I have a pair of RCF N850 clones I didn't end up using for the project I had bought them for. Unity horns are catching my eye and might be a good use for these.
Lots of degrees of freedom in Unity designs... one that seems particularly difficult for me to get my head around is phase/time alignment between the mids and compression drivers.
What's the workflow like for that? If I use those N850s crossed at 650Hz, 24 dB/octave, and then choose the horn outer dimensions based on intended pattern control cutoff... then I can model the compression drivers and mids separately in hornresp and shift the mid entry points until they match phase at the crossover point... right? Any other gotchas here? Easier ways to do this?
Lots of degrees of freedom in Unity designs... one that seems particularly difficult for me to get my head around is phase/time alignment between the mids and compression drivers.
What's the workflow like for that? If I use those N850s crossed at 650Hz, 24 dB/octave, and then choose the horn outer dimensions based on intended pattern control cutoff... then I can model the compression drivers and mids separately in hornresp and shift the mid entry points until they match phase at the crossover point... right? Any other gotchas here? Easier ways to do this?
In order to achieve the famous sound of the Danley Synergy series of MEHs with their time alignment and very low phase growth vs. frequency, I'd recommend using no steeper than first order low pass filters.
I'd let the distance to the lower frequency driver ports determine the notch frequencies at the crossover frequency. Hornresp is your friend in determining the position of the off-axis ports. In general, they should be about a 1/4 wave axially along the main horn axis to determine the point at which they will have a cancellation notch.
The issue of course is the high pass roll off slope for your compression drivers--you can use two sets of attenuating filters--one at the crossover frequency and one at lower frequencies to cut short the interference band of the compression driver shallow slope.
I would not recommend using any steeper slope filters, especially those having orders greater than second order.
Just remember that there is no reason to use higher order filters in drivers used along an MEH. All the overlaps of driver acoustic output vs. frequency are summed within the horn's aperture, so the usual goal of using steep slope filters in the crossovers to minimize the interference band really has no place in the design. The real issue is trying to get the phase vs. frequency curve as flat as you can, and that demands different thinking--sort of like Dunlavy or Duntech like crossover filters.
JMTC.
Chris
I'd let the distance to the lower frequency driver ports determine the notch frequencies at the crossover frequency. Hornresp is your friend in determining the position of the off-axis ports. In general, they should be about a 1/4 wave axially along the main horn axis to determine the point at which they will have a cancellation notch.
The issue of course is the high pass roll off slope for your compression drivers--you can use two sets of attenuating filters--one at the crossover frequency and one at lower frequencies to cut short the interference band of the compression driver shallow slope.
I would not recommend using any steeper slope filters, especially those having orders greater than second order.
Just remember that there is no reason to use higher order filters in drivers used along an MEH. All the overlaps of driver acoustic output vs. frequency are summed within the horn's aperture, so the usual goal of using steep slope filters in the crossovers to minimize the interference band really has no place in the design. The real issue is trying to get the phase vs. frequency curve as flat as you can, and that demands different thinking--sort of like Dunlavy or Duntech like crossover filters.
JMTC.
Chris
Last edited:
One other comment: if you find yourself trying to get flat response vs. frequency out of your design without using notch filters to attenuate response peaks, instead trying to get flat response as the most important requirement, I think that you'll defeat the flat phase curve vs. frequency requirement.
The Danley crossovers use plenty of notch filters--especially in the midrange and compression driver bands--to flatten the response. This is where something like XSim or some other filter simulating app using "frd" files from the driver's frequency response on the horn to simulate the overall response after the electrical filter notches are designed to compensate for the constant coverage hump in response around 2-3 kHz.
Chris
The Danley crossovers use plenty of notch filters--especially in the midrange and compression driver bands--to flatten the response. This is where something like XSim or some other filter simulating app using "frd" files from the driver's frequency response on the horn to simulate the overall response after the electrical filter notches are designed to compensate for the constant coverage hump in response around 2-3 kHz.
Chris
Reading with interest. I cheated. I bought a pair of Yorkville Unity U15 "parts or repair" and modified them, active EQ. Due to limitations in testing gear, I have been able to EQ frequency to my tastes, but not get timing good. Using active EQ (JRiver) I have a lot of flexibility with my x-over, but the delay (tweeter to mids) was at best a guess. I used the wavelength of the cancellation notch (1.9KHz) which delay equates to about 0.24 msec. Recently I tried some "hunt and peck" longer delays (using pink noise around the x-over frequency) and come up with 80 msec. This seems to get a lot more of the sharp image and depth that "successful" Unity and Synergy owners crow about 
I guess the point of this blab is that there is still room for experiment, even with a (mostly) pre-built horn. If it matters, I'm using 48 dB/oct JRiver x-over. Cask05, I am no expert, but isn't using a first-order filter not recommended with the typically fragile compression driver?
