Hi there!
I have played a lot with simulation software - a quite simple to start one - BoxSim from UweG.
The problem is that you have to start with the measurement files of the Visaton drive units - the new one with the 19 measurements in 5 degree steps.
I have tried for comparison reasons also VituixCAD 2 - it is much more effort to get a first result with this software.
With BoxSim it is much easier to learn how an active crossover works together with the frequency behavior of the driver itself.
Here a small example how it looks like - sorry for the text in German, i think you can turn the language to english:
I have played a lot with simulation software - a quite simple to start one - BoxSim from UweG.
The problem is that you have to start with the measurement files of the Visaton drive units - the new one with the 19 measurements in 5 degree steps.
I have tried for comparison reasons also VituixCAD 2 - it is much more effort to get a first result with this software.
With BoxSim it is much easier to learn how an active crossover works together with the frequency behavior of the driver itself.
Here a small example how it looks like - sorry for the text in German, i think you can turn the language to english:
You've got almost 6 ms of delay on the lower frequency driver...
I recommend cranking in about 5.7 ms of channel delay on the higher frequency channel, i.e., the AMT-1.
That's quite a lot of--almost 2m of acoustic offset of the two drivers. What order of crossover filters are you using?
In an MEH, the woofers should lead the HF driver by 90 degrees if no crossover filter is used, not the reverse, and not by that amount..
Chris
Last edited:
This is actually funny. When I try dual channel measurement with no crossover (maybe I forgot to turn off the Mid highpass at 300 Hz), I measure the AMT around .5 ms "closer" to the mic than the mids. But it seems that the Behringer soundcard is not really good for this, I get multiples of 1 kHz peaks in distortion and some other strange stuff. It is slightly better when using the ASIO drivers instead of standard windows, but still strange.
The measurement above was made by UMIK, the filters were all LR24, mids - 300 Hz HP, 1000 Hz LP, HF 1000 Hz HP - the 1 kHz crossover point can be actually a bit lower, but I think it is exactly at 1 kHz in this measurement. Now I see there is also the midbass unit included, which is a 12" bass reflex with 300 Hz LP and 120 (or 140?) Hz HP. That one is not in the horn. The horn contains only the AMT and mids (2x 3FE22).
I think I need to start from the top, just the crossover between AMT and mids. I will try to do some better measurements tonight (I failed yesterday in Win, need to do in Linux, but there I can only do with UMIK). AMT only, Mids only, then both together without any filters or delay.
The measurement above was made by UMIK, the filters were all LR24, mids - 300 Hz HP, 1000 Hz LP, HF 1000 Hz HP - the 1 kHz crossover point can be actually a bit lower, but I think it is exactly at 1 kHz in this measurement. Now I see there is also the midbass unit included, which is a 12" bass reflex with 300 Hz LP and 120 (or 140?) Hz HP. That one is not in the horn. The horn contains only the AMT and mids (2x 3FE22).
I think I need to start from the top, just the crossover between AMT and mids. I will try to do some better measurements tonight (I failed yesterday in Win, need to do in Linux, but there I can only do with UMIK). AMT only, Mids only, then both together without any filters or delay.
It's not a specious argument at all.....
Because being in phase thru xover (along with flat magnitude) equals complementary acoustic summation.
Can smooth acoustic summation throughout xover not be symmetrical? I can't see how.
I think one thing folks need to see past when picturing acoustic summation, is the picture of symmetry like with electrical LR xovers.
Unlike LR xovers, acoustic symmetry doesn't need to happen at any integer order or number of orders. It could occur at 8.5 dB/oct for example, and have a varying order away from xover freq.
Just like driver rolloffs don't fall on integer orders, and can vary in slope/order as they move away from their passband.
Bottom line goal is that smooth mag and phase summation occurs acoustically post electrical filtration.
The electrical filter that makes that happen can be anything really....whatever delivers the desired shape of the acoustical summation symmetrically.
The big thing to remember imo, is that once a desired acoustic xover order (or shape) and frequency are chosen, there is one unique net electric filter that will provide it. (The net filter being the transfer function of the xover or high/low pass, EQ's, all-pass filters, etc, whatever is in play through the xover region)
All techniques, all methods of getting there, have to ultimately reach that one unique net filter, or bogus.
Again, the unique filter can be reached through nothing but PEQs and shelving and all-pass and whatever, or have named xovers thrown in, whatever....it doesn't matter as long as the net electrical filter provides the desired symmetric acoustical summation.
