May I suggest that you do one clean measurement at 1 meter / 90 dB (fairly loud) and then post the .mdat file as a zipped attachment.
Or post the .mdat file for what you get the graph in post #35 from.
//
Or post the .mdat file for what you get the graph in post #35 from.
//
Looks like both woofer (5 ms) and mids (less than 1 ms) are ahead of the tweeter, at this distance. Somewhat further out, 1m from mouth like Mark suggests should be saver to see proper timing. Do you use phase linear crossovers? While Tom Danley shared some hints how he did his crossover, I don't think he ever shared the complete topology.
I'm glad I nagged John (maker of REW) for these type of wavelets though, as it provides a nice visual tool.
I'm glad I nagged John (maker of REW) for these type of wavelets though, as it provides a nice visual tool.
Try flopping the woofer channel polarity and remeasure. The good thing about MEHs is that the crossover regions are usually pretty easy to get very close. Using only first order crossover filters will likely get you quite close to time aligned. No channel delays are needed in the case of using first order crossover filters-since internal summing of drivers within the aperture occurs. This also means that you don't need sharp slope crossovers since the reason for using them is absent with MEHs (due to internal summing).
The way that I measure time alignment of drivers (on any loudspeaker) is by looking at the excess group delay curve within REW. You have to generate this curve using "generate minimum phase" or "minimum group delay" curves, then select the excess group delay curve checkbox below the plot to see this curve.
If you're using "zeroth order" crossover filters (PEQ and low shelf/high shelf only), you'll need to add delay to the lower frequency driver channel equivalent to ~90 degrees of phase shift at the crossover point (first throat bounce frequency for the woofer or midrange) as a starting point. This is the beauty of MEHs--time alignment is straightforward and isn't sensitive to the microphone position vertically (i.e., the drivers are effectively coaxial)--unlike "conventional" loudspeakers having separate apertures for each driver, which makes the required delays sensitive to where you put your microphone.
BTW: the K-402 runs out of 1/4 wavelength axial path length at ~172 Hz, so if your horn is larger than a K-402, you're probably crossing the woofer-midrange right at the point where the internal wavelengths are transitioning to using the room boundaries including horn flanges, so you might get a little SPL discontinuity there (usually very narrow in bandwidth that you can't detect it). Try moving the horn mouth closer to a room boundary to see if if changes the depth of the local 150 Hz null.
Chris
I would like to see if the concept can produce low distortion. I think it is desirable with low distorsion for a faithful reproduction.
//
//
I believe it can based on the writings of Tom Danley and my own experiments. However Danley does do more with higher levels such as outdoor products. I thought we were past looking at distortion as being important with speakers that are otherwise well designed.
Hi, hope the tuning process is going well 🙂
Hey, if you're having any trouble nailing down delays as precisely as you'd like, I'd be happy to offer the process i use.
It's really just a way of using accurate time-of-flights to each driver section after they have been individually tuned, but there are a few tricks to making it easy.
I find using ToF's much more accurate, and easier, than using spectrograms or group delays...
If you're interested, lemme know.
Hey, if you're having any trouble nailing down delays as precisely as you'd like, I'd be happy to offer the process i use.
It's really just a way of using accurate time-of-flights to each driver section after they have been individually tuned, but there are a few tricks to making it easy.
I find using ToF's much more accurate, and easier, than using spectrograms or group delays...
If you're interested, lemme know.
Indeed, the visual confirmation isn't all that accurate to show the real time delay differences. Be it a wavelet or a group delay graph.
It's just an easy way to visually see if you're in the ball park and nothing more. Use the IR to determine time of flight or to check time differences between the two sources.
For instance, if you don't have a dual channel measurement setup available to help you find exact time of flight, one can add a specific amount of delay to one of the drivers (or driver groups) and measure both drivers (or driver and driver group) of interest simultaneously. With enough delay on one of them, you can see both IR peaks clearly and determine the exact time difference between both drivers (or driver groups) that way. Choose a band limited signal that both driver can play. There are other options as well, for instance the use of an acoustic timing reference feature. Basically it works very similar to the trick I just mentioned using 2 drivers (or driver groups).
For each measurement it will first send a chirp which is used to determine it's timing. For timing difference the first option should be time of flight using dual channel, if that's not an option the above text could help figure things out.
It's just an easy way to visually see if you're in the ball park and nothing more. Use the IR to determine time of flight or to check time differences between the two sources.