I guess the point of this blab is that there is still room for experiment, even with a (mostly) pre-built horn. If it matters, I'm using 48 dB/oct JRiver x-over. Cask05, I am no expert, but isn't using a first-order filter not recommended with the typically fragile compression driver?
Last edited:
I've used a lot of compression drivers (2" mostly). I've not blown a single one, and all of them are being crossed below 500 Hz. Two of them I've currently had and run full time for the last 10 years are beryllium dome drivers (TADs).
Now if you're using your home loudspeakers like they're night club loudspeakers, I can see why you'll be worried about the diaphragms.
Soldermizer, I recommend reverse engineering the passive crossovers in your Yorkville Unity horn. I think that you will be surprised and also will think about abandoning that 48 dB/octave JRiver crossover. You should run measurements of phase on your Unities with passives and then of your active setup. Look closely at the dramatic phase growth with your active settings. This is I believe very important to understanding why so many people like the Danley synergies.
Chris
Now if you're using your home loudspeakers like they're night club loudspeakers, I can see why you'll be worried about the diaphragms.
Soldermizer, I recommend reverse engineering the passive crossovers in your Yorkville Unity horn. I think that you will be surprised and also will think about abandoning that 48 dB/octave JRiver crossover. You should run measurements of phase on your Unities with passives and then of your active setup. Look closely at the dramatic phase growth with your active settings. This is I believe very important to understanding why so many people like the Danley synergies.
Chris
I think the advice about using only first order filters succeeds if and only if you have very robust CDs, like CASK05, and in addition have placed the woofer ports such that the cancellation notch is well above the intended XO frequency - so that you have plenty of woofer-CD overlap.
When using more ordinary 1" aperture CDs that struggle to get below the 1100 Hz XO that typical mids on Unity horns struggle to reach up to, you have much less overlap and end up with steeper acoustical and electrical slopes. You are crossing over where the bandpass mid chamber's implicit acoustical low pass filtering generally gives you a steep slope. If I recall correctly in my design, I achieved 8th order acoustical slopes and used 4th order electrical filters to get there. The reason for that is that the BMS4550 CD had a 24 db acoustical high pass slope naturally and when I added to that to protect the diagphragm I ended up with 48 db.
I also raised the XO frequency as high as I could to minimize strain on the CD.
I don't think my sound suffered from employing high order filters; I tried lower order filters and didn't like the result. The difference is when you have a design with the driver overlap that allows first order XO filters then you have a chance to do a passive XO with flat phase. Without it you are forced to use the active approach but going active you have powerful tools in hand capable of dealing with the issues. One of those tools is FIR filters with which you can linearize the higher order XO filters and flatten the phase curve.
As to how to achieve time alignment, it requires careful measurements in a clean environment. The issue is discussed in some detail in my thread as I struggled with it initially. BWaslo talks about it also in some of his threads. In general, you want the start of the rise of the impulse of the CD to be aligned with the start of the rise of the mid's impulse response when each is measured separately with all the XO filters in place. It may take some iterations to get there...
When using more ordinary 1" aperture CDs that struggle to get below the 1100 Hz XO that typical mids on Unity horns struggle to reach up to, you have much less overlap and end up with steeper acoustical and electrical slopes. You are crossing over where the bandpass mid chamber's implicit acoustical low pass filtering generally gives you a steep slope. If I recall correctly in my design, I achieved 8th order acoustical slopes and used 4th order electrical filters to get there. The reason for that is that the BMS4550 CD had a 24 db acoustical high pass slope naturally and when I added to that to protect the diagphragm I ended up with 48 db.
I also raised the XO frequency as high as I could to minimize strain on the CD.
I don't think my sound suffered from employing high order filters; I tried lower order filters and didn't like the result. The difference is when you have a design with the driver overlap that allows first order XO filters then you have a chance to do a passive XO with flat phase. Without it you are forced to use the active approach but going active you have powerful tools in hand capable of dealing with the issues. One of those tools is FIR filters with which you can linearize the higher order XO filters and flatten the phase curve.
As to how to achieve time alignment, it requires careful measurements in a clean environment. The issue is discussed in some detail in my thread as I struggled with it initially. BWaslo talks about it also in some of his threads. In general, you want the start of the rise of the impulse of the CD to be aligned with the start of the rise of the mid's impulse response when each is measured separately with all the XO filters in place. It may take some iterations to get there...
Last edited:
I should have clarified the reason that I threw out a higher order crossover number: I'm planning on crossing over in software using FIR filters (Equalizer APO). As far as I know in that scenario I can tune the crossover phase response to what I want - I would still like to end up with a physical construction that makes sense. I'm still working through the patents to figure out what exactly that would mean.
For what it's worth, the N850's frequency response falls off a cliff below 500Hz, and so do the clones I have when I measured. Somewhere above that seems to me like a reasonable cutoff, although exact values could vary.
I'll check out your threads, nc535. I'm planning on 3D printing the horns as Mr. Waslo has done, so I'll have to think about how I could make that work with an iterative design... maybe something interesting could be done there with movable low range ports.
For what it's worth, the N850's frequency response falls off a cliff below 500Hz, and so do the clones I have when I measured. Somewhere above that seems to me like a reasonable cutoff, although exact values could vary.
I'll check out your threads, nc535. I'm planning on 3D printing the horns as Mr. Waslo has done, so I'll have to think about how I could make that work with an iterative design... maybe something interesting could be done there with movable low range ports.
Correction: current delay is 0.80 msec. Cask05, thanks for the suggestion. However, I am pretty sure the crossovers were blown when I got the speakers (units did not measure the same.) In any case, after I gutted the speakers to make them active, the unused hardware sat around for two years until I tossed it out. Even if I'd had a perfect unit, I do not (yet) have the capability to test a speaker properly. My bottle neck, if you want to call it that, is using JRiver for the active EQ. I simply can't get REW to work properly through JRiver. I have no problem running REW to a single ASIO channel to sweep a single driver. But how to do phase, etc.? One thing I'm going to try is using a 2nd PC to run REW, if I can figure out how to get the "JRiver" PC to take analog inputs and see if REW can get sweeps without JRiver's PC hiccupping too much during the measurements Even failing that, I may well take your suggestion and re-measure (Freq.) and optionally, try different x-over slopes (e.g. I is smart enough to match levels.)
Cask05, thanks for the suggestion. However, I am pretty sure the crossovers were blown when I got the speakers (units did not measure the same.) In any case, after I gutted the speakers to make them active, the unused hardware sat around for two years until I tossed it out. Even if I'd had a perfect unit, I do not (yet) have the capability to test a speaker properly. My bottle neck, if you want to call it that, is using JRiver for the active EQ. I simply can't get REW to work properly through JRiver. I have no problem running REW to a single ASIO channel to sweep a single driver. But how to do phase, etc.? One thing I'm going to try is using a 2nd PC to run REW, if I can figure out how to get the "JRiver" PC to take analog inputs and see if REW can get sweeps without JRiver's PC hiccupping too much during the measurements Even failing that, I may well take your suggestion and re-measure (Freq.) and optionally, try different x-over slopes (e.g. I is smart enough to match levels.)
Last edited:
The original driver that the OP mentioned has enough frequency response overlap using woofers only. I have no data on the "clones", however. You do not need midrange drivers.
Yes, but be careful of the pre-ringing issues.
I've found that it isn't an issue, if you know where to look. I use REW, and the REW plots of impulse spectrogram response are EXTREMELY sensitive in showing driver alignment. You can verify this time misalignment by also looking at the excess group delay plot, but you need to put down a lot of absorption between the microphone of the mouth of the loudspeaker before you take a measurement. The impulse spectrogram seems to be much less sensitive to measurement damping issues.
I've found that 475 Hz crossover point works well, and if you place the notch frequency at 400-450 Hz, you're good to go, but you need to remember to use crossover filter overlap to do that (just like Danley uses), and measure from the beginning lip of the off-axis port--not the center of the port. There are good reasons why you want to place the notch frequency close to the crossover frequency (as does Danley).
I also found that Hornresp yielded a fairly accurate assessment of the notch frequency. I'd aim the design notch frequency at 550 Hz within Hornresp and you should have enough margin. The woofers will have plenty of loaded response bandwidth to get you there, up until the 1/4 wave notch frequency begins to become significant, so you can set your notch frequency where you want to assure the handover from the compression driver to the woofer. I chose 475 Hz and got it on the first try.
