Ya I seen his name mentioned when I ddid a search, but couldn't find anything directly from him. Seems the idea has merits, though it demands a specific kind of driver for the filler.
Go here and ask your question. JohnK might answer or at least some of those who have tried his version might respond. The "filler driver" concept was thoroughly discussed there 2-3 years ago.
http://www.madisound.com/cgi-bin/audio_forum/index.pl
http://www.madisound.com/cgi-bin/audio_forum/index.pl
If I remember it correctly, JoknK once used the Vifa M11 as the filler driver. Later he switched to Vifa MG10. Unfortunately, both drivers were discontinued.
Wide bandwidth and gentle roll off at the upper end are the keys.
Wide bandwidth and gentle roll off at the upper end are the keys.
The filler driver concept work very well. However, as one poster all ready noted, finding a good filler is a tough act. You need a driver with very wide bandwidth that can handle the excrusion of a true 1st order at fairly low frequency. I currently favor my transient perfect 2nd order approach http://www.pvconsultants.com/audio/tp/tpdgam.htm
or if you have DSP and multiamp capability the subtractive delay approach
http://www.pvconsultants.com/audio/tp/tpsd.htm
I have a variety of transient perfect spread sheet here,
http://www.pvconsultants.com/audio/frdgroup.htm
or if you have DSP and multiamp capability the subtractive delay approach
http://www.pvconsultants.com/audio/tp/tpsd.htm
I have a variety of transient perfect spread sheet here,
http://www.pvconsultants.com/audio/frdgroup.htm
I've got the article from the Journal of The Audio Engineering Society on my hard drive. Email me if you want it.🙂
Hi, Read here:
(1) http://www.soton.ac.uk/~apm3/diyaudio/Filler_drivers.html
And if you download:
(2) http://www.geocities.com/kreskovs/GenFiller.html
Using (2) and setting Gamma = 1 gives you the graphing of the filter in (1) for all branches and in my opinion better than B&O’s because the resulting Q’s can be kept at 0.5.
B
(1) http://www.soton.ac.uk/~apm3/diyaudio/Filler_drivers.html
And if you download:
(2) http://www.geocities.com/kreskovs/GenFiller.html
Using (2) and setting Gamma = 1 gives you the graphing of the filter in (1) for all branches and in my opinion better than B&O’s because the resulting Q’s can be kept at 0.5.
B
Let's talk about practical suggestion.
Now the Vifa MG10 is discontinuted. IMO the other potential filler driver from the Vifa catalog would be PL11MH.
The little Fostex fullrange might also work, such as the FE103E. Though the response curve looks a bit rugge.
Scan 12M looks like a fit, but is very expensive. Esspecially when 2 filler per channel is need for WMTMW design.
Other interesting driver is the Dayton RS52, very well extended freq response. But there might be a problem of low end power handling.
Now the Vifa MG10 is discontinuted. IMO the other potential filler driver from the Vifa catalog would be PL11MH.
The little Fostex fullrange might also work, such as the FE103E. Though the response curve looks a bit rugge.
Scan 12M looks like a fit, but is very expensive. Esspecially when 2 filler per channel is need for WMTMW design.
Other interesting driver is the Dayton RS52, very well extended freq response. But there might be a problem of low end power handling.
Hi John, have you ever tried this one with an analog topology - i.e. making the delay with a chain of allpass filters ?
Regards
Charles
(JPK) It can be done, but it would require an excessive number of stages of 1st order allpass delays. The problem is that the delay has to be constant across the audio band which meams the Fc of the allpass has to be up around 100K with about 0.032msec delay per stage. Thus for a crossover based on a LR4 at 1.5k with DC GD = 0.3 msec you would need about 10 staged of 1st order delays. It's just not pratical.
I am afraid that it would be even more than that. OTOH 10th order can also be done by using 5 sections of 2nd order allpass filters.
I am particularly interested in trying it in the "some 100Hz" range where the ear is most susceptible to group-delay distortion at the cost of increased group delay (due to a finite number of allpass filters) above some kHz.
What do you think about that ?
Regards
Charles
I posted this question over on the digital forum, but got no takers.
If I set up my DCX2496 for overlapping 2nd-order Linkwitz-Riley slopes, reverse the tweeter polarity, and add apropriate boost at the XO point with the PEQ function, will I not achieve transient-perfect behavior, as described by Mr. Kreskovsky?
Seems logical to me, assuming IIR filters affect phase the same as analog circuits, but I'd appreciate some reality-check feedback.
You won't get a perfect step response and amplitude response that way but you should be able to get quite close.
Keep in mind that the 2nd order lowpass-filter with the lowest and flattest group-delay for a given pole frequency isn't the one with critical damping (Q=0.5) but the one with a Q of 0.577. And the higpass-filter would have a different Q and pole frequency than the lowpass.
But you could do much more stupid things with your spare time than that if you already have a digital x-over on hand to play around with.
Regards
Charles
Keep in mind that the 2nd order lowpass-filter with the lowest and flattest group-delay for a given pole frequency isn't the one with critical damping (Q=0.5) but the one with a Q of 0.577. And the higpass-filter would have a different Q and pole frequency than the lowpass.
But you could do much more stupid things with your spare time than that if you already have a digital x-over on hand to play around with.
Regards
Charles
I guess the difficult part is to maintain acoustic LR2 with addition overlap.
For reasonable amount of Q boost, say 6dB, Gamma would be 1.55. That means Flp = 2.4*Fhp
It might work better if you lower the electrical filter slope a little, trying to achieve an acoustic slope closer to LR2.
For reasonable amount of Q boost, say 6dB, Gamma would be 1.55. That means Flp = 2.4*Fhp
It might work better if you lower the electrical filter slope a little, trying to achieve an acoustic slope closer to LR2.
Bill F. said:
(JPK) You can never get a TP response if drivers are connected with inverted polaity (unless you introduce much more complex DSP than a DCX2496).
The problem with all of the TP design except the subtracted, delayed type is that they have rather shallow roll offs. If you want to try a 2nd order TP then you should down load my spread sheet from the FRDC. It will allow you to find the targets for the HP and LP filter Q and the Eq Q. I recommend an overlap factor, Gamma of 2 or more to avoid excessive boost from the eq. With Gamma = 2 the HP and LP are overlaped 2nd order with Q = 0.4. You can get Q = 0.4 by starting with an LR HP or LP filter and using a Q cut eq to reduce the filter Q. I presume the DCX2496 can introduce a delay, so the tweeter off set c an be handeled through that. The drivers should be connected in phase for any TP crossover.
That's very helpful. Thanks, John.
When the DCX2496 is set for LR2, I believe it inverts highpass polarity internally. That's why I mentioned inverting polarity externally--to return the outputs to same-polarity.
Yes, your spreadsheets are very helpful, too. I've been playing with them off and on for some time.
When the DCX2496 is set for LR2, I believe it inverts highpass polarity internally. That's why I mentioned inverting polarity externally--to return the outputs to same-polarity.
Yes, your spreadsheets are very helpful, too. I've been playing with them off and on for some time.
- Status
- Not open for further replies.