Converted broken KEF LSX to passive speakers

[1096bimu]

I recently picked up a pair of broken KEF LSX speakers and also a WiiM amp for cheap and decided I was going to try and put these together to replace the KEF LSX II I have been using for a long time as computer speakers.
I originally picked the LSX II because it was a good deal on facebook marketplace, but it's also perfect for this purpose because it is small, coaxial, high performance and looks amazing. But I also know it is worth a lot of money so I thought if I could recreate the same feature set with the cheap parts I had, I can sell the more expensive LSXII.
I have quite a bit of general knowledge and know-how but have not dabbled in speaker engineering, but somehow I got lucky and everything worked out amazingly well, I think the passive set I got now is even better than the original product.

Firstly I know personally the first generation KEF LSX speakers were unreliable, I had a set of those as well, it didn't die but it would occasionally disconnect from the wifi and require a manual reboot (if you want to use any of its smart features like remove via internet, or streaming, obviously). The broken set I got, once I opened it up, it was obvious a part of the main board literally caught on fire, and burnt surrounding parts into crisp, complete with a distinct burning plastic smell.

I learned the method to open this up on a post here but I'll quickly repeat the tips, the back panel surrounding the I/O part, is a single piece of plastic, held in by friction alone, you can pry it open. Then 8 screws along the edges will be available that hold the front face onto the cabin. However I can't offer any advice on fixing one of these because I couldn't even figure out how to remove the main board without damaging stuff. It looks like the main board is super stuck behind some big components on the power supply board, but the screw that holds the power supply board is almost covered by the main board, so I don't know how this is supposed to be done, so I put on some thick gloves and just used brute force.

After that I ordered some generic crossover circuit on aliexpress, I know that these are supposed to do, but I had no idea what I needed exactly, so I just basically tried my luck, and I somehow got lucky. The board came with some configuration jumpers, and I know the configuration I ended up with basically, a 4.7uf capacitor in series with the tweeter, and straight connection for the woofer. This woofer on the Uni-Q driver is very wide band, it's good up to 4 or even 8khz, the tweeter is only good for super high frequency so that just turned out to work really well.

I hijacked the power plug on the speaker for signal input, I connected neutral to positive and ground to negative, I took the original power cables, cut off the wall plug and just used that wire as speaker wire to connect to the WiiM Amp. Now it looks even cleaner than before because each speaker only has one power cable coming out, no more ethernet cable to connect them! I ran some tests for just the speaker response and other than the bass region, it honestly looks very similar to the review of the original thing posted here, with my working LSX II as comparison I also thought they sounded about the same.

It also worked amazingly well, better than original I would say. Firstly it just has way lower noise floor with this WiiM amp thing, I wouldn't say the KEF LSX noise floor is annoying, but it can become slightly noticeable sometimes, since I have an ultra quiet PC build. Now with the passive conversion, the noise floor is totally impossible to hear without being so close you turn the speaker into a headphone, and yes I know there's like cutoff tricks or wahtever, I also tried this with extremely low volume signal playing, it really is just a lot cleaner than before.
Secondly being a smart amp, it has way more powerful DSP features than the stock LSX, the Roomfit thing it came with, just works super well, it does a better job than I'm able to do manually, maybe it's a skill issue but now I think I've got a result more balanced than I've ever achieved previously.
The amp also has better subwoofer support, so with my sub connected, settings tuned, room equalized and even delay synchronized, I couldn't be more happy with the results. I also like small features like having a volume knob, volume display, more responsive app-based control, and volume limiter so I don't accidentally start shaking the floor with the connected sub.

 

[staticV3]

This woofer on the Uni-Q driver is very wide band, it's good up to 4 or even 8khz

For distortion yes, but for smooth directivity you want to cross at ~2kHz: https://www.erinsaudiocorner.com/driveunits/kef-ls50-drive-unit/
as KEF are doing.

 

[1096bimu]

For distortion yes, but for smooth directivity you want to cross at ~2kHz: https://www.erinsaudiocorner.com/driveunits/kef-ls50-drive-unit/
as KEF are doing.

I did try that before closing the enclosure and it didn’t sound right so I left it full range, I will try that again with full assembly then, thanks.
Since I only sit straight facing the speakers it didn’t make that much of a difference but now this reminds me how quickly the treble fell off off-axis

EDIT
I re-attached the inductor for the woofer, there is an expected drop around 3khz but that is easily corrected with EQ and yea it certainly looks better
red: full range woofer
aqua: woofer with low-pass
blue: slapping on a simple 3k boost and 16k drop with EQ

 

[Arindal]

After that I ordered some generic crossover circuit on aliexpress, I know that these are supposed to do, but I had no idea what I needed exactly, so I just basically tried my luck, and I somehow got lucky. The board came with some configuration jumpers, and I know the configuration I ended up with basically, a 4.7uf capacitor in series with the tweeter, and straight connection for the woofer. This woofer on the Uni-Q driver is very wide band, it's good up to 4 or even 8khz, the tweeter is only good for super high frequency so that just turned out to work really well.

Do you happen to have done any measurements?

Have never been doing anything with this (KEF-proprietary) coaxial driver, but I was of the impression that such tiny fullrange coaxials in a tiny enclosure usually require a pretty complicated crossover, as woofer and tweeter are of different sensitivity, tweeter requiring rather low crossover freq but not too low, with the woofer potentially playing pretty high in range which can lead to a lot of funny effects, but usually comes with a lack of bass and overshoot of treble. I mean, there is a reason why this thing has DSP and there is no passive equivalent on the market, isn´t it? It looks appealing so that would be an easy win for KEF if easily doable.