For instance, if you don't have a dual channel measurement setup available to help you find exact time of flight, one can add a specific amount of delay to one of the drivers (or driver groups) and measure both drivers (or driver and driver group) of interest simultaneously. With enough delay on one of them, you can see both IR peaks clearly and determine the exact time difference between both drivers (or driver groups) that way. Choose a band limited signal that both driver can play. There are other options as well, for instance the use of an acoustic timing reference feature. Basically it works very similar to the trick I just mentioned using 2 drivers (or driver groups).
For each measurement it will first send a chirp which is used to determine it's timing. For timing difference the first option should be time of flight using dual channel, if that's not an option the above text could help figure things out.
Wesayso...what do you think the time alignment requirement is?
(If you don't already know for this sort of large MEH, the answer is "less than 1/4 wavelength"). That's actually all it really is. If you don't believe me, build yourself a K-402-sized MEH and cross the woofers at 150 Hz and do the iterations to see the results. I think you'll see and hear what I'm talking about--quite clearly.
What is the period of a 1/4 wavelength that corresponds to 170 Hz? No time delay at all is actually necessary on the woofer, but a 1/4 wavelength delay ref. the first notch frequency that comes automatically with a first order filter will get you a lot better than "visual confirmation on a wavelet or group delay plot". Please do not dismiss the methods that you just did, without first understanding the requirement that's a result of this particular instance of the physics involved.
I find that the resulting phase and group delay curves should stay under the "just audible thresholds" at the frequencies being discussed (i.e., 150 Hz). That really is the secondary requirement.
But per your text quoted just above, it appears that you've already decided that the requirement is something else--basically "much, much tighter" because you know how to do it that way. Where do you get your requirement from for MEH time alignment of drivers? I got mine from Danley himself and my own measurements of his SH-50 MEH, along with my prototype K-402-MEH.
For communicating with someone just starting out using time alignment techniques and REW (i.e., the OP), your text that I highlighted just above is--well--not terribly useful. Think about this--and the comment that I made about "internal summation within the aperture". If you did understand my comment about internal summing within the aperture, I don't believe you would have written the text quoted just above.
If you're trying to teach a newcomer (that I just happen to have known for several years on another forum) "the right way to do things", consider the likely outcome that you simply drive him away. That would be unfortunate, indeed.
Chris
I think you may have misread my intentions and even the difficulty of my proposed method. It's far easier than you assume here. I do believe the getting the time of flight method would be far more challenging for a newcomer. That's why I wanted to offer an easy alternative.
If you realize how simple this method actually is I think you'd embrace it yourself, you'd just have to try it once. It works with an USB mic or a phantom mic and gets you there real fast. The only requirement would be to temporarily delay one of the driver(groups) far enough to see both pulses clearly. What's not to like? My wording perhaps, as I'm not native English. If I were in proper health today I'd start up an REW session to demonstrate it with pictures. I guess it could take less tan 15 minutes. A tutorial sort of post, useful for more than one situation would have been more helpful.
At 150 Hz, the visual confirmation would be all one would need. Yet at the first crossover, just over 1.2 KHz an easy IR trick could come in handy.
If you want to help @ncbluetj, go right ahead, no problem. I'll gladly step aside.
If you realize how simple this method actually is I think you'd embrace it yourself, you'd just have to try it once. It works with an USB mic or a phantom mic and gets you there real fast. The only requirement would be to temporarily delay one of the driver(groups) far enough to see both pulses clearly. What's not to like? My wording perhaps, as I'm not native English. If I were in proper health today I'd start up an REW session to demonstrate it with pictures. I guess it could take less tan 15 minutes. A tutorial sort of post, useful for more than one situation would have been more helpful.
At 150 Hz, the visual confirmation would be all one would need. Yet at the first crossover, just over 1.2 KHz an easy IR trick could come in handy.
If you want to help @ncbluetj, go right ahead, no problem. I'll gladly step aside.
Last edited:
Sorry to hear about your health. Hope that you are recovering quickly. Also, I was aware of your native language difference. We have actually seen eye-to-eye in the past, so the current disagreement is a statistical outlier.
Keeping the learning process as simple as possible I think is a crucial requirement to allow those just coming up to speed a moment to use the standard plots that are available within REW without having to take any extra steps. Using existing plots within a typical REW measurement is really about as simple as it gets.
Also note that my reaction, above, was not unfounded. The problem, I think, is that someone thought that the method I briefly described (using excess group delay, step function, phase response, and spectrogram plot, etc.) is not "good enough" for this application. It is, in fact, much more than good enough for the issue at hand--an MEH woofer-midrange crossover.
MEHs don't require time alignment in the traditional sense, but rather the use of first order crossover filters to provide the amount of delay necessary. Anyone that uses high-order crossover filters in an MEH simply doesn't understand MEHs very well. The very reason for using higher order crossover filters does not exist in MEHs, i.e., crossover interference band polar lobing. Phase fidelity without the use of FIR filtering and coaxial driver performance without polar lobing effects have been Mr. Danley's prime message regarding MEHs for more than 15 years.