Which equals flat mag and in-phase throughout xover.
For me, it's just a matter of finding the easiest method to get there.
Because being in phase thru xover (along with flat magnitude) equals complementary acoustic summation.
Can smooth acoustic summation throughout xover not be symmetrical? I can't see how.
I think one thing folks need to see past when picturing acoustic summation, is the picture of symmetry like with electrical LR xovers.
Unlike LR xovers, acoustic symmetry doesn't need to happen at any integer order or number of orders. It could occur at 8.5 dB/oct for example, and have a varying order away from xover freq.
Just like driver rolloffs don't fall on integer orders, and can vary in slope/order as they move away from their passband.
Bottom line goal is that smooth mag and phase summation occurs acoustically post electrical filtration.
The electrical filter that makes that happen can be anything really....whatever delivers the desired shape of the acoustical summation symmetrically.
The big thing to remember imo, is that once a desired acoustic xover order (or shape) and frequency are chosen, there is one unique net electric filter that will provide it. (The net filter being the transfer function of the xover or high/low pass, EQ's, all-pass filters, etc, whatever is in play through the xover region)
All techniques, all methods of getting there, have to ultimately reach that one unique net filter, or bogus.
Again, the unique filter can be reached through nothing but PEQs and shelving and all-pass and whatever, or have named xovers thrown in, whatever....it doesn't matter as long as the net electrical filter provides the desired symmetric acoustical summation.
Which equals flat mag and in-phase throughout xover.
For me, it's just a matter of finding the easiest method to get there.
Last edited:
Allen the symmetry might be needed as well as the importance of phase for FIR xovers.
For the xover i use those are prerequisite.
I think there is misunderstanding between us because we don't explicitely tell when we use them. Eg: i almost never linearize outside band more than half an octave in practice as i use 48db/octave linear phase filters... but for IIR 1 octave is a minimum... sometimes i make shortcuts and tell what i use without warning it could not work with IIR.
Pelanj you very probably have windows messing things. Don't ever let windows take care of audio. Make a search about optimisation of win for audio ( studio dedicated forums as well as magazine like Sound On Sound ight have article or tutorials).
For the xover i use those are prerequisite.
I think there is misunderstanding between us because we don't explicitely tell when we use them. Eg: i almost never linearize outside band more than half an octave in practice as i use 48db/octave linear phase filters... but for IIR 1 octave is a minimum... sometimes i make shortcuts and tell what i use without warning it could not work with IIR.
Pelanj you very probably have windows messing things. Don't ever let windows take care of audio. Make a search about optimisation of win for audio ( studio dedicated forums as well as magazine like Sound On Sound ight have article or tutorials).
Last edited:
If you have access to a USB DAC (or some other preamp input means than a soundcard), I'd try another means to inject the REW driving signal. You may also have issues with your DSP crossover. I've seen issues similar to this sort of behavior before with others that I've helped, and it was the most pernicious difficulty that I've experienced in helping others with the dialing in process.
When I started out using REW over 14 years ago, I had terrible issues with the soundcard I used to drive the preamp. It wasn't until much later (a few years) when I had more time to devote to the process that I replaced the older Behringer microphone/mixer (phantom power source and ADC) and laptop soundcard with a UMIK-1 (having self-calibrating output) and a HDMI output from the laptop to a pre/pro that the measurements began to be usable and repeatable. After I replaced the older EV Dx38 crossover with a Yamaha SP2060 that the measurements, and therefore control of the acoustic output, finally became predictable. But that was its own sort of a "horror story" that I had to deal with. I always hope that others don't ever have to deal with any of those kind of problems.
Chris
If you want something stable use what we use in professional application. This is the first parameter of judgement from this kind of gear.
Iow, Asio soundcard if on win.
Iow, Asio soundcard if on win.
That sounds like a good place to start. The VHF sets the timing refence, and then work down the ladder.
I'd get your dual channel with loopback running. Measure the AMT's TOF, and set that as the Delay for the measurement such that a measurement shows zero delay when taken.
Then measure the mids raw, and the Delay shown will be very close to the distance between acoustic centers.
I agree with Chris regarding your posted measurement, and also his comments about not needing a 4th order xover with a MEH.
