Fair enough.
Having accidentally destroyed an expensive driver by driving it too hard too low, I'd be hesitant to apply low-shelving boost to a driver that's effectively unloaded.
I do see that your application makes sense, and while I'd use a different EQ method myself, if it works for you then I won't argue.
Chris
Chris, you can certainly EQ a dipole in several different ways. For example one EQ stage for the 6dB dipole rolloff, andother for woofer or low end boost EQ, another for the dipole peak, etc., etc. I find I can do much of it with just the LT type biquadratic filter.
It strikes me how many posters harp on how the LT can easily overdrive amp, speaker, etc. and this is "bad". Well perhaps the DIYer should pay close(r) attention to the total boost being deployed and dial back the maximum boost by reducing the total gain for that channel? For example, a 2 octave extension from 100HZ to 25Hz will require 24dB of gain. That is a huge amount of voltage gain and this will easily clip on even modest signal levels. But if you reduce the gain on the woofer (and the rest of the system) by say 12dB or more you get a much more managable situation. The problem is that you then lose overall system gain and your speaker can no longer play all that "loudly".
With this problem in mind, I wrote a LADSPA DSP lookahead limiter back in 2016 based on the Linkwitz Transform. The way it works is that you tell it what LT boost you want to apply. For example my 24dB example to extend the cutoff from 100 to 25 Hz. In software DSP, the maximum output signal is +1.0 to -1.0 in the internal representation of the signal. The program looks ahead in the audio stream (this requires latency) for peaks that will exceed this level. It then smoothly transitions the LT boost to a less extended one, e.g. instead of the 25Hz cutoff it can dial it back to 40Hz 50Hz 60Hz or whatever is needed, all the back to no boost. This reduces the gain, and will eliminate the offending peak. This approach permits the use of lots of boost, and when the signal level is low you will still get the maximum extension. When the signal rises the LF extension is automatically dialed back, and when the loud passage passes it's extended again. I think this is pretty useful. Unfortunately, there was no real interest in it (a total of one person tried it), and it was put down by forum members as a "cheap" trick where a more capable driver should be used. So I never published it. More info here:
IDEA: Linkwitz-Transform LADSPA plugin with lookahead boost control
Charlie,
It's a real shame that it didn't go further. A few years back, I was using a bunch of 15" sealed subs in my PA system, which of course required EQ. Having a limiter that works how you've said would've been amazing. If the latency was sensible, I'd've been happy to pay for that!
If there's a way of applying similar excursion limiting for ported boxes (perhaps feed it a particular EQ'd curve, and then apply the inverse post-limiting), that would still be a really interesting project. The system processing I use at the moment is very good, but doesn't have excursion limiting.
I'll read through the thread you've linked.
All the best,
Chris
I completed development in 2016. The code is just sitting around, unused, but is IIRC fully functional. I could dust it off and let any one interested give it a try.
The minimum latency is one audio frame (e.g. 20msec but this depends on the LADSPA host application). Longer latencies on the order of 100-200msec provide a slower release and would be better suited to very low frequency (e.g. 10-30Hz) applications because they are more in line with the period of those frequencies (e.g. the period of 10Hz is 100msec). I fear that with too low a release time the code will do some funny things with the waveform that will look more like clipping.
Keep in mind that, to keep the audio sync'd with the other drivers, you should delay them by the same latency that is generated by the lookahead limiter.
The wonderful Behringer DCX2496 has the most amazing elaborate set of set and dynamic limiters you could imagine. Not to mention EQs and crossovers that can be set sub-sonically or any way you want.
If you want to go to an OB sub or really push any kind of bass boost, it sure would make sense to buy one (cheap on eBay).
B.
If you want to go to an OB sub or really push any kind of bass boost, it sure would make sense to buy one (cheap on eBay).
B.
Mmm, not really. The limiters in the DCX are not "lookahead" types. Big, big difference.
While the DCX has several preset filter types with adjustable parameters, it's not really all that flexible because Q is fixed by the filter type you choose and it is often useful to tweak the Q and corner frequency of each stage to do some response shaping. I had one back around 2006, but I got rid of it when the miniDSP came out. In about 2012 I migrated from miniDSP to doing all my signal processing under linux using LADSPA plugins that I wrote.
If there is something new and different that I want to do I have the freedom to code it up. Fractional delay, lookahead limiter, etc. With a computer-based DSP platform, I have freedome to choose the ADC and DAC, meaning I can control the level of performance or match it to my budget. The software platform gives me everything I could ever want for signal processing, and is more capable and flexible than most hardware DSPs. At least for me, this is where it's at for DIY DSP crossover work.