Chris
I think you are correct that a well constructed Synergy with proper driver selection and port placements, lends itself very well to using low order, even first-order, crossovers. I've read and studied all the patents too, and think i understand them well.
I also realize the whole Synergy invention was designed to meet the install market's needs, where passive xovers are often a major advantage.
IMO, first order in xovers are just a component of maintaining a simple passive design taking advantage of favorable z-axis driver spacing, to achieve low phase rotation. But they are neither strictly necessary, or entirely optimal.
I think you are incorrect if you say no further delays are ever needed. I have read Tom Danley talk about the difficulty in getting the needed delays with different horn coverage angles with their associated different horn depths, when using passive xovers.
And it's off the mark in many ways to say "Anyone that uses high-order crossover filters in an MEH simply doesn't understand MEHs very well".
I too have had the opportunity to discuss xover/tuning processes with Mr Danley, and received strong positive confirmation of my complementary linear phase high-order method.
There's always multiple methods to approach electrical tuning of a good acoustical design.
It's worth noting that the upcoming Danley Signature series speakers for home audio have been touting FIR filtering in their preview literature.
It seems he would disagree with the idea phase fidelity without the use of FIR filtering remains his prime message.
Ok, do Synergies need anything beyond being well laid out and using low order xovers, ....no they don't have to have anything else, that's their glory.
Can they be made even better ?
Measurements?...certainly.
I know from measurements, that phase alignment and acoustic summation between sections can be improved with precise delays. As compared to relying only on 1/4 WL spacing, which can sum anywhere within a 3 dB range.
I know mag, phase, and impulse can all be improved vs just first-order and rough delays.
Audibly?...surely always debatable....and for each to decide. Never worth disagreeing over imho.
All that said, I'll still offer a method to nail timing down, that i think is both more accurate and easier, than using group delay and/or spectrographs.
Again, it's the simple use of ToF measurements (best on drivers that are first individually tuned.)
It doesn't depend on FIR or dual channel measurements.
It does need at least REW, and a USB soundcard with loopback, and a non-UMIK mic. (UMC202HD and ECM8000 will do, for a total of $105 .)
Simply measure each driver (or driver sections) one at a time, from a fixed point in space, and compare their Delays relative to loopback on their measurement tabs. These are their individual ToF's to a point in space (which include any fixed delays like processor latency etc)
The differences in those 'Delay vs loopback' times, are dang close to being the exactly needed delays. (Certain little tricks can make them exact.)
On a syn, the lowest driver section (LF) will measure the greatest delay vs loopback (furthest from mic).
The mid driver section (MF) the next. And the CD the least time. So mid delay = LF-MF, and CD delay = LF-CD
Make sure Timing offset is set to zero as shown for all sections, and make sure all non-tested section are either shorted (or connected to an amp.)
Oh, and it must be relative to loopback, and not acoustic reference which depends on a 5-20kHz chirp. Chirp is too high to work, and becomes the reason UMIKs don't work, and a soundcard is needed.
Like said, the process gives delay settings far more accurate, and than looking at group delay or spectros, ime/imo.
And again, it's independent of using FIR or steep xovers.
I use both of those because they simply measure better, and are much easier to implement than low-order xovers, no matter whether low order is IIR or linear phase FIR.
Here's a tuning made two days ago on a 4-way syn on top of a bass-reflex sub.
The synergy needed 2.4ms delay relative to sub.
The syn's delays are: Low 0ms, Mid 0.2ms, CD HF section 0.68ms, and CD VHF section 0.76ms.
Hey, at least such exact tuning produces a pretty spectro ! And was easy-peasy to do !!
Last edited:
Happy Monday everyone! I had a very productive weekend with the big horns. As Chris predicted, I had my woofer polarity reversed. Once I fixed that and EQ'ed everything again, I got some good results. I was also able to improve my driver time alignment considerably. I'm not sure I have it perfect, but it is close, I think. Anyway, here is a FR and phase graph:
I tried to get three different distortion measurements, at progressively louder volume. I discovered something interesting. It seems my woofers are not keeping up as the volume increases. I'm not sure what to make of that. Maybe I am simply running out of power? I have a 150wpc rated amp on the bass drivers, which I would think is plenty, but maybe not? Or perhaps it is a gain structure issue of some kind? Everything seems to scale normally at sane volume levels,it only happens when it gets pretty loud. Any ideas?
85dB distortion:
85dB distortion: