I'm modifying a pair of ATC SCM20 Pro PSL Mk2 monitors - making them active and lowering the crossover point as low as realistically possible. The included crossover point is advertised to be 2.1khz, but 1.975khz seems to be what it actually is.
I did a frequency sweep of the woofer at 0 and 45 degrees. The woofer's -3db point at 45 degrees is 2150hz. Ignoring (for now) the potential for harmonic distortion from the tweeter, and aiming for a design with the best on axis frequency response and also the widest horizontal dispersion, what should the crossover point be? With the 10 inch baffle, the tweeter's on axis response will experience a suckout starting at ~1400hz, increasing in severity until it's down 6db at ~700hz (baffle step), so it would seem like a 4th order at 1.6-1.7khz would be needed to have the signal down 6-8db by the beginning of the suckout, but I'm probably making a new box with a wider baffle to move the suckout to 1000hz - 1.2-1.4khz is my target range (the bigger box will also allow for a lower q alignment (sealed design).
I know that harmonic distortion can be an issue for some tweeters, but it very likely won't be for this tweeter - I'll be doing tests later to confirm the lowest viable crossover frequency at the power levels I'll be running. Because the tweeter is essentially omnidirectional at any considered crossover frequency, it's just the woofer that needs attention.
For an excellent power response and the associated optimal horizontal dispersion, how many db down at 90 degrees can the driver doing the lower frequencies be? 3db? 6db?
If a driver is 3db down at 45 degrees, can you halve the frequency and extrapolate that the -3db point then doubled to 90 degrees? (This woofer at 1075hz???)
Is there a rule for how many db down at 90 degrees at crossover a driver can be for a speaker with good power response?
I'll be considering this frequency along with others yet to be determined/measured/calculated for the crossover, picking the lowest that allows for widest flattest frequency response, with acceptable distortion
I did a frequency sweep of the woofer at 0 and 45 degrees. The woofer's -3db point at 45 degrees is 2150hz. Ignoring (for now) the potential for harmonic distortion from the tweeter, and aiming for a design with the best on axis frequency response and also the widest horizontal dispersion, what should the crossover point be? With the 10 inch baffle, the tweeter's on axis response will experience a suckout starting at ~1400hz, increasing in severity until it's down 6db at ~700hz (baffle step), so it would seem like a 4th order at 1.6-1.7khz would be needed to have the signal down 6-8db by the beginning of the suckout, but I'm probably making a new box with a wider baffle to move the suckout to 1000hz - 1.2-1.4khz is my target range (the bigger box will also allow for a lower q alignment (sealed design).
I know that harmonic distortion can be an issue for some tweeters, but it very likely won't be for this tweeter - I'll be doing tests later to confirm the lowest viable crossover frequency at the power levels I'll be running. Because the tweeter is essentially omnidirectional at any considered crossover frequency, it's just the woofer that needs attention.
For an excellent power response and the associated optimal horizontal dispersion, how many db down at 90 degrees can the driver doing the lower frequencies be? 3db? 6db?
If a driver is 3db down at 45 degrees, can you halve the frequency and extrapolate that the -3db point then doubled to 90 degrees? (This woofer at 1075hz???)
Is there a rule for how many db down at 90 degrees at crossover a driver can be for a speaker with good power response?
I'll be considering this frequency along with others yet to be determined/measured/calculated for the crossover, picking the lowest that allows for widest flattest frequency response, with acceptable distortion
Last edited:
Why not measure and find out. The nature of lobing per frequency makes a two point measurement less certain.
What is the suckout you refer to?
What is the suckout you refer to?
I'm not an experienced diyer, though have refurbished speakers filters already. Often to make them new again (loudspeaker crossover drift causse lythic after a decade, it is slow but can change a good design in something very less good faster than one believes.
The big force of your system is the mid-woofer. If it was for me, I do it that way and AllenB who is filter guru labeled will correct me as often btw if I mistake :
- I will measure the T&S of the raw drivers, if they are more than 20 years will look if the tweeter is ferrofluided. Measure the lythic caps if any saying myself they could have drifted.
- then according the T&S I will simulate if a sealed box is good enough, they are most of the time because a better delay group and a smooth response towards the end, look at the -F10 vs a vented. But of course sealed is also meaning distorsion more easily.
- I will measure the distorsions of the raw drivers to decide on the filter cut off and order, having in mind the guts at ATC certainly did a goo joob (is it a 18 dB à la Proac on thye negative polarity of both drivers for the signal?)
- then I will adapt the filter from the listening position as you talk about active system with a delay of the tweeter according the filter choice for a better transcient and soundstaged system.
Be aware than cheap adc/dac active system have also trade off on sounding quality but there is certainly not better than a bespoke loudspeaker with the room at listening position.
The big force of your system is the mid-woofer. If it was for me, I do it that way and AllenB who is filter guru labeled will correct me as often btw if I mistake :
- I will measure the T&S of the raw drivers, if they are more than 20 years will look if the tweeter is ferrofluided. Measure the lythic caps if any saying myself they could have drifted.
- then according the T&S I will simulate if a sealed box is good enough, they are most of the time because a better delay group and a smooth response towards the end, look at the -F10 vs a vented. But of course sealed is also meaning distorsion more easily.
- I will measure the distorsions of the raw drivers to decide on the filter cut off and order, having in mind the guts at ATC certainly did a goo joob (is it a 18 dB à la Proac on thye negative polarity of both drivers for the signal?)
- then I will adapt the filter from the listening position as you talk about active system with a delay of the tweeter according the filter choice for a better transcient and soundstaged system.
Be aware than cheap adc/dac active system have also trade off on sounding quality but there is certainly not better than a bespoke loudspeaker with the room at listening position.
Last edited:
I'd like to applaud for aiming at good power response! Way to go!
In order to get a grip on power response, you'll need to measure it. This means you'll have to measure the anechoic response of the speaker at many different angles, both horizontally and vertically. That's a lot of work, but it's worth it!
The best approach is to measure all drivers separately (just the drivers without any xovers). Measure their anechoic SPL response curves at horizontal and vertical angles of -180°, -170°, ... -10°, 0°, +10°, +20°, ... , +180°. Then import these data to a suitable software package that allows you to model the xover filters and to calculate the power response of the entire loudspeaker system. Vituix CAD is a good choice, and it comes with very detailed step-by-step instructions (but it does not describe the theoretical background very much).
Here's some more stuff on power response measurement:
https://www.princeton.edu/3D3A/Publications/Tylka_3D3A_DICalculation.pdf
GitHub - mbrennwa/mat_lspr: m-file to calculate power response of a loudspeaker from sound-pressure measurements arranged along two orbits around the speaker.
In order to get a grip on power response, you'll need to measure it. This means you'll have to measure the anechoic response of the speaker at many different angles, both horizontally and vertically. That's a lot of work, but it's worth it!
The best approach is to measure all drivers separately (just the drivers without any xovers). Measure their anechoic SPL response curves at horizontal and vertical angles of -180°, -170°, ... -10°, 0°, +10°, +20°, ... , +180°. Then import these data to a suitable software package that allows you to model the xover filters and to calculate the power response of the entire loudspeaker system. Vituix CAD is a good choice, and it comes with very detailed step-by-step instructions (but it does not describe the theoretical background very much).
Here's some more stuff on power response measurement:
https://www.princeton.edu/3D3A/Publications/Tylka_3D3A_DICalculation.pdf
GitHub - mbrennwa/mat_lspr: m-file to calculate power response of a loudspeaker from sound-pressure measurements arranged along two orbits around the speaker.
I plan to, it's just that weather and my schedule have conflicted over the past while, and I'm getting impatient strategizing. When I begin measuring, it would be nice to have an idea of what I'll be seeing. In addition to satisfying my curiosity in the mean time, I'll know my measurements are close to what's expected when I start doing them, and if I have things set up reasonably well. I want to measure distortion, frequency response, and off axis response of both the woofer and tweeter for my project and to put online for others. The woofer sounds great and is a topic of discussion, so it'd be nice to put numbers to its reputation (like the 3" ATC midrange dome). The tweeter is one of two ATC makes in house. It's the "S-Spec" version (used from this speaker up to the SCM150, sans SCM40). It is different from the other tweeter in probably more than a few ways, but what's listed is its 33% stronger magnet, upper -2db and -6db point 5khz higher, and 3.5db/w more efficient. I think to pull this off it has a slightly narrower magnetic gap and a lighter moving mass, but these details aren't available in the datasheet (almost nothing is...). I think it has a higher resonance, probably 200hz higher. If I measure the tweeter, someone else may measure the other version in the future, and the differences could be known.
About the suckout, in another thread I made a few weeks ago, someone brought up that by lowering the tweeter's crossover point, on axis response could be affected. He said he put a 1 inch dome on a 10 inch baffle in a simulator and that it showed an on axis a suckout began at 1400hz. It makes sense to me because the wavelength of 1400hz is just under 10 inches, so I assume that because the tweeter radiates omnidirectionally at that frequency and below, the edges of the wave would start wrapping around the cabinet and reflecting outwards, stealing energy that would otherwise be directed forward. I don't know by how much, or how sound would be affected, but I plan to measure the tweeter's on axis response in the default enclosure, then temporarily widen the baffle with wings and measure it again to see how much loss the baffle causes
Thanks for the advice!
This speaker is quite new, manufactured in late 2018 if my memory of the serial number card is correct. I bought them about a year ago.
I haven't examined the crossover in detail yet. It's a very large board at the back of the enclosure. Everything is separated and generously sized. No iron core and no electrolytics. I was told by the dealer that the crossover is a 3rd order butterworth - my favourite. I haven't measured it electrically, just acoustically (and not with a corrected mic), but it does seem to have an 18db/oct roll off on the tweeter, and the tweeter is flat through where I measured that.
I know that a cheap DSP would be horrible. I have a 4 channel interface (RME Babyface Pro) with ~113db SNR on 2 channels and 108 on the other 2. I'll be using DSP on PC, sending it throug the interface to a 140wpc RMS class a amp for the tweeter (Kinergetics KBA-280), and 140wpc RMS class ab amp for the woofer (Arcam A32). I'm probably going to make a passive preamp to get the best performance I can at lower levels.
This is an older version unfortunately. The crossover is different, the tweeter is different, and I think the woofer has had a couple incremental improvements since. I know my cabinet is different inside too.
I'll read these this afternoon, thank you. I was planning on doing measurements outside for minimal reflected sound, and just to 90 degrees (I have no chamber lol). I'll see how these links may change that plan
Ah, the second link is a file to put measurements over two axis into and get power response. Very nice.
If I do the measurements outside, would it be best to point the speaker at the sky for vertical?
If I do the measurements outside, would it be best to point the speaker at the sky for vertical?
Please clarify;
"This" refers to the speakers within the link that I posted << or >> "this" refers to your speakers ??
🙂
You will rotate the speaker around it's horizontal or vertical axis, keeping the microphone at a fixed position. Also, you'll need anechoic measurements, so you'll likely want gated measurements to remove echoes. I don't think pointing the speaker at the sky would not be very practical.
This may be a storm in a teacup. Some energy can get redirected at the edge. Maybe that lines up for a dip in response on some axis. It probably averages out over the wider range of angles, so if you try to "fix" it you could create problems elsewhere.
Better to round your edges if it is a problem. If not, and if unsure about a diffraction problem then it is probably better to ignore the issue until you get more information.
I don't think that diffraction is as bad as some people say, and I don't understand why people think that the typical 1/2" "rounded edges" will solve diffraction problems.
Note that the radius of the "rounded edge" would need to be similar to (or slightly less than) the wavelength of the frequency where the diffraction "issues" are observed. In this case, the frequency is 1.4 kHz, which corresponds to a wavelength of 0.25 m. The radius of the "rounded edge" would need to be at least 10 cm to have an effect.
Sigfried Linkwitz presented some nice data and explanations on diffraction: Diffraction from baffle edges
Finally, remember that if some frequency shows a diffraction-related dip in the SPL curve observed at some angle, there will be other angles that will show a SPL peak at the same frequency. This will compensate to an overall flat power response. In other words, the energy dissipated from the speaker into the listening room at any given frequency is not affected by diffraction.
Last edited:
So it's very important you measure the capacitors and coils out of their filter (raw) as all is new enough to have not drifted yet. Good loudspeakers have often their capacitors with a precise capacitance value which is not the one on the case marking (+ there is a tolerance) and sometime a matching with left and right speaker. Plus the measure of the drivers at impedance cut off, it will give you the exact crossover typology by simulation and the possibility to sell your speaker in the futur for a new project. It's also a good basis to work on for the actual project with your sealed new enclosure 😎
Yeah I'm thinking measure it first to see just how much a problem it is, and if it's enough of a problem to even be a problem
Good point with the rounded corners. I do think smaller rounded edges have an effect on lower frequencies though, just not a maximal effect.
I didn't think baffle interaction would be a problem myself, but when the issue was brought up I thought about it a bit - woofers of most sizes are becoming directional by 1400hz, minimizing baffle interaction at the frequency, while tweeters are completely non directional at 1400hz, maximizing interaction. The effect could be small and sonically benign, or small and offensive. Or large and unobtrusive. It seems measuring and listening will be the only way to determine, because dome tweeters are rarely crossed over so low
I get your point about the power response being unaffected by baffle diffraction, but it still could affect on axis listening as the character of the direct sound, used to determine sonic character, would be different.
I think I will, and I'll write the values down to store them for the future if they need repair. Or to include with the speakers if I sell them so they can be repaired properly by the new owners. I like them a lot though, so I'll probably keep them til I die. Or they rot, whichever comes first.
It sucks when crossovers go and wreck the performance of a speaker... It happened to me once before and I couldn't fix it because one component was weird - I think it was supposed to be a resistor that changed with temperature (which increased with power) to prevent cooking. Pretty sure the default resistance changed and it made the midrange permanently quiet. I didn't even overdrive them! The thing was in a horrible spot of the glued-up mess of a crossover, so I couldn't measure the other one without risking breaking it - I couldn't even find it's legs to measure in circuit. Very annoying. Infinity SM-112 - a lively 3 way, extremely dynamic and efficient. It had a recessed mid-bass between 120 and 220hz, and they could have imaged better, but I liked them.
Back to these speakers - for now this project is active, but in the future, say if I need to use my hardware differently, I may make a passive crossover
Last edited:
Moving the microphone with the speaker facing the sky would be complicated I guess lol. Do you have any tips on rotating the speaker on its vertical axis for a stationary mic?
I think taking the measurements carefully outside is good enough for my purpose - the extra work and hardware required for gated measurements don't seem worthwhile because I'll be crossing the tweeter as low as I can, maintaining low distortion and flat horizontal dispersion to 90 degrees (until tweeter directionality of course).
Edit: my goal is to have the best power response possible, and even half space radiation from the drivers at frequencies above the baffle step
Last edited:
Just lay the speaker on it's side and repeat what you did for the horizontal measurement.
Think again! The echoes from the floor will mess up the acoustic measurements throughout the entire frequency band.
Hardware: you'll need a microphone and a computer with a sound card. I am pretty sure you already have that.
Software: there are many good software packages out there that will do the gating for anechoic data. Some of them are free (as in free beer), and some are even free (as in free speech).
I'd think a suitable software tool will not only allow you to get cleaner data, but also to get the job done quicker and easier. For your purpose, Arta and Vituix CAD might be a good couple. (I would use MATAA for everything, but that's just me)
Well it seems obvious when you put it that way haha
I'll have to think about it. At this point I'm thinking I'll probably do it when I'm finished to see what the exact power response is. I have a feeling the tweeter's maximum linear excursion will ultimately dictate the minimum crossover point (and how good the power response can be). If my off axis sweeps of both drivers from 500hz to 2khz end up showing issues with the tweeter near where I plan to cross (1.2-1.4khz), I might end up doing it sooner
I was being told by someone that there would be a suckout at 1400hz, but this baffle simulator, configured to the SCM20's dimensions, shows a boost of 2db at about the same frequency. Was I advised improperly?
Tweeter:
And the Woofer:
Ignoring for now the 6db transition from half to full space radiation:
The woofer comes to a 2db peak at 800hz, which is slowly risen to from ~480hz. This returns to 0db by 1370hz. Above is effectively flat to infinity.
The tweeter rises 1db from 660hz to 1khz. From 1khz to 1.375khz it rises further to 2.5db. Then it falls to 0db by 1.8khz. The response remains flat to infinity, other than a very narrow 1.2db dip centered at 2.7khz.
This baffle seems benign to me. Things could even be improved a bit by moving the tweeter 2 inches to one side - it'd get rid of the 1.2db dip at 2.7khz. If I don't make a new box, I won't do that though.
Listening to the thing 20 degrees off axis horizontally (how I would have the speakers pointing anyway) would lower the 2.5db peak centered at 1.375khz to just over 1db, and the woofer's 2db peak to 1.5.
Am I doing anything wrong? Does this spreadsheet seem to be doing an accurate job?
Tweeter:
And the Woofer:
Ignoring for now the 6db transition from half to full space radiation:
The woofer comes to a 2db peak at 800hz, which is slowly risen to from ~480hz. This returns to 0db by 1370hz. Above is effectively flat to infinity.
The tweeter rises 1db from 660hz to 1khz. From 1khz to 1.375khz it rises further to 2.5db. Then it falls to 0db by 1.8khz. The response remains flat to infinity, other than a very narrow 1.2db dip centered at 2.7khz.
This baffle seems benign to me. Things could even be improved a bit by moving the tweeter 2 inches to one side - it'd get rid of the 1.2db dip at 2.7khz. If I don't make a new box, I won't do that though.
Listening to the thing 20 degrees off axis horizontally (how I would have the speakers pointing anyway) would lower the 2.5db peak centered at 1.375khz to just over 1db, and the woofer's 2db peak to 1.5.
Am I doing anything wrong? Does this spreadsheet seem to be doing an accurate job?