If you use either a 1st order or any LR between the AMT and mids, the timing you found above process, AMT vs mids, will be the Delay you need to put in your processor. No need to account for any xover induced delay then. If you use something non complementary, then additional delay finding may be needed,
fwiw, I use a Behringer UMC404 interchangeably with a RME Baby face Pro....both give me the same results consistently. Oh, and with Asio drivers on WIn10pro.
ECM8000's mics are dang close to my Isemcon too...
Doesn't take much for speaker work (as opposed to digging in to electronics, which i only sometimes do)
ECM8000's mics are dang close to my Isemcon too...
Doesn't take much for speaker work (as opposed to digging in to electronics, which i only sometimes do)
I have a great FW card, but my new laptop does not have any option to get a FW port. I used it for real time sample/trigger playing from DAW and I had a highly optimized WXP setup for that, using ASIO drivers. I tried a borrowed cheapo UMC22, but I think it is unusable. The UMC1820 I have should work much better I hope. I found out that my ECM8000 most probably does not work after being unused for a year and a half (need to try once more - maybe it was the UMC22 that was not fully OK), so I think I will need to wait for the replacement. I guess cardioid overhead mics cannot be used for anything...since I do not have another measurement mic at hand at the moment.
I have some things to try this afternoon, so I hope I will at least move forward a bit.
I really appreciate all your feedback here!
I have some things to try this afternoon, so I hope I will at least move forward a bit.
I really appreciate all your feedback here!
Hey pelanj, i have a UMC1820 too....it's better with REW than my UMC404 because it has way more line output, and lets mic SPL level calibration have more maximum headroom. I'd be surprised if your ECM is whacked...i bought several of them years ago, don't use them often, but they still check out.
You clearly know your way around... and will be sailing once you get the bugs killed.....Good luck!
You clearly know your way around... and will be sailing once you get the bugs killed.....Good luck!
Instead of linearising the natural rolloff it is possible to "bend" the natural rolloff of drivers to achieve a desired traget function by the use of biquadratic filter (of which the Linikwitz transform is a variant) or other types of equalisers. This target function is a part of the acoustic crossover function.
Regards
Charles
Last edited:
Totally agree Charles. And implemented it.
This way you don't have ( a dedicated) high pass processing in the final complementing filter, it's done 'assisted natural' behavior of box+driver. For the low pass however a complementary electrical filter of same kind is used in what i've done ( for a target 'named' config). Sometimes i used a filter in top of that hp for eg a 24db LR ( another 2pole 0,707 butterworth on top of the LT with same spec).
For reference target so far i used some profile from Rephase and before i recorded the output of my processor's curve ( all digital loop, no conversion so as clean as you could get. Both works well but you need harware with comprehensive digital in/out routing).
I didn't wanted to complexify the subject even more in case Pigmy decided to be part of the discussion again, and why i took caution in what i've said: there is multiple possible approach to this, some way clever than 'my brute force' approach.
Pelanji, what kind of expension ports do you have on your laptop? Adaptec had some cards availlable to expand laptop's capability. I use some for my older laptop for scsi and FW ( old studio gear in FW and circa 90's samplers) without glitchs for years.
Maybe they have new version of them for recent laptop?
This way you don't have ( a dedicated) high pass processing in the final complementing filter, it's done 'assisted natural' behavior of box+driver. For the low pass however a complementary electrical filter of same kind is used in what i've done ( for a target 'named' config). Sometimes i used a filter in top of that hp for eg a 24db LR ( another 2pole 0,707 butterworth on top of the LT with same spec).
For reference target so far i used some profile from Rephase and before i recorded the output of my processor's curve ( all digital loop, no conversion so as clean as you could get. Both works well but you need harware with comprehensive digital in/out routing).
I didn't wanted to complexify the subject even more in case Pigmy decided to be part of the discussion again, and why i took caution in what i've said: there is multiple possible approach to this, some way clever than 'my brute force' approach.
Pelanji, what kind of expension ports do you have on your laptop? Adaptec had some cards availlable to expand laptop's capability. I use some for my older laptop for scsi and FW ( old studio gear in FW and circa 90's samplers) without glitchs for years.
Maybe they have new version of them for recent laptop?
There is no expansion port on the laptop I have now, just USB3. I still have the old one stored.
I made some proper two channel measurements and the results are much better now. Good news is that the ECM8000 with UMC1820 works on Linux. The only problem is that the measurement is clipping as -57 dBFS (some Linux driver issue/setting I guess), but I got pretty good results for now. Still not perfect, but much better than before.