See "A K-402-Based Full-Range Multiple-Entry Horn" for a two-way design using IIR filters from the DSP crossover (currently using a Xilica XP series crossover).
frequencyresponseandphasewithminphase.jpg.62d5d8add00e49ed06d45474dc2fcd86.jpg)
An externally hosted image should be here but it was not working when we last tested it.
Chris
Last edited:
You guys got me experimenting now...that, and me finally having a big TV screen so I can sit at listening position and tweak. I'm trying lower order x-overs: 12 db @ 1000 Hz for the "tweeter", and 6 dB/oct on high pass for the mids. Using my very limited adjustment tool (pink noise 1/3 oct, 1000 Hz), I get the best null with no delay. Go figure. I had not previously factored the advance or delay imposed by the x-over (since I only did raw driver measures, no active x-over in ckt.) Guess I will have to figure out how to run reputable tests on my system
I had a load of posts saved by Tom D somewhere but can't seem to find them. I'm sure he mentioned that on the Synergys he often ends up with a '4th order or higher' electrical HP on the tweeter. Something about crossing over higher than the drivers natural roll off as well to keep the phase smooth.
Rob.
Rob.
I found that the K-402-MEH could almost run without crossover filters or delay. I didn't do that, however, since I've always used a DSP crossover and had filters available to create a better interference band frequency response....I'm trying lower order x-overs: 12 db @ 1000 Hz for the "tweeter", and 6 dB/oct on high pass for the mids. Using my very limited adjustment tool (pink noise 1/3 oct, 1000 Hz), I get the best null with no delay. Go figure. I had not previously factored the advance or delay imposed by the x-over (since I only did raw driver measures, no active x-over in ckt.) Guess I will have to figure out how to run reputable tests on my system
The delay values that I currently use are small: 0.3 ms on the midrange diaphragm and 0.6 on the tweeter diaphragm (using a dual diaphragm compression driver).
Chris
Thanks for inspiration
Good news! I got off my lazy a$$ and hooked up the laptop + measurement stuff to the "TV Box" (media player + JRiver) and fired up REW and damned if it doesn't work perfectly. No hiccups and sweeps sound and, at least a first appearance*, appear to work as they should. At last, after two or more years, major opportunity to muck with (or some thing that rhymes! ) the system again! More precisely: now I can properly use REW or similar tools to test my system with the DSP in the circuit.
In the "Too good to be true?" department, I did a few sweeps of the upper x-over (mine is set at 1KHz) and phase looks too good to be true. Other sweeps, at this position with wider frequency sweep and also several at LP, look more honest, with multiple phase flips as any (?) system would show in an untreated room.
Good news! I got off my lazy a$$ and hooked up the laptop + measurement stuff to the "TV Box" (media player + JRiver) and fired up REW and damned if it doesn't work perfectly. No hiccups and sweeps sound and, at least a first appearance*, appear to work as they should. At last, after two or more years, major opportunity to muck with (or some thing that rhymes!
) the system again! More precisely: now I can properly use REW or similar tools to test my system with the DSP in the circuit.In the "Too good to be true?" department, I did a few sweeps of the upper x-over (mine is set at 1KHz) and phase looks too good to be true. Other sweeps, at this position with wider frequency sweep and also several at LP, look more honest, with multiple phase flips as any (?) system would show in an untreated room.
Attachments
I've learned that there really aren't any of those when it comes to measurement. However, you may have found a particular "sweet spot" with your microphone and loudspeaker/room position that shows minimum phase behavior.
In any case, never throw away particularly good looking phase measurements.
As you move down in frequency, and particularly if you've got the MEH close to a floor, wall or corner, the 1/4 wave within the horn begins to use these boundaries for its fundamental wave (i.e., boundary gain), so you will see multiple sources from your microphone measurements. The trick is to position the microphone to minimize extra nearfield reflections and use frequency-dependent windowing (FDW) to exclude those early reflections that obliterate the phase information. I use at least 6 inches of absorption material on the floor between the loudspeaker and the microphone and extra absorption on the side of the cabinets wherever possible to attenuate the early reflections.
Great to know that my measurement microphone, taped to the back of a straight back wooden chair, sitting atop an Ikea Lack coffee table, perhaps 4 feet and 30 degrees off center of left speaker, is a sweet spot
Good point is that I can re-eq my system with some hope of accurate measurements, at least of frequency. Phase is still a chimera
Good point is that I can re-eq my system with some hope of accurate measurements, at least of frequency. Phase is still a chimera
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.