DCX is flexible to a fault in almost any imaginable way including Q for EQ (the "fault" being the steep learning curve).
Respectfully, you are looking at the problem from an odd perspective, as if you know what is needed and you plan the correction from theory. I look at the problem empirically: the measurements are the only criteria and the DCX lets you produce the sort of output you want by a combination of input and output controls for EQ, xo (and even bass and treble that do not have the same turnover freq, like I run mine), time alignment, phase tweaking, and limiting.
For low notes, given the unpredictability of any speaker in a room and double that for dipoles, sure makes sense to have DSP tools or the CharlieLaub software ready to go.
B.
Last edited:
UPDATE:
I've been listening to this speaker for several weeks now. I recently made some changes to the crossover that allowed me to reduce the crossover frequency between the mid/woofer tweeter and flatten the response above 300Hz. I'm listening now as I write this, and it sounds great. This is worth keeping around for sure, if for nothing else as an example of how to do a nude 3-way dipole.
I've been using a closed box sub so far. I wanted to compare CB and dipole subs, and I happened to have a couple of subs ready to go so it made sense to start there. Next up will be a dipole sub. My plan is to build an "M frame" using two Eminence Alpha 15A woofers per side in an H-frame with D=14". Some preliminary measurements using only a single woofer (as an H-frame) are encouraging. The plan is to parallel two of the Alpha 15As, resulting in a 4R load that has very high voltage sensitivity and that can cross over to the mid/woofer around 250-300Hz. This will also allow the mid/woofer to be used in a more efficient way then when it is crossed to the CB subs around 100Hz. As a result the efficiency of the entire system should go up.
There is currently a heat wave in place here, but next week should be cooler and I hope to be able to cut some wood and build the M-frame assembly, or at least make good progress towards that goal.
I've been listening to this speaker for several weeks now. I recently made some changes to the crossover that allowed me to reduce the crossover frequency between the mid/woofer tweeter and flatten the response above 300Hz. I'm listening now as I write this, and it sounds great. This is worth keeping around for sure, if for nothing else as an example of how to do a nude 3-way dipole.
I've been using a closed box sub so far. I wanted to compare CB and dipole subs, and I happened to have a couple of subs ready to go so it made sense to start there. Next up will be a dipole sub. My plan is to build an "M frame" using two Eminence Alpha 15A woofers per side in an H-frame with D=14". Some preliminary measurements using only a single woofer (as an H-frame) are encouraging. The plan is to parallel two of the Alpha 15As, resulting in a 4R load that has very high voltage sensitivity and that can cross over to the mid/woofer around 250-300Hz. This will also allow the mid/woofer to be used in a more efficient way then when it is crossed to the CB subs around 100Hz. As a result the efficiency of the entire system should go up.
There is currently a heat wave in place here, but next week should be cooler and I hope to be able to cut some wood and build the M-frame assembly, or at least make good progress towards that goal.
With a passive XO, makes no sense to design a XO from First Principles and then have to open up the speaker box to - inevitably - tweak the xo once it is installed.
Let me think about that a minute. You added capacitance (in parallel) to bring down the upper range driver. That's easy. And you..... um, um..... did something to the coil in the lower range driver......
What did you do to the coil? Have the test gear to accurately measure these items or kind of wing it?
B.
Last edited:
Huh?
Ben, I have not employed a passive crossover in a speaker project since 1985 or so. I'm the author of the ACD active crossover design program, and write my own LADSPA DSP plug-ins. I guess you didn't notice a bunch of posts in this thread about that kind of thing...
The crossovers for my loudspeaker projects are implemented using IIR filters in software running on lightweight Linux boxes.
My bad, sorry. I must have been confused by the title of this forum ("Conventional loudspeakers with crossovers") and gotten mixed up with another thread.
I hope some readers who do use passive XOs will take to heart your mod'ing experience and my misplaced ribbing about it.
Now, I've been using active XO and bi-amp'ing since 1967, about. But - as we have disagreed - I've found there's nothing to compare to DSP since about 6 yrs ago. With DSP, you just touch a knob and then you instantly can hear the results.
B.
Here is POST 366
What did I write that makes you think I do not like or use DSP? I don't really find anything to that effect when I re-read the post. Also, we have been debating some aspects of DSP, and I mentioned (e.g. in post 366) several of the DSP related things that I have worked on. Then you make some accusations about the passive crossover that I am (not) using... It's all quite baffling.