I have another question. How to choose the HP slope for the HF driver not to overload it with low frequencies? First order would work for low volumes, but most loudspeakers use 2nd or 3rd order there. But that means inverting the phase for 2nd order, correct? Or is that compensated better with delay?
And a second question - in REW, should I watch the (generated) minimum phase?
I will clean up the measurements a bit and will post them later.
I made some proper two channel measurements and the results are much better now. Good news is that the ECM8000 with UMC1820 works on Linux. The only problem is that the measurement is clipping as -57 dBFS (some Linux driver issue/setting I guess), but I got pretty good results for now. Still not perfect, but much better than before.
I have another question. How to choose the HP slope for the HF driver not to overload it with low frequencies? First order would work for low volumes, but most loudspeakers use 2nd or 3rd order there. But that means inverting the phase for 2nd order, correct? Or is that compensated better with delay?
And a second question - in REW, should I watch the (generated) minimum phase?
I will clean up the measurements a bit and will post them later.
This isn't a simple question to answer for myself. I will refer you to Mulcahy's write-up on this subject, which is better than a "sound bite response":
https://www.roomeqwizard.com/help/help_en-GB/html/minimumphase.html
In general, it's faster to look at excess phase and excess group delay plots (the difference between the minimum phase and total phase, etc.) to see what you need to know, but there are some cases where you need to see the minimum phase plot--especially as you start to EQ at midrange and lower frequencies--toward the roll-off portion of the SPL plot. Flat portions of the minimum phase curve tell you that (generally) the driver is still in a good region. If you're looking at excess phase and it's not in a flat region chances are that you're trying to EQ something that's really the result of room (or internal loudspeaker/driver) reflections.
Chris
You can also use PEQs (large attenuating ones in the stop band below the crossover interference band) and shelving filters (negative slope looking toward lower frequencies, or positive slope looking from the LF stop band) to avoid the phase shifts of the "named" crossover filter types.
The way you can tell if you need to increase the effective high pass (combined) filter slope is to run the driver at 110 dB and look at the non-linear distortion. If you've got a problem, you'll see it there. Generally, modulation distortion will go as the harmonic distortion goes, so using HD is a way of seeing MD--which is the problem that I can hear (i.e., lower order harmonic distortion is generally not audible, but the modulation distortion it causes is audible).
Unless you've got a Klippel NFS with the nearfield distortion package, you're not going to be able to get a good idea of the modulation distortion curve. I've asked Mulcahy for a better test than a dual-tone test that he provides, but I got no response on AV Nirvana from him. I think it would be more than just nice to have a sweeping long-term dual-tone system available in REW, but thus far--no cigar.
Chris
Here's the link to "AN10"--Klippel's Application Note for AM and FM distortion: https://www.klippel.de/fileadmin/kl...es/AN_10_Loudspeaker_FM_and_AM_Distortion.pdf
Look at the last two figures in that PDF file for an idea of what modulation distortion curves look like. This is what sounds bad because it's all inharmonic, and therefore quite audible.
Chris
Look at the last two figures in that PDF file for an idea of what modulation distortion curves look like. This is what sounds bad because it's all inharmonic, and therefore quite audible.
Chris
Have done so many active setups so I don't use any frequencies measurment anymore. This is my approach in general.
I measure group delay with sound impulse and run an LR 12 db at the desired frequences. After that, balance the volumes on the amps. Most often important, is baffle step compensation (Q0.5) that you force down when you listen to a song you know. Finally, I look for a standing wave with a frequency generator and make a slim cut. Then runs the steps a few rounds until it sits.
With digital filter:
The last step is to find a volume that is physical without getting too loud when maximizing the digital volume control (to contain maximum bit depth on lower volumes). I trimming down the low end and then meet up with the other amps to get back to the desired blend achieved in the first part.
I measure group delay with sound impulse and run an LR 12 db at the desired frequences. After that, balance the volumes on the amps. Most often important, is baffle step compensation (Q0.5) that you force down when you listen to a song you know. Finally, I look for a standing wave with a frequency generator and make a slim cut. Then runs the steps a few rounds until it sits.
With digital filter:
The last step is to find a volume that is physical without getting too loud when maximizing the digital volume control (to contain maximum bit depth on lower volumes). I trimming down the low end and then meet up with the other amps to get back to the desired blend achieved in the first part.
Attachments
Last edited: