Hi, I am currently building a pair of desktop speakers using some drivers from Dayton : TCP115-4 and PS95-8.
Since I want to cross them low I need quite" huge" inductors to go 2nd order passive crossovers, I was thinking the best way would be bi-amping. I've already build boards with TDA7375 ( from an old Blaupunkt car stereo, I think most likely they are original, and powerfull enough for desktop use ).
Now for " active crossover " would a 3db passive RC style at the amp input work ? ( a simple 47uF capacitor at the amp output on the PS95's seems to do the job just fine) but i duno for the woofers.
I might build Rod Elliott's project 123, https://sound-au.com/project123.htm , what about the phase shift? I need to invert the ps95's polarity ?.
What x-over freq u guys recommend ? . I was thinking around 250hz. Pleasse take a look at the drivers as well.
Thank you, Bruno.
Since I want to cross them low I need quite" huge" inductors to go 2nd order passive crossovers, I was thinking the best way would be bi-amping. I've already build boards with TDA7375 ( from an old Blaupunkt car stereo, I think most likely they are original, and powerfull enough for desktop use ).
Now for " active crossover " would a 3db passive RC style at the amp input work ? ( a simple 47uF capacitor at the amp output on the PS95's seems to do the job just fine) but i duno for the woofers.
I might build Rod Elliott's project 123, https://sound-au.com/project123.htm , what about the phase shift? I need to invert the ps95's polarity ?.
What x-over freq u guys recommend ? . I was thinking around 250hz. Pleasse take a look at the drivers as well.
Thank you, Bruno.
Attachments
It's nice to see some DIY active XOs for a change!
The trouble with passive line level filters is that it messes with the impedance between the source and the amplifier, which can add noise and distortion. Hence, op-amp based buffers can help.
I would do a hybrid: a simple air-cored inductor + 2nd order active filters for a total of 3rd order, so you would need to drop C4 and R5. That way, at mid-range to high frequencies, the inductor presents a high impedance in series with the TCP115 and helps to reduce distortion from inductance modulation. The output will fade with increasing frequency anyway, but each octave only gives a nominal 18dB of attenuation.
Is the TCP115 locked in? Its 50cm^2 cone area seems only a little bigger than the PS95.
The trouble with passive line level filters is that it messes with the impedance between the source and the amplifier, which can add noise and distortion. Hence, op-amp based buffers can help.
I would do a hybrid: a simple air-cored inductor + 2nd order active filters for a total of 3rd order, so you would need to drop C4 and R5. That way, at mid-range to high frequencies, the inductor presents a high impedance in series with the TCP115 and helps to reduce distortion from inductance modulation. The output will fade with increasing frequency anyway, but each octave only gives a nominal 18dB of attenuation.
Is the TCP115 locked in? Its 50cm^2 cone area seems only a little bigger than the PS95.
@abstract Already bought the drivers, yes the tcp wasn't the best choice here, a 5.25" would be better, but I thought it's high excursion would help bass a lot, they are ok. I didn't want to build huge cabinets.
I have mixed a song to mono, duplicated it and sent it to both channels , one to left one to right, and applied high pass and low pass in audacity, woofer on L and mid-high on R, the ps95 can work happy down to even 150hz, weird thing, free air they sound best with 250hz on the ps95's and 600-800 on the woofers, maybe because it's free air and the ps95's lack those frequencies ? need to test them in box.
Also tried in FL Studio , I can do it real time (high pass, low pass) polarity, I observed how it changes based on frequencies, and how they are out of phase on certain frequency range, fun stuff!.
I have mixed a song to mono, duplicated it and sent it to both channels , one to left one to right, and applied high pass and low pass in audacity, woofer on L and mid-high on R, the ps95 can work happy down to even 150hz, weird thing, free air they sound best with 250hz on the ps95's and 600-800 on the woofers, maybe because it's free air and the ps95's lack those frequencies ? need to test them in box.
Also tried in FL Studio , I can do it real time (high pass, low pass) polarity, I observed how it changes based on frequencies, and how they are out of phase on certain frequency range, fun stuff!.
Not finished yet, I need the paint to fully cure , sand , and paint again then wet sand and polish them .
Already tried on a subwoofer box I painted red a while back, sanded (400 to 3000 grit )a
and polished to see how it works( by hand with a cloth on the red box) Got a kit for polishing, haven't tried it yet but I think it would work wonders if the results from " hand polishing" were good enough.
Already tried on a subwoofer box I painted red a while back, sanded (400 to 3000 grit )a
and polished to see how it works( by hand with a cloth on the red box) Got a kit for polishing, haven't tried it yet but I think it would work wonders if the results from " hand polishing" were good enough.
Attachments
Seems reasonable.......being 'very old school', the way I figure it is by matching polar responses, so FWIW, etc.:
TCP's fVC = ~34400/pi/2.54 = ~4311 Hz
PS's fVC = ~34400/pi/2 = ~5475 Hz
TCP's 1st approx XO: ~sqrt(53.8*~4311) = ~482 Hz
2nd = sqrt(482*4311) = ~1441 Hz, etc.
PS's = sqrt(119.8*5475) = ~810 Hz
2nd = sqrt(810*5475) = ~2106 Hz, etc.
1st mean = sqrt(482*810) = ~625 Hz
2nd = sqrt(1441*2106)~1742 Hz
While most would probably go higher, which brings the gap closer together with increasing XO point; based on your listening choices it seems the ~625 Hz is a good comprise.
Otherwise, KISS method 😉 = sqrt(250*800) = ~447 Hz
Common rule of thumb can work.
Virtuix Cad has extensive design calculators
For OPA active filters.
You could simulate the baffle /drivers in full space
with the driver FRD and create a reasonably
accurate filter network.
And observe behavior of phase as well
to sum correctly
Virtuix Cad has extensive design calculators
For OPA active filters.
You could simulate the baffle /drivers in full space
with the driver FRD and create a reasonably
accurate filter network.
And observe behavior of phase as well
to sum correctly
Crossovers and phase go hand in hand as rightly pointed out already. What you dont have to contest with in active xovers is speaker rising impedence assuming amp has high damping factor. Dear i say one less factor for active xovers to correct for. When is a speaker manufacturer going to invent a driver with no inductance, no reasonance and be perfectly resistive linear. Also no beakup nodes and no distortion.
IMO inductance modulation is the elephant in the room that active designs often struggle with overlook altogether. A 1st order passive low-pass is a quick fix (the series impedance helps keep the current proportional to the amplifier input voltage, in spite of non-linearity of the speaker).
2nd order passive filters lose this benefit because of the capacitor added in parallel. It short-circuits the speaker at high frequencies, forcing current to flow, causing problems (and maybe even fuel myths about capacitors having peculiar sounds, when there's some real physics going on).
Some of the First Watt amplifiers (especially F2 with its 80 ohm output resistance? IIRC) could be a good match to a well-made active XO. That way the output impedance is built into the amplifier, and the active XO can be designed from the ground up to EQ the response.
2nd order passive filters lose this benefit because of the capacitor added in parallel. It short-circuits the speaker at high frequencies, forcing current to flow, causing problems (and maybe even fuel myths about capacitors having peculiar sounds, when there's some real physics going on).
Some of the First Watt amplifiers (especially F2 with its 80 ohm output resistance? IIRC) could be a good match to a well-made active XO. That way the output impedance is built into the amplifier, and the active XO can be designed from the ground up to EQ the response.
Attachments
The source being the amplifier or the speaker? If looking from the amplifier, then yes, but if looking at spurious voltages generated on the speaker side, then that would reduce the impedance seen from that side. (But please double check as I'm often wrong).
Last edited:
The source impedance could be described as the impedance the driver sees looking out at it's terminals. In other words if there was no crossover then it would only see the short circuit of the amplifier, so any finite impedance you see in that plot is the result of the crossover components.
You are correct that the capacitor shorts this, but as can be seen comparing the red/blue (first/second order) it happens above resonance, when the response is already down.
You are correct that the capacitor shorts this, but as can be seen comparing the red/blue (first/second order) it happens above resonance, when the response is already down.
Said it before and will say it again: without in-box measurements of the individual speakers, filter design is only shooting in the dark. This will lead to sub-optimal (best case) to way off (worst case) results. Textbook filters will not do the job properly because they -mostly- do not deal with in-box driver behaviour. Furthermore, relative acoustic centres are not taken into consideration.
Start here:http://audio.claub.net/software/FRD_Blender/White Paper - Accurate In-Room Frequency Response to 10Hz.pdf
Start here:http://audio.claub.net/software/FRD_Blender/White Paper - Accurate In-Room Frequency Response to 10Hz.pdf
And what exactly is the reason, being on a budget, to not consider NPE capacitors?
DIYers ought to be smarter than propaganda developers, I should think. 🙂
Testing a particular configuration to "sound best" is something elso than rationally designing a system in a step-by-step fashion.
I've found that polypropylene sounds much better than polyester in Sallen-Key filters. The different capacitor types have a thing called a "dielectric constant" and generally, the higher it is, the more prone they are to distortion.
You could call it a "studio monitor", and explain that the worse it sounds, the more 'revealing' it is of all the flaws everywhere else in the system and production chain (but not here).
You as user define/set the optimization targets. If you set the target slopes and x/o points wrongly, the results will be equally wrong.
Furthermore, I yet have to encounter a loudspeaker system that sounds bad, but measures good.