Ben,
The link will explain in detail everything I attempted to. 6o6 (Bob) is a better writer than me, I should have just linked it from the git go, instead of doing the "Cliff's Notes" version
.I have included the phase and frequency response of my 3 way system as a visual aid to Bob's article, as you can see, the 24 dB per octave crossovers result in 720 degree phase shift, the horn loaded driver's response are responsible for the remaining shift.
Even though it is a "time train", all the cars are smoothly connected in phase, the crossover points are undetectable. One advance in technology since Bob's article is that using FIR filters, a flat phase over the entire passband is now possible.
You could measure delay on either side of the horn mouth, since it is a symmetrical path. The distance to your listening position would, as you noted, be different from either side of the mouth, though the 2 foot distance difference is still within 1/4 wavelength at 100 Hz, so is of no concern.
The Behringer DCX2496 DSP latency is <1ms, about half the latency of the DriveRackPa DSP I presently use.
Attachments
Last edited:
Many thanks, Art. With the good advice from you and 6o6, I can now present better information, with apologies for my earlier errors.
REW provides a time-zero setting using a loop-back signal (when I remember to use it). That is what I use here. But I also eyeball the impulse responses as they were scope traces to verify stuff (and because you should never let machines do all your thinking).
But I can't find a means of filtering the mid and tweeter within REW in order to "level the playing field" with the Klipsch sub.
With no DSP delay added, the signals get through my pre-amp and Behringer DSP in these times:
Klipsch - 7.23 ms
DW mid - .82 ms
ESL tweeters - .80 ms
In a post a few spaces up, I falsely said the delays were under 1/4 ms. Sorry.
With the DSP delays to the mid and tweeter, switched back in, the measured total delays are arithmetically right-on, which I find surprising but reassuring.
How big is the Klipschorn? As of this stage of analysis, it looks like about 6.5 ms is differentially added to the Klipsch inside the DSP. So, that makes the Klipsch about 7 feet shorter than previously thought and a lot closer to the expected value. Crunching tape-measure distance and impulse time values, the previous estimate of 20.25 feet should now be about 13 feet for the insides of a Klipschorn.
I still am unsure of other subtle aspects of using impulse responses to measure distances and doing so acoustically and when comparing subs and mids, but that's it for today.
Anybody else with a folder corner horn with something to add?
As far as the raging discussion here of best method for aligning subs and mids, that's next.
Ben
REW provides a time-zero setting using a loop-back signal (when I remember to use it). That is what I use here. But I also eyeball the impulse responses as they were scope traces to verify stuff (and because you should never let machines do all your thinking).
But I can't find a means of filtering the mid and tweeter within REW in order to "level the playing field" with the Klipsch sub.
With no DSP delay added, the signals get through my pre-amp and Behringer DSP in these times:
Klipsch - 7.23 ms
DW mid - .82 ms
ESL tweeters - .80 ms
In a post a few spaces up, I falsely said the delays were under 1/4 ms. Sorry.
With the DSP delays to the mid and tweeter, switched back in, the measured total delays are arithmetically right-on, which I find surprising but reassuring.
How big is the Klipschorn? As of this stage of analysis, it looks like about 6.5 ms is differentially added to the Klipsch inside the DSP. So, that makes the Klipsch about 7 feet shorter than previously thought and a lot closer to the expected value. Crunching tape-measure distance and impulse time values, the previous estimate of 20.25 feet should now be about 13 feet for the insides of a Klipschorn.
I still am unsure of other subtle aspects of using impulse responses to measure distances and doing so acoustically and when comparing subs and mids, but that's it for today.
Anybody else with a folder corner horn with something to add?
As far as the raging discussion here of best method for aligning subs and mids, that's next.
Ben
Last edited:
Ben,
Now your measurements agree within .5ms with my estimate using an envelope to measure the path length of the "exploded" Khorn I posted.
With that as a start, IIRC the Khorn is off to one side, so splitting the difference between the two mains plus the 6.5ms will probably be as good a compromise as any.
Looking at the phase and frequency response in the crossover region will help you fine tune the delay, but the room modal response and the distance differences between your mids and subs to the LP will make an even phase match an impossibility in any location, but you knew that coming in.
One possibility you may not have considered would be centering the Khorn, laying it sideways using the floor/wall junction rather than a corner for the horn mouth. If that were possible, you could achieve phase alignment in the center, other than the contribution of the other sub...
Have fun!
Art
Thanks, Art.
Glad to hear my crude experiment resemble your estimates.
The two mids are same distance to chair and so Klipsch is same distance behind either, even though it is behind just right one. Which doesn't mean the three aren't mixed up in other ways.