 

[1096bimu]

Do you happen to have done any measurements?

Have never been doing anything with this (KEF-proprietary) coaxial driver, but I was of the impression that such tiny fullrange coaxials in a tiny enclosure usually require a pretty complicated crossover, as woofer and tweeter are of different sensitivity, tweeter requiring rather low crossover freq but not too low, with the woofer potentially playing pretty high in range which can lead to a lot of funny effects, but usually comes with a lack of bass and overshoot of treble. I mean, there is a reason why this thing has DSP and there is no passive equivalent on the market, isn´t it? It looks appealing so that would be an easy win for KEF if easily doable.

 

[AnalogSteph]

For starters, you need better measurements. This looks like single point, at too large a distance, under less than ideal acoustic conditions, with little to no windowing to speak of. (And you should probably make it a habit to document these things.) You've got a coax, so don't be afraid to get closer. What's the worst that can happen, some proximity effect on the bottom end, some mic distortion and perhaps some interference of the mic body? If you can place the speaker on a stand in the middle of a large room, all the better. For a good idea of what the speaker is doing in general, take some moving microphone measurements - while you don't get any timing information, the spatial averaging irons out the raggedness for virtually dB-level accuracy.

Taking and Interpreting Measurements with REW (FREE eBook)

A few years ago, Amir wrote Room Measurement Tutorial for Dummies Part 1 and Part 2. The third part has never materialized despite many requests. In the meantime, there is a clear need for clear guide on how to take room measurements because we are seeing more and more people take and post their...

www.audiosciencereview.com

You also want to be able to take raw measurements of tweeter and woofer separately. (Start tweeter sweep at 1 kHz or so.) This will give you a much clearer picture of which driver is contributing what and what a crossover should be acting like. The woofer is a metal membrane affair so I'd expect some rather nasty breakup modes that may need addressing via a notch filter. (Rule of thumb: They should end up 20 dB below tweeter level.) Tweeter fundamental resonance may also need some extra attention - when combined with a single series capacitor, it can make the crossover slope very poor and negatively impact level handling. (One more thing to be measured.)

If you can also take some electrical / TS parameter measurements alongside acoustic measurements at various angles, you'd have enough data to give simulation in VituixCAD a try.

 

[Ktacos]

Your data shows some pretty major crossover issues. Your implementation is simply not working right at all. Definitely need to analyze the speakers with a particular process and design the filters in software, anything else is unfortunately just guessing and your in room on axis response is not enough to tell whether drivers are being integrated properly.

As another mentioned, your window scaling is misleading, a range of about 50db is what most people use.

If we import your passive response into vituixcad, we can see that your tweeters response is sloped way too far up towards the top end, and its fighting the woofer at 4khz causing a cancellation.

 

[1096bimu]

Your data shows some pretty major crossover issues. Your implementation is simply not working right at all. Definitely need to analyze the speakers with a particular process and design the filters in software, anything else is unfortunately just guessing and your in room on axis response is not enough to tell whether drivers are being integrated properly.

As another mentioned, your window scaling is misleading, a range of about 50db is what most people use.

If we import your passive response into vituixcad, we can see that your tweeters response is sloped way too far up towards the top end, and its fighting the woofer at 4khz causing a cancellation.

View attachment 504642

oh so that's what was causing the big swing from 3 to 4khz, makes a lot of sense.

it is gone after I implemented the low-pass on the woofer though.

I do wonder whether the much gentler dip around 3khz is also due to this cancellation and it would be better if I made the low-pass even lower?

EDIT, I just tried playing a 3khz tone and covering the tweeter with my hand causes the sound to be louder, I guess that means there is destructive interference, first I’m gonna try to just reverse the polarity on the tweeter, and try to get a crossover with a sharper and lower crossover point?

 

[Ktacos]

oh so that's what was causing the big swing from 3 to 4khz, makes a lot of sense.

it is gone after I implemented the low-pass on the woofer though.

I do wonder whether the much gentler dip around 3khz is also due to this cancellation and it would be better if I made the low-pass even lower?

EDIT, I just tried playing a 3khz tone and covering the tweeter with my hand causes the sound to be louder, I guess that means there is destructive interference, first I’m gonna try to just reverse the polarity on the tweeter, and try to get a crossover with a sharper and lower crossover point?

You'll have driver integration problems pretty much everywhere unless you can measure the speaker and develop a crossover properly. That or buy some sort of DSP/amp combination that gives you 4 channels and just run them active. Otherwise there's not really anything anyone can do to help without measurements of the individual drivers.

 

[1096bimu]

Update, there was a crossover issue, tweeter and woofer fighting each other around 3khz, I added lowpass to the woofer and reversed the polarity of the tweeter, it's much better now.
I've also took the time to bring everything to an empty room with a mattress on the back for sound treatment, stacked them together in a symmetrical way and did some better measurements, lots of work.

I think doing a true side by side comparison now I can subjectively identify the differences I'm seeing in the measurements, there is certainly some difference in frequency response but nothing an EQ can't fix.

I know the mattress is too much dampening but at least it's a very smooth effect, I just wanted to make sure there's no comb filtering affecting the results.

 

[Ktacos]

I think doing a true side by side comparison now I can subjectively identify the differences I'm seeing in the measurements, there is certainly some difference in frequency response but nothing an EQ can't fix.

Crossovers aren't just about frequency response on axis. You have to consider phase which you can't get right without measuring the drivers individually and import the data into design software. Considering you have a measurement mic one has to wonder why you