Last edited:
This thread had been dormant for awhile, but I wanted to post that I have finally completed the first system that follows my "20-20k dipole" paradigm. I just returned from demoing these at a DIY event in Grand Rapids hosted by Meniscus Audio and CSS. The speakers were generally well received, and sounded good in the large conference room listening space where they had room to breathe.
I was so busy during the event that I didn't have time to take a picture of my speakers when they were set up, and I had to quickly take them apart to make room for other speakers.
For now I can provide the frequency response measurement of the system in the attached pic. I made this in my basement listening space before the event, at about 20 degrees off axis, at about 2m, with a long gating window (resulting in about 20Hz resolution) and 1/6th octave smoothing. I also have a distortion measurement that was taken at about 90dB@1m that I will post later when I can find it.
.
I was so busy during the event that I didn't have time to take a picture of my speakers when they were set up, and I had to quickly take them apart to make room for other speakers.
For now I can provide the frequency response measurement of the system in the attached pic. I made this in my basement listening space before the event, at about 20 degrees off axis, at about 2m, with a long gating window (resulting in about 20Hz resolution) and 1/6th octave smoothing. I also have a distortion measurement that was taken at about 90dB@1m that I will post later when I can find it.
.
Attachments
Congrats!
Hello Charlie....
Seems you have most of it figured out. Your frequency response domain looks about as good as it gets. Wish I could have listen to them! Suspect on large orchestral pieces they would be stunning! What woofers did you use? Sorry, I am a bit lazy and did not read all of your previous posts.
Hello Charlie....
Seems you have most of it figured out. Your frequency response domain looks about as good as it gets. Wish I could have listen to them! Suspect on large orchestral pieces they would be stunning! What woofers did you use? Sorry, I am a bit lazy and did not read all of your previous posts.
Here is some info and a couple of pics (will try to find some better ones).
This is a test platform for developing nude open baffle speakers using a variety of drivers. Drivers are freely suspended within the frame using wires or other means. There are several 1/4" OD aluminum rods running front-to-back in the top wood section. This provides a spot to hang the drivers and allows me to easily reposition them in the front-back direction by sliding the loop of wire back and forth along the rod. I do this so that I can physically align the acoustic centers of the drivers.
In this system I have chosen the drivers to be (A) as small in number as possible (e.g. a 3-way system) yet still cover 20-20k Hz as much as possible and (B) to have identical forward and rearward radiation, frequency response, etc. Careful outdoor measurements of individual drivers was used to screen them for this purpose.
This is a fully active system and uses a DSP crossover. The crossover is implemented in software on a laptop using Gstreamer and LADSPA plugins. Only IIR filtering is used. A pro-audio interface (Behringer UMC1820) provides analog and digital input, and 10 analog outputs. Amplification was an Adcom GFA-2535 (4-channels @ 60w/ch into 8R) and an ATI1502 (125Wch into 8R).
The drivers used included two Eminence Alpha 15A in an M-frame, an Eminence Deltalite II 2515, and a back-to-back pair of Dayton ND25FW-4 tweeters. The crossover frequencies are at approximately 250Hz and between 1600-1800Hz.
This is a test platform for developing nude open baffle speakers using a variety of drivers. Drivers are freely suspended within the frame using wires or other means. There are several 1/4" OD aluminum rods running front-to-back in the top wood section. This provides a spot to hang the drivers and allows me to easily reposition them in the front-back direction by sliding the loop of wire back and forth along the rod. I do this so that I can physically align the acoustic centers of the drivers.
In this system I have chosen the drivers to be (A) as small in number as possible (e.g. a 3-way system) yet still cover 20-20k Hz as much as possible and (B) to have identical forward and rearward radiation, frequency response, etc. Careful outdoor measurements of individual drivers was used to screen them for this purpose.
This is a fully active system and uses a DSP crossover. The crossover is implemented in software on a laptop using Gstreamer and LADSPA plugins. Only IIR filtering is used. A pro-audio interface (Behringer UMC1820) provides analog and digital input, and 10 analog outputs. Amplification was an Adcom GFA-2535 (4-channels @ 60w/ch into 8R) and an ATI1502 (125Wch into 8R).
The drivers used included two Eminence Alpha 15A in an M-frame, an Eminence Deltalite II 2515, and a back-to-back pair of Dayton ND25FW-4 tweeters. The crossover frequencies are at approximately 250Hz and between 1600-1800Hz.
Attachments
