A how to for a PC XO. - Page 52 - diyAudio
Go Back   Home > Forums > Loudspeakers > Multi-Way

Multi-Way Conventional loudspeakers with crossovers

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 6th August 2006, 04:34 PM   #511
diyAudio Member
 
Join Date: Jan 2004
Location: London
Quote:
Originally posted by dwk123

I use SoundEasy to measure 'through' Allocator, and it mostly works ok. The big issue I'm having I believe is due to the Emu drivers and not anything to do with Allocator, which is that latency is wildly unpredictable from measurement to measurement. Makes it really tough to look at time alignment. I'm planning to swithc measurement over to an M-Audio card, but then Allocator won't be in the loop. However, it should at least allow me to get solid reference FRD's that are normalized.

Please have a look at the nice little utility that I wrote (Simple Automated IR Measuring Tool) to capture impulse responses available from the room correction wiki:
http://www.duffroomcorrection.com/wi...Measuring_Tool

It should run on Windows (ASIO/Direct Sound), Linux (Alsa, OSS or Jack) and OS X

Sure it's command line, but it's supposed to be part of a larger tool which can do measurement and analysis.


Many soundcards have trouble doing sample accurate sync of playback and recording. It's a driver level issue, but compounded by many audio API's not offering this as a solid option, eg it's only possible under ASIO on windows, not direct sound.

The app above will allow sample accurate recordings using ASIO, Alsa and possibly Jack and others. OSS would also be possible with a simple patch but I haven't bothered to add it.

However, to work around this you can use a second "loop back" channel in the software - by recording the sweep on a second channel (which goes straight out and back in again) we can then normalise the signal for freq response and also for delays in starting the recording and playback.

If you are pretty technical then you will have no problem adopting the basic code to measure all kinds of cool things such as static tones, etc.


With regards to Mic's. The basic ECM8000 is extremely good, it's just not calibrated. Add some random generic calibration such as you will find on the internet to make things a bit better, or buy one ready calibrated for a touch more money (lots of firms make mics based on the Panasonic capsule, but ready calibrated).

Good luck all

Ed W
  Reply With Quote
Old 6th August 2006, 05:25 PM   #512
diyAudio Member
 
Join Date: Feb 2002
Quote:
Originally posted by Thunau


If you wanted to just confirm the combined frequency response you could open the preset saved in Allocator in Allocator Light (the preset files are compatible). This way your measurements would not have the extra delay in them.
If you wanted to confirm the phase as well, you have to use the Full allocator and deal with the latency issues.
The problem really isn't the amount of latency (although that does require a bit of scrolling around the window playing 'find the impulse'), it's the variability. One test may put the main inpulse at 120ms, the next at 60 or 150. I'm pretty sure this is due to the Emu WDM driver, but since I can't run SE via asio, I don't actually know that for sure.
Either way, though, the 'combined system' stuff is OK, the only problem is in trying to get aligned impulses for each driver so that offset can be determined.


Quote:

I think that all you need is a mic preamp in the audio interface. The M-audio 410 has it, the Presonus Firebox has it, the Emu 1820 has it.

Are you running into problems acessing yours from two applications simultaniously?
Given the request for AES/EBU, I was guessing that the original poster is using a Lynx AES-16 card, since most other AES/EBU cardshave some level of analog section to it.
  Reply With Quote
Old 6th August 2006, 05:35 PM   #513
diyAudio Member
 
Join Date: Feb 2002
Quote:
Originally posted by ewildgoose



Please have a look at the nice little utility that I wrote (Simple Automated IR Measuring Tool) to capture impulse responses available from the room correction wiki:
http://www.duffroomcorrection.com/wi...Measuring_Tool
I've seen that, and have been meaning to check it out for DRC impulse generation. My speakers and xovers are still evolving, though, so I haven't gotten to that stage yet.
Plus, my current PC can't handle xover +DRC, but I'm almost done building out a much quieter and faster box which will handle both.


Quote:

However, to work around this you can use a second "loop back" channel in the software - by recording the sweep on a second channel (which goes straight out and back in again) we can then normalise the signal for freq response and also for delays in starting the recording and playback.
I've stopped using a reference channel for now, since I was getting completely bogus results when doing so - the reference channel was far from flat, which screwed up the results. I didn't spend much time investigating, though. Since the results I'm getting without the reference channel tend to aggree with other folks measurements of the same drivers, running without one seems to be working OK.
You're right that this makes determining the offset far easier, though.
  Reply With Quote
Old 6th August 2006, 05:43 PM   #514
Thunau is offline Thunau  United States
diyAudio Member
 
Thunau's Avatar
 
Join Date: Mar 2001
Location: PA USA
Quote:
Originally posted by dwk123




I've stopped using a reference channel for now, since I was getting completely bogus results when doing so - the reference channel was far from flat, which screwed up the results. I didn't spend much time investigating, though. Since the results I'm getting without the reference channel tend to aggree with other folks measurements of the same drivers, running without one seems to be working OK.
You're right that this makes determining the offset far easier, though.
I wonder if wiring the loopback internally for example in the Console would solve the problem? You would not compensate for the card's frequency response irregularities, but the impulse and delay should be right on.
All you would have to do is perform a measurement of latency of the DA/AD section first, then insert a delay equal to that latency in the path of the loopback.
__________________
"Most people just say what they know, the wise ones know just what to say."
  Reply With Quote
Old 6th August 2006, 05:56 PM   #515
Thunau is offline Thunau  United States
diyAudio Member
 
Thunau's Avatar
 
Join Date: Mar 2001
Location: PA USA
Quote:
Originally posted by ewildgoose



Please have a look at the nice little utility that I wrote (Simple Automated IR Measuring Tool) to capture impulse responses available from the room correction wiki:
http://www.duffroomcorrection.com/wi...Measuring_Tool

It should run on Windows (ASIO/Direct Sound), Linux (Alsa, OSS or Jack) and OS X

Sure it's command line, but it's supposed to be part of a larger tool which can do measurement and analysis.



Ed W
I took a look at it and even managed to run it once in the cmd window.
My question is, how hard would it be to make a GUI in Visual Basic that lets one take advantage of the options without typing long commands?
Also, will MLS measurement be possible using your program?
__________________
"Most people just say what they know, the wise ones know just what to say."
  Reply With Quote
Old 7th August 2006, 09:05 AM   #516
diyAudio Member
 
Join Date: Jan 2004
Location: London
Quote:
Originally posted by Thunau


I took a look at it and even managed to run it once in the cmd window.
My question is, how hard would it be to make a GUI in Visual Basic that lets one take advantage of the options without typing long commands?
Also, will MLS measurement be possible using your program?
Writing a GUI should be a process of about 2 mins... There is a fixed command line and you usually don't even change any params except filename..

To be honest I don't bother because I just cut and paste the sample command out of the usage help and that's about all you need to run it! It even guesses the channel and card you are likely to want to use, so you can leave most params unentered. Basically I really don't feel the need for a gui myself...

Also as a cmd line program I tend to use it as part of a longer script that I just click on and it measures everything, calculates the filters and then drops them into my convolver. Probably it's *harder* to use a GUI than this!

MLS is not likely to be something that I implement. Log Sweeps are so much better for many reasons and have only one disadvantage which is that it's reasonably easy to melt your tweeter if you run them very loud. The SNR is vastly better and they are unaffected by speaker distortion or overload, all in all I see little reason for an MLS measurement (convince me otherwise?)

One of the biggest reasons for log sweep is the immunity to many forms of distortion... Overload the speaker into distortion and it's no problem because the distortion component drops out of the IR and can be seen (and easily measured) as seperate spikes just in front of the main IR... Nice feature!

Ed W
  Reply With Quote
Old 8th August 2006, 02:45 AM   #517
Daveis is offline Daveis  United States
diyAudio Member
 
Join Date: Dec 2003
Location: Des Moines, IA
I've been using Thunau Allocator Lite this evening. Latencies of 6 msec seem stable on my PC using an RME card and Athlon64 3500. 3msec works but with occasional glitches.

Can anyone tell me how I can figure out the relative delay between my woofer/mid/tweeter?

I've encountered the variable latency problem even when using a reference channel.

As far as MLS vs. log sweep... isn't MLS purely an attempt to remove the room's response from the measurement. When you do long sine sweeps you are recording alot of room echo even if the mic is up close. At least with an MLS response I've been able to "see" the mismatch in delay between mid/tweeter/woofer.

I never was able to get Ed's rec_impASIO to work. And currently, I'm less interested in summed frequency response as I am impulse response. From eyeballing an MLS graph I can see at least 3 msec difference between woofer and mid.

Anyone have experience with Praxis, Sample Champion, and Sound Easy? I've heard them all mentioned for speaker measurement. I used a trial of Sample Champion for a while.
  Reply With Quote
Old 8th August 2006, 03:56 AM   #518
Daveis is offline Daveis  United States
diyAudio Member
 
Join Date: Dec 2003
Location: Des Moines, IA
Ok, after doing some research on MLS vs. log sweep, it appears that log sweeps can indeed be used for both distortion measurements and impulse response measurements. And that there are numerous benefits to log sweeps, including better S/N, etc.

I even read something about being able to eliminate room response, but it has something to do with using very long log sweeps... longer than the room decay.

A link

http://support.supermegaultragroovy...._Sweeps_vs_MLS
  Reply With Quote
Old 8th August 2006, 10:01 AM   #519
diyAudio Member
 
Join Date: Jan 2004
Location: London
Please read the first paper referenced in the page at the link below. It's extremely thorough and there is also an even longer version of the same paper floating around with much more detail still

However, the basic point is that MLS and low swept sine produce the *same* output result. Both of them are intended to capture the same information, and yes you can see both room reflections and direct sound on both - no it does not really matter about the length of the sweep as long as we are talking about real rooms and sensible length sweeps (sure measuring some big Cathedral may take some slightly different params!). More than 2 seconds or so should be plenty and in general I would suggest a 45 second sweep

Basically both MLS and sweeps are mathematical tricks to calculate the room IR without using significant CPU power. There is a trick with the sweep where you basically do a trick by inverting the original signal and convolve it with the recorded signal and the IR drops out. However, in theory at least you could play the radio out through your speakers, record the sound and from this recording you could figure out the IR - this requires doing some FFTs though and hence is less common (although not that complicated to code up if anyone is actually interested)

Sweeps also have the nice property that any *harmonic* distortion gets automatically removed from the impulse response, ie if you crank up your speakers to any appreciable volume then they will distort quite a bit - all this distortion is *included* in the MLS results, but *excluded* from the sweep results - cool huh!

However, there are still some types of non-linear distortion which will mess up even a log sweep measurement - typically the effects of a cheap soundcard which is resampling everything to 48Khz fall into this bucket.


If anyone is having trouble with rec_imp then please drop me an email offlist and I will endevour to fix the problem. Windows has really bad audio apis and so we need to just get problem reports for each audio card and work around the windows bugs...

If you tried a much older version of rec_imp then it was known to have plenty of windows issues, but the latest version should work reasonably well for many people. Also try installing Asio4all which can often give you more stable drivers and often allows sample accurate record and playback (ie the delay is constant and you can more easily measure driver distance)

Using a loopback channel *will* allow you to correct for soundcards which don't have sample accurate record and playback. The only situation which would break this system is where you had a soundcard which had different clocks for left and right audio channels, but such a card would never play back decent stereo signals either so I can't imagine that such a card exists. The basic idea is that you do a sweep through your audio system and also a sweep straight out and back into the soundcard. Then the gap you are looking for is the gap between the loopback impulse and the speaker measured impulse - this gap will be constant even though the absolute positions of each impulse varies


Please all consider buying something like ETF or Sample Champion if you enjoy fiddling with this stuff. I used to muck around with SPL meters for hours and yet the information you can get from a single sweep and 10 mins on ETF is *VASTLY* better than spending hours of your life with an SPL meter trying to guess what is going on! Honestly, there is a bit of learning to figure out how to interpret the graphs, but after that you will have a tool which just gives you all the crucial info about your system in one easy measurement.

Good luck
  Reply With Quote
Old 21st August 2006, 10:58 PM   #520
diyAudio Member
 
ShinOBIWAN's Avatar
 
Join Date: Feb 2004
Location: UK
Hi Jan

I've finally gotten around to checking out the Allocator/Arbitrator.

Things are looking pretty good so far.

I'll post a thorough writeup in after I've had a lot more exposure.

A couple of things missing that immediately struck me as being obviously needed:

- Ability to setup seperate XO schemes for the left and right channels. This is important however trivial it may sound. Imaging in particular really takes on a pin sharp quality once you have very closely matched pairs of speakers. I should imagine this feature will be fairly easy to implement.

- Dynamic skins; The program chops information off the display and those controls are inaccessible with any less than 1024x768. This isn't much good for me as my main display is a widescreen projector with a resolution of 1280x720 and my secondary display is a small 7.5" TFT with a resolution of 800x600. This may not be a problem for the masses but its something thats annoying for us that have less tradition displays. I always really like the waves stuff because they were compact but had all the options laid out nicely.

The following isn't essential but would greatly increase the appeal of the program:

- FIR filters implemented. If this is done I'd like to see the ability to customise the coefficients and window functions, as well as the more generic types. The design package LEAP is excellent in this regard but it doesn't have a crossover emulator to allow you to 'listen' to these. Again this is a big ask and I'd totally understand if you turn around and say 'ain't gonna happen buddy'. Didn't you mention a possible FIR version of the software in our emails?

- DRC integrated into the whole package so as to prevent cluttered plugins syndrome (perhaps you could get in touch with Dennis and see about using his freely available source and packaging it with your own program, all you'd then need to do is implement a convolver). This can be achieved through using Console with the addition of Voxengo Pristine Space or a similar convolver such as Waves IR1.

- The ability to measure drivers and generate .frd profiles within the program without relying on 3rd party apps to do the job, Without the .frd files, I've found that Allocator/Arbitrator is very limited in scope otherwise. Its full potential is only realised when your using actual driver data measured from the final enclosures. I know this is NOT a trivial thing to implement and perhaps relying on 3rd party apps is essential to maintain a reasonable cost for the app but I'd strongly suggest writing a more thorough manual just for this subject as you'd certainly be aiding ease of use for those newer to this. For those who've been messing with this stuff for a while now and have programs like ETF and SC, this isn't a problem.


Finally I don't mean to sound negative. All of these were just things that I noticed within the first 20 minutes of playing around with the program. Overall I think it looks promising. I'll be comparing it to the DEQX in terms of sound quality alone. I've been extremely pleased with the DEQX over the last 6 months so it will be interesting to see if I can approach and hopefully exceed what I've done with that.

I'd also love to compare it the Waves/Voxengo/DRC setup I had going very early on this year - I still swear it was better sounding than the DEQX but alas. I some 'issues' seemed to creep into the implementation. These still seem to be at large so I doubt any meaningful comparison can be drawn.

I'll get back to you with more thoughts as I get into testing more deeply.

Cheers
Ant
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



New To Site? Need Help?

All times are GMT. The time now is 09:36 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2