In your specific case you might build the high pass filter using biquads, and then only linearize its phase in the FIR.
What slope and type of filter are you using there?
Hi guys
First I aplogise if these questions have been raised before, I've read several pages on this thread but not all of them! Yet!!
I've tried rePhase but can't make sense of it because I can't be using it correctly!
I exported a full range (EQ'd flat) sweep from REW and imported this to rePhase.
The phase trace was just a mass of zigzag lines. Tried playing with the eq but couln't get it close to a normal line. Crossover is already in place, minidsp and active horns.
Obviously I'm missing the basic's so is there anywhere that can step me through the process?
Appreciate any assistance.
Dave
First I aplogise if these questions have been raised before, I've read several pages on this thread but not all of them! Yet!!
I've tried rePhase but can't make sense of it because I can't be using it correctly!
I exported a full range (EQ'd flat) sweep from REW and imported this to rePhase.
The phase trace was just a mass of zigzag lines. Tried playing with the eq but couln't get it close to a normal line. Crossover is already in place, minidsp and active horns.
Obviously I'm missing the basic's so is there anywhere that can step me through the process?
Appreciate any assistance.
Dave
Hello Dave
If your phase trace is a mess that probably means that you are either measuring from too far at distance and/or with too long a gating (and in that case the magnitude measurement will also be a mess), or that you are not centering the t=0 of the measurement impulse correctly.
Can you show us your measured response?
If your phase trace is a mess that probably means that you are either measuring from too far at distance and/or with too long a gating (and in that case the magnitude measurement will also be a mess), or that you are not centering the t=0 of the measurement impulse correctly.
Can you show us your measured response?
Thanks POS,
I'm away at work this week so can't send anything through. Because the horns are a bit further apart than might be normal in a box speaker I measure further back, in my listening spot. How do i change the gating? I can produce a minimum phase trace in REW which looks normal ish, starts hi in the low bass area , flat for most of it then drops off at the top end. But I understand this isn't true phase.
Is there a real simple process for me to follow so I can get usable results?
If you give me a basic run down I'll re measure when I get back home next week.
Appreciate your help.
Sent from my Moto G (4) using Tapatalk
I'm away at work this week so can't send anything through. Because the horns are a bit further apart than might be normal in a box speaker I measure further back, in my listening spot. How do i change the gating? I can produce a minimum phase trace in REW which looks normal ish, starts hi in the low bass area , flat for most of it then drops off at the top end. But I understand this isn't true phase.
Is there a real simple process for me to follow so I can get usable results?
If you give me a basic run down I'll re measure when I get back home next week.
Appreciate your help.
Sent from my Moto G (4) using Tapatalk
I will first try with a very 'vintage' windows XP platform using VSTHost (freeware/donationware already installed an tested on xp old laptop i have around), Voxengo Pristine space vst as (multichannel) deconvolution engine and my old trusty Aardvark 20/20 (asio). All that in 32bit system (limitation imposed by Pristine Space) and for as cheap as possible...
No synchronisation issue and latency tolerant, it'll be for stereo digital 44.1k/16 bit music only.
Let's dig in the pc dumpster to find a mobo and processor and ram... It'll be fun, don't have done it this kind of thing for maybe 10 years.
I want to reuse Aardvark because it sound VERY nice. This is not a surprise as Aardvark specialized in doing converters and wordclock generator in the end of the 90's.
This was one of the dirty secrets in digital studio of this area: when you couldn't afford a Sony 3348hr, you had a Protools system running horrible 888 converters... you clocked the whole thing by Aardvark clock generator and it suddenly sounded 'good'.
One of the founder of Aardvark just relaunched a brand: Antelope audio.
You have some lineage in the sound of the Aardvark from what i've heard.
Only real issue is the lack of drivers past XP os...
So let's go back in time! And headhache begin: will have to find a solution for install an SSD on win XP first!
I'll try the linux solution latter using the Rme (which is compatible with ALSA).
Last edited:
If the minimum-phase generated curve looks good then the real phase curve should not be a mess (that is unless your crossovers are a mess
).You can play with the "time offset" entry in the measurement tab in rephase and see if it gets better.
You should end up with a phase curve that looks like the REW's minimum-phase one plus a few all-pass around your crossover points.
Also do not hesitate to play with polarity.
I would like to use linear phase 96 LR for the mid HP and sub LP, just for simplicity. (@100Hz)
I've tried extensively with 96LR and 48LR, using Hann widowing. And trying brickwall HP and LP using Albrecht-8, after seeing one of your recent posts.
The linear phase of the HP and LP filters themselves seems to hold up OK with 6144 taps.
It falls apart when I try to add in phase correction for the mid driver.
The mid is a horn loaded design with ports in the horn to aid lower extension. I've attached a pict of it's phase response.
I currently measure the mid's response below from the FIR file in the openDRC, with no attempt there to correct driver phase.
Then I put the driver phase correction in the 2300 taps of an ice ice amp...(using -700usec, inversion, 1st order all pass, and some para EQ down low...
is there a better approach for this driver specific phase correction?
Never toyed with bi-quads. Looked at the spreadsheet from miniDSP this morning. It appears you need to stack biquads to get steep slopes...yes?
I really just want to simplify..........with more taps it seems
Attachments
Last edited:
I will first try with a very 'vintage' windows XP platform using VSTHost (freeware/donationware already installed an tested on xp old laptop i have around), Voxengo Pristine space vst as (multichannel) deconvolution engine and my old trusty Aardvark 20/20 (asio). All that in 32bit system (limitation imposed by Pristine Space) and for as cheap as possible...
wow. you are a brave person. I would never,EVER! attempt this with windows. slow as molasses Graphical interface hell using 2-4 different programs to try to achieve the result.
before you export it with REW, try this.
click on IR WINDOWs
then check the add frequency dependent window
and change the width in cycles to 10
export that and it should now look normal.
Nzlowe
In REw preference measurements tab check mark use acoustical timing reference
It will emit three small chirps and do the t=o offset for you. It works pretty good good enough to measure with ...
Or go to IR tab and manually set you gate before you export the measurement...you can click on the line right at bottom of peak and crosshairs will give you how much delay there is before the IR....
Than export. Don't feel bad I didn't know about that for months and almost gave up on it completely. Once that is sorted out it all will start to make sence
In REw preference measurements tab check mark use acoustical timing reference
It will emit three small chirps and do the t=o offset for you. It works pretty good good enough to measure with ...
Or go to IR tab and manually set you gate before you export the measurement...you can click on the line right at bottom of peak and crosshairs will give you how much delay there is before the IR....
Than export. Don't feel bad I didn't know about that for months and almost gave up on it completely. Once that is sorted out it all will start to make sence
Thx !!
What cascading strategy do you use?
I posted a strategy a few posts up,.... openDRC 6144 taps, then ICE amp 2300.
Just beginning down this route looking for clues
I wish someone would write up a definitive guide to how REW handles timing.
I get the logic behind the starting loopback and acoustic reference timings......but exactly WHAT POINT in frequency /time space are these times relative to ?
And how does Estimated IR delay overlay on top of either of the starting time references?
And why doesn't Estimated IR delay locate 0 timing at impulse peak?
Maybe because it's estimated against a calculated minimum phase curve, which is different than measured curve? (just guessing here..)
In trying to align multiple drivers, REW has driven me near crazy lol
SmaartLive is soooo much easier for this ...albeit so much more $ too...
if you use a physical loopback then the time is relative to the flight time of the reference signal
if you use acoustic loopback then the time is relative to the flight time of the acoustic reference signal which is played by the selected high frequency source
both of these are based on impulse peaks
if you don't use a loopback then it just puts the impulse peak at T=0
that's all there is to it
if you use acoustic loopback then the time is relative to the flight time of the acoustic reference signal which is played by the selected high frequency source
both of these are based on impulse peaks
if you don't use a loopback then it just puts the impulse peak at T=0
that's all there is to it
Yep, I'm good with all that, thx.
Do you maybe know what freq is used for acoustic ref?
I guess my biggest confusion comes from how Estimated IR delay works.
For instance, if no timing reference is used, and T=0 is put at impulse peak, why does Estimated IR delay nearly always still make a shift?
Which can be pretty substantial for measuring passbands in mid and lower frequencies....
IIRC it's a sweep that starts at 5kHz
not sure I understand the question, peak at T=0 is just an arbitrary position.
It's more likely I don't understand enough to ask a decent clear question
I guess I think of it like this....whatever the impulse response is, be it full spectrum or some pass-band, it has an unfiltered peak in time.
With no timing reference, it looks like REW sets T=0 to that peak, right upon initial measurement.....yes?
If so, what is 'Estimate IR delay' relative to that initial T=0 setting ?
about t:0 placement with "estimate IR delay "
JohnM:
"To do this, REW cross-correlates the measurement with a minimum phase version "
IR estimated delay, something wrong - Home Theater Forum and Systems - HomeTheaterShack.com
JohnM:
"To do this, REW cross-correlates the measurement with a minimum phase version "
IR estimated delay, something wrong - Home Theater Forum and Systems - HomeTheaterShack.com
Isn't the polarity reversed in that measurement?...I would like to use linear phase 96 LR for the mid HP and sub LP, just for simplicity. (@100Hz)
I've tried extensively with 96LR and 48LR, using Hann widowing. And trying brickwall HP and LP using Albrecht-8, after seeing one of your recent posts.
The linear phase of the HP and LP filters themselves seems to hold up OK with 6144 taps.
It falls apart when I try to add in phase correction for the mid driver.
The mid is a horn loaded design with ports in the horn to aid lower extension. I've attached a pict of it's phase response.
I currently measure the mid's response below from the FIR file in the openDRC, with no attempt there to correct driver phase.
Then I put the driver phase correction in the 2300 taps of an ice ice amp...(using -700usec, inversion, 1st order all pass, and some para EQ down low...
is there a better approach for this driver specific phase correction?
Never toyed with bi-quads. Looked at the spreadsheet from miniDSP this morning. It appears you need to stack biquads to get steep slopes...yes?
I really just want to simplify..........with more taps it seems
Anyway, if you want a 48dB/oct LR you simply need to stack two 24dB/oct Butt. And in turn each 24dB/oct butt can be obtained by stacking two 2nd order filters, one with Q=054 and the other with Q=1.31.
Similarly a 96dB/oct LR is obtained by stacking two 48dB/oct Butt filters. Each 48dB/oct Butt is obtained by stacking four 2nd orders filters, with Q values of 0.51, 0.60, 0.90, and 2.56.
So you need 4 biquads to emulate a 48dB/oct LR and 8 for a 96dB/oct one.
I'm interested in experimenting with group delay equalization.
When I make a measurement and put it in rePhase, is there any way to EQ the group delay away for low frequencies? I see in Parametric Phase EQ I can play around with it, but they are "peak" filters when I want more of a "shelf" filter. Any way to do that?
Thanks
When I make a measurement and put it in rePhase, is there any way to EQ the group delay away for low frequencies? I see in Parametric Phase EQ I can play around with it, but they are "peak" filters when I want more of a "shelf" filter. Any way to do that?
Thanks