Hi all,
I built a fun speaker a while back based on the Mark Audio CHP90 in a bass reflex design using some big oak slab baffles. Very rustic chic. I've been using active DSP on them to handle any issues from the overall baffle changing the response. I recently took them outside to do some measurements to get a better pseudo-anechoic response so that I could export it as an in-baffle FRD and used my DATS to generate a ZMA from in-baffle as well. The goal being to take these measurements of the driver in the cabinet and work on them in VituixCAD to alter the frequency response a bit in a passive crossover, even though its not really crossing over.
I'm attaching my VituixCAD file for anyone that wants to play with it to help out (renamed to *.txt, just remove the .txt and it will be good to go)
Overall I'm just trying to flatten it a bit. Doesn't have to be a heavy natural house curve, but just flattening it would be a nice thing for me. I don't particularly want to make $100 crossovers for this, but I'm just seeing what I can do and what I can tolerate with some basic passive components. Any advice or criticism is appreciated!
I did farfield impulse response gated measurement of the driver, and then did near field of the driver and port and scaled the port's output via area and then summed the near field response. Then I corrected it for the baffle to full space. Final measurement:
As you can see there are a few big spikes I'm trying to potentially tame down, and then maybe lower the upper end in general to flatten it out. Those spikes are around 4200hz, 6400hz and 11800hz.
I used this measurement to generate an in-baffle FRD.
I used this DATS measurement in-baffle for producing the ZMA.
I'm currently working in VituixCAD. A few options, the response on its own and some of my efforts so far.
RAW response in VituixCAD with no components:
Here's one approach I'm looking at:
And then I went overboard with notch filters to hit those 3 spikes. Overall this seems silly and wasteful. Maybe there's a better way? Looking for advice on that.
Any suggestions appreciated!
Very best,
I built a fun speaker a while back based on the Mark Audio CHP90 in a bass reflex design using some big oak slab baffles. Very rustic chic. I've been using active DSP on them to handle any issues from the overall baffle changing the response. I recently took them outside to do some measurements to get a better pseudo-anechoic response so that I could export it as an in-baffle FRD and used my DATS to generate a ZMA from in-baffle as well. The goal being to take these measurements of the driver in the cabinet and work on them in VituixCAD to alter the frequency response a bit in a passive crossover, even though its not really crossing over.
I'm attaching my VituixCAD file for anyone that wants to play with it to help out (renamed to *.txt, just remove the .txt and it will be good to go)
Overall I'm just trying to flatten it a bit. Doesn't have to be a heavy natural house curve, but just flattening it would be a nice thing for me. I don't particularly want to make $100 crossovers for this, but I'm just seeing what I can do and what I can tolerate with some basic passive components. Any advice or criticism is appreciated!
I did farfield impulse response gated measurement of the driver, and then did near field of the driver and port and scaled the port's output via area and then summed the near field response. Then I corrected it for the baffle to full space. Final measurement:
As you can see there are a few big spikes I'm trying to potentially tame down, and then maybe lower the upper end in general to flatten it out. Those spikes are around 4200hz, 6400hz and 11800hz.
I used this measurement to generate an in-baffle FRD.
I used this DATS measurement in-baffle for producing the ZMA.
I'm currently working in VituixCAD. A few options, the response on its own and some of my efforts so far.
RAW response in VituixCAD with no components:
Here's one approach I'm looking at:
And then I went overboard with notch filters to hit those 3 spikes. Overall this seems silly and wasteful. Maybe there's a better way? Looking for advice on that.
Any suggestions appreciated!
Very best,
I have never done a contour filter for a full range driver, but just received a pair of CHP-90s from Madisound today and look forward to the feedback that you get.
One thing I had bookmarked in anticipation is the User Manual for the Copperhead kit from Parts Express. This might be helpful, as it shows a couple different options for the filter.
https://www.parts-express.com/Copperhead-Desktop-Full-Range-Speaker-Kit-Pair-300-7180?quantity=1
One thing I had bookmarked in anticipation is the User Manual for the Copperhead kit from Parts Express. This might be helpful, as it shows a couple different options for the filter.
https://www.parts-express.com/Copperhead-Desktop-Full-Range-Speaker-Kit-Pair-300-7180?quantity=1
Thanks,
The drivers are great. I've enjoyed them a lot, they do vocals so well. I run them active on a DSP to flatten them out and they're great. So now I'm just exploring if I can do a passable job with a filter to go passive on them. Nothing crazy, just enough to flatten it out. The top end is a lot brighter than the bass end, subjectively, as expected from the large baffle (relative). So just bringing that down alone will help. But while at it, maybe shaping it just a bit more. I don't mind some sparkle on top. Just trying to flatten it down more to the bass output level. It sounds great like that and that's all I really need really. Just seeing what's possible beyond that too.
The manual there has notch filter info. I'm using a bit more than that. This also has quite a few more problems to address too.
Very best,
These are so small that they appear to be eigenmodes across the dust cap, so suggest trying the $0.98 tweak or some variation thereof.
Interesting. Though that seems to be for whizzer cones. I will likely try this on my other build that uses a full range driver with a whizzer cone (my BG20 open baffle). But this Mark Audio CHP90 doesn't have a whizzer cone.
Unless there's another way to implement that?
Thanks!
CHP90 with its dust cap:
BG20 with a whizzer cone (as a tweeter here in this open baffle):
Very best,
If by 'contour', I assume you're wanting a frequency contour filter (AKA 'tank' ckt/CD horn EQ), in which case I'd start by simming with the two -6F points
Full rangers really need dsp EQ to shine IME. A full ranger with some very precise parametric EQ filters can sound pretty good, otherwise they're just meh with poor dispersion and breakup modes and resonances all over the upper range. I would personally not even bother trying to fix a full range with passive components.
I could try to do it pasive, but i can't read those files, even with virtuixcad, as they are missing the frd and zma files. If you post these we can try.
But with that much resonances (much more than the Alpair 10.3 or CHN110 that i did equalize passive) i also think you will need dsp to make it relative flat. They are not that strong (luckely), but there are many and that is harder to tame passive than a few strong ones with passive filters.
But with that much resonances (much more than the Alpair 10.3 or CHN110 that i did equalize passive) i also think you will need dsp to make it relative flat. They are not that strong (luckely), but there are many and that is harder to tame passive than a few strong ones with passive filters.
Thanks, 🙏
Doh, I figured the driver's FRD and ZMA would be present in the file saved, like WinISD, etc. I'll attach them (added *.txt to them to attach, just remove that or I think the software reads txt too).
Very best,
I could make it a lot simpler and still got a +/- 5dB flat between 14khz and FS (scake is 3dB increments):
But you need to use serial filters, not parrallel, those can't do this kind of eq i think.
Program i used here is xsim. It's bettter for this job i think (altough i also use virtuixcad to check more complex crossover, i use both).
But you need to use serial filters, not parrallel, those can't do this kind of eq i think.
Program i used here is xsim. It's bettter for this job i think (altough i also use virtuixcad to check more complex crossover, i use both).
Thank you, that looks pretty good to me. While I realize active DSP is easier and better for this, which is what I'm currently doing, the idea of having the option to plop a filter on the back terminals to go passive on any stereo setup is what I'm looking to do. This looks pretty solid. I'll work some on this idea of yours and see what else can be done. Very inexpensive components which is a plus!
Edit: Interesting that Xim doesn't show the spike at 11600hz, it's tame and smoothed out there. I guess that software is smoothing or is there a setting you're using to smooth? When I do this filter in VituixCad I see a lot more like my measurements.
Thank again!
Very best,
Last edited:
You use different values than I in the filter, that's why you need an extra filter. I tried to minimise the components in mine. And a small peak like that won't affect it that much.Thank you, that looks pretty good to me. While I realize active DSP is easier and better for this, which is what I'm currently doing, the idea of having the option to plop a filter on the back terminals to go passive on any stereo setup is what I'm looking to do. This looks pretty solid. I'll work some on this idea of yours and see what else can be done. Very inexpensive components which is a plus!
Edit: Interesting that Xim doesn't show the spike at 11600hz, it's tame and smoothed out there. I guess that software is smoothing or is there a setting you're using to smooth? When I do this filter in VituixCad I see a lot more like my measurements.
View attachment 1389623
Thank again!
Very best,
I changed some values in mine, but when I do an exact copy of your values in Vituix I see a lot more of the original measurement spikes. Xsim seems to smooth it out. Does it have some sort of smoothing enabled in the view?
Here's your values with no smoothing, the 11.6khz spike being obvious. This is why I changed a few things in my other attempts, just seeing how I could potentially notch out some of those bigger spikes.
Very best,
You have DSP already and are trying to go backwards? Everyone wants a nice passive filter they can slap between the amp and drivers and call it a day but it's often at odds with reality and what you want to do isn't going to happen with passive filtering. You already have the tools to filter the driver right, use them.
Trying do this manually is also just kind of a waste of time with DSP. Just let something like room EQ wizard do the work. You can add more bands or remove them based on what you're hearing. With full range drivers I used, which is not many, the very first step I take after building them is to linearize their rather poor response and adjust to taste. We live in the golden age of filtering and it's bonkers to me that people keep wanting to go backwards trying to make things harder on themselves.
Here's predicted response just running automated EQ in REW using the data you shared.
Here's the filter tasks, good luck creating these passively.
Just as a reference point...
In your initial contour in VituixCad, you can't have an Lpad out front that way. Since there is no cap out front of it, the amplifier will see those resistors as the main load running full bandwidth, and they will get hot. That 15 ohm resistor needs to be removed to fix this.
After the series filter, shunt components are okay and safer. If shunt components are out front, they will see a lot more current if not bandwidth limited.
As to using/designing a contour filter on a wide range, start with a simple single 0.1mH coil, and adjust it larger until the FR flattens out as a whole. This accounts for the baffle step, and reduces the inherent shout factor most single drivers have. It usually does not take a very large coil to do this, likely 0.6mH maximum. Sometimes a resistor across the coil is needed to not tank the treble. I would say 2-15 ohms is normal here, depending on the driver. However, this reduces the efficiency of the coil itself and the coil will then need to be larger, but not larger than 2.2mH in most cases. If you have the coil value too large, it will greatly reduce overall sensitivity, and that is what you should try to avoid with this kind of design. Not to mention widerange designs are not known for their bass extension, so don't expect miracles there. If you after that still have the treble rolling off, then place a cap across the LR filter to lift it back up, start with 1uF and increase until the FR levels out as a whole between treble and bass. Adjust resistor to flatten the middle. Then add a notch like @waxx or you did to target the main problem areas.
It I not uncommon to require 2-3 notches and a tilt coil on a single driver design to flatten it out.
One more thing- if this is a small driver, say 4" or less, you may not get full BSC comp in the end. A lot of hindrances come forth. Larger than 3.5" loses off axis response. Smaller than 4" loses bass capabilities. To keep the smaller drivers happy and still get reasonable output, you have some options. You can:
--- Whatever you do, do not use full BSC, use a small sealed box alignment, forget the highpass, and just expect that little driver to handle the bass itself and be happy. Sensitivity is a plus here, and more than 84dB would definitely be preferred in most cases. Below there drivers tend to struggle more at doing the whole job, no matter how you tried to make it work. In a larger room, and/or at longer distances, the driver will likely need to be bigger and/or lot more sensitive just to stay linear, while still applying the above options.
In your initial contour in VituixCad, you can't have an Lpad out front that way. Since there is no cap out front of it, the amplifier will see those resistors as the main load running full bandwidth, and they will get hot. That 15 ohm resistor needs to be removed to fix this.
After the series filter, shunt components are okay and safer. If shunt components are out front, they will see a lot more current if not bandwidth limited.
As to using/designing a contour filter on a wide range, start with a simple single 0.1mH coil, and adjust it larger until the FR flattens out as a whole. This accounts for the baffle step, and reduces the inherent shout factor most single drivers have. It usually does not take a very large coil to do this, likely 0.6mH maximum. Sometimes a resistor across the coil is needed to not tank the treble. I would say 2-15 ohms is normal here, depending on the driver. However, this reduces the efficiency of the coil itself and the coil will then need to be larger, but not larger than 2.2mH in most cases. If you have the coil value too large, it will greatly reduce overall sensitivity, and that is what you should try to avoid with this kind of design. Not to mention widerange designs are not known for their bass extension, so don't expect miracles there. If you after that still have the treble rolling off, then place a cap across the LR filter to lift it back up, start with 1uF and increase until the FR levels out as a whole between treble and bass. Adjust resistor to flatten the middle. Then add a notch like @waxx or you did to target the main problem areas.
It I not uncommon to require 2-3 notches and a tilt coil on a single driver design to flatten it out.
One more thing- if this is a small driver, say 4" or less, you may not get full BSC comp in the end. A lot of hindrances come forth. Larger than 3.5" loses off axis response. Smaller than 4" loses bass capabilities. To keep the smaller drivers happy and still get reasonable output, you have some options. You can:
- port the design and tune to yield a broad humped bass range centered around 100-150Hz with about +4dB at the peak. This will yield some BSC without hurting the sensitivity further. A large cap can then roll it off below tuning if it needs further protection.
- use a TL/Voight or back loaded horn design to amplify the bass range acoustically.
- place close to the front wall to reduce the BSC required for fullness in the bass.
- seal the box AND use a passive or active highpass to hand off to a sub or woofer for the bass.
- seal the box and use a series cap as a passive assist. Box Qtc before cap should be about 1.1, Xmax should be good and long, and 7<Qms<10 for best results.
- allow a gradual rolloff into the bass and keep the sensitivity while accepting the thinner bass.
--- Whatever you do, do not use full BSC, use a small sealed box alignment, forget the highpass, and just expect that little driver to handle the bass itself and be happy. Sensitivity is a plus here, and more than 84dB would definitely be preferred in most cases. Below there drivers tend to struggle more at doing the whole job, no matter how you tried to make it work. In a larger room, and/or at longer distances, the driver will likely need to be bigger and/or lot more sensitive just to stay linear, while still applying the above options.
Thanks, yes, working backwards. Yes, it's less good. Yes, active DSP is better. I'm already doing this actively. I'm just looking to see what can be done, worse, passively. Just an exercise for me. I'm not trying to do passive better. Learning experience for me for contour filters like this as I have no done this. New to me. So just experimenting and seeing what I can do.
Very best,
Thank you so much for all that, very interesting. So many things I'm unaware of with how these filters work and alter things beyond what I'm seeing in the software, where I don't get what I see, per what a few of you fine folk have mentioned.
Looks like I need to get a few components and do some comparison testing to see how it will work out real world vs simulated here.
Very best,
That's why i made that xsim for you, it's showing how it's done easier to filter. And i would use this driver in combo with a woofer so you can use higher sensivity and volume. This driver is not made to go very low. That's why i used in my own bookshelf the 10.3 (NLA) with a 5.5" cone and way better damped resonances, so it's easier to tame them passive.
That CHP90 will probally be use full in your way to about 90dB, not much louder. So only as nearfield. When you cross it higher it can reach above 100dB without distorting, but it will need a woofer below 200Hz or so for that.
But i also think your box design is a bit flawed, looking at the bass response. I tried to get the bet out of it, but a better box would probally the best solution. Now the bass is limited because you tune it below FS and this driver is not fit for that. You should tune it to 55Hz idealy, 50 would probally also be possible, but not lower like you do.
That CHP90 will probally be use full in your way to about 90dB, not much louder. So only as nearfield. When you cross it higher it can reach above 100dB without distorting, but it will need a woofer below 200Hz or so for that.
But i also think your box design is a bit flawed, looking at the bass response. I tried to get the bet out of it, but a better box would probally the best solution. Now the bass is limited because you tune it below FS and this driver is not fit for that. You should tune it to 55Hz idealy, 50 would probally also be possible, but not lower like you do.