Big corner horn on the side? Even too weird for our decor (and besides, you need a cap on "top" of Klipschorn... I made one but uses gravity and my wine rack to stay in place).
Tried one-octave pink noise (as narrow as I could get REW). Maybe there's a way to read output in all the jumping around but too bumpy for my eyes.
Tried REW's "CEA 2010 Burst" which is a 6-cycle burst. Seems to produce stable readings on the SPL simulator when you look at the max reading. Doing pole reversal moves the reading about just 1-2 dB (but, glad to say, favouring the polarity I had settled on previously). With sharp crossovers and my geometrically complex speakers, couldn't expect more. Of course, a tone-burst contains a lot of frequencies besides the nominal one. Indeed, freq response changes a fair amount with polarity reversal to my Klipsch and/or giant OB sub - but often quite hard to decide which set of bumps and dips I like better than the other polarity, no kidding.
I may someday have the energy to trace down optimum phase settings using a tone burst (which is a speaker oscilloscope test I like a lot).
As 6o6 points out, you need to work at the acoustic balance point of the crossover, not the nominal electrical crossover point.
Ben
Glad to hear my crude experiment resemble your estimates.
The two mids are same distance to chair and so Klipsch is same distance behind either, even though it is behind just right one. Which doesn't mean the three aren't mixed up in other ways.
Big corner horn on the side? Even too weird for our decor (and besides, you need a cap on "top" of Klipschorn... I made one but uses gravity and my wine rack to stay in place).
Tried one-octave pink noise (as narrow as I could get REW). Maybe there's a way to read output in all the jumping around but too bumpy for my eyes.
Tried REW's "CEA 2010 Burst" which is a 6-cycle burst. Seems to produce stable readings on the SPL simulator when you look at the max reading. Doing pole reversal moves the reading about just 1-2 dB (but, glad to say, favouring the polarity I had settled on previously). With sharp crossovers and my geometrically complex speakers, couldn't expect more. Of course, a tone-burst contains a lot of frequencies besides the nominal one. Indeed, freq response changes a fair amount with polarity reversal to my Klipsch and/or giant OB sub - but often quite hard to decide which set of bumps and dips I like better than the other polarity, no kidding.
I may someday have the energy to trace down optimum phase settings using a tone burst (which is a speaker oscilloscope test I like a lot).
As 6o6 points out, you need to work at the acoustic balance point of the crossover, not the nominal electrical crossover point.
Ben
Last edited:
Haven't read the REW manual, but with Smaart you use full range pink noise for both phase and frequency response, which have two separate screens you can look at. It has different time constants, when the longer averaging is used, the traces stop "jumping around", but wind noise can make the trace "goofy", so for outdoor testing I have to go shorter and "pull the trigger" between gusts..
Recalled this thread as I'm learning REW.....
Decided to check DCX2496 latency as the sole DUT before trying to move on to speakers....
Found the DCX2496's latency with no processing engaged other than routing, to be 0.7ms.
X-over page: With X-over engaged as 'low pass' to subs....(no 'high pass' in use)
LR12 2.9ms
LR24 6.7ms
LR48 13.3ms
So roughly 3ms per 12db, after subtracting base latency.
One example with both LP and HP........ LR24, 18.2ms
EQ page didn't add latency.
I also measured a QSC PL340 amp at 3.1ms at 30hz & 50hz filter settings... and the manual says it's a 2nd order (12db) filter.
Do all digital xovers add latency in the neighborhood of 3ms per 12db? (Excluding FIR filters)
Decided to check DCX2496 latency as the sole DUT before trying to move on to speakers....
Found the DCX2496's latency with no processing engaged other than routing, to be 0.7ms.
X-over page: With X-over engaged as 'low pass' to subs....(no 'high pass' in use)
LR12 2.9ms
LR24 6.7ms
LR48 13.3ms
So roughly 3ms per 12db, after subtracting base latency.
One example with both LP and HP........ LR24, 18.2ms
EQ page didn't add latency.
I also measured a QSC PL340 amp at 3.1ms at 30hz & 50hz filter settings... and the manual says it's a 2nd order (12db) filter.
Do all digital xovers add latency in the neighborhood of 3ms per 12db? (Excluding FIR filters)
DSP is a "tower of babble", I would not make any blanket statements without testing, but I think you will find the latency is a function of the crossover frequency as well as the slope- lower crossovers add more delay, higher less.
Each pole in an analog emulating crossover adds 90 degrees of phase shift, 360 degrees is the the time of a full wavelength, the time delay a pole adds at 1000 Hz is 1/10th that at 100 Hz.
Art
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.