offline Points: 10919 |
Post Options
Thanks(0)
Quote Reply
Posted: 14 September 2017 at 12:15am |
|
|
Even if you do achieve linear phase response in most of the passband, you may cause more problems such as pre-ringing, affect dispersion in one or more directions, or many others. Just whacking an FIR filter over the response is not enough and will likely make issues worse - if you look into it, you'll see that even the biggest and best sounding manufacturers do not chase fully linear phase response even in their flagship products – and if they do, they spend a lot of time, money and their best brains to even attempt it.
The other question is, can you even tell the difference; lots of research says no. |
||
|
||
|
Timo Beckman
Registered User
Joined: 09 August 2011 Location: Netherlands Status: Offline Points: 197 |
Post Options
Thanks(0)
Quote Reply
Posted: 14 September 2017 at 12:19am |
|
|
Hi "LeruSound".
Before starting to do a "linear phase" system (which is possible if the Xilica you use has 2nd order allpass filters) check the filter orders you're using. Past 2nd order high-pass filters from 1K going up is no fun and you need a lot of AP2 filters. I saw some screen shots from the xilica controller and the phase plot they show in the controller is inverted compared to what a Allpass filter does (i did sent a email to them in 2011/2012 regarding this subject but seems that it's not their thing to reply or to change the plot in their controller which is wrong) I did a "couple" of posts on how to change the phase respons of a system. In short you have to add delay to every processor output. After that synchronize your FFT analyzer. After synchronizing the analyzer decrease delay on the processor output(s) and by doing so the phase trace will go up instead of going down. At that point you can use allpass filters to get the phase trace back to the 0º line...... It's a pain in the but performing this procedure with higher high-pass filters the 2nd order. Here are the links to the blog posts on the subject: And a "golden oldie" from the first time i tried this trick: The processors i use that have linear phase stuff (Linea and Lake) use a algoritme (i think) where they've integrated the allpass trick shown in the blogposts but you do not have to do the work yourself. If i synchronize on separate outputs i see about the same stuff i see when doing the allpass trick but with less latency and a better controlled phase respons in the outputs frequency range. Also these processors can go to higher filter order (4rd 8th or even more in the lake) without the accompanying phase delay. Enjoy Edited by Timo Beckman - 14 September 2017 at 12:20am |
||
|
The sound will be as good as the band play's
|
||
|
||
|
gen0me
Young Croc
Joined: 20 February 2016 Location: UK Status: Offline Points: 999 |
Post Options
Thanks(0)
Quote Reply
Posted: 14 September 2017 at 12:26am |
|
No. Formula is in the 4th post. Constant slope is not beautifull phase. Look at hornresp record where group delay is constant, phase is getting more shifted with frequency rise. Better show us group delay or waterfall 
|
||
|
||
|
corell
Young Croc
Joined: 19 August 2013 Location: Berlin Status: Offline Points: 1161 |
Post Options
Thanks(0)
Quote Reply
Posted: 14 September 2017 at 3:43am |
|
|
I think you are mistaken here. GD is the derivation of phase, the rest is math.
What should be bad about a constant GD >0 ? It just means the signal needs time to be transformed, but the same for every frequency. |
||
|
||
|
Teunos
Old Croc 
Joined: 23 November 2008 Location: The Netherlands Status: Offline Points: 1799 |
Post Options
Thanks(0)
Quote Reply
Posted: 14 September 2017 at 7:45am |
|
I'm surprised this is a point everybody seems to miss. BUT!!! This graph shown is NOT linear phase. The plot axis is logarithmic! Also, if we are at it anyways, group delay is negative derivative of phase.
|
||
|
Best regards,
Teun. |
||
|
||
|
Post Reply
|
Page <123> |
| Tweet |
| Forum Jump | Forum Permissions You cannot post new topics in this forum You cannot reply to topics in this forum You cannot delete your posts in this forum You cannot edit your posts in this forum You cannot create polls in this forum You cannot vote in polls in this forum |
Topic Options
snowflake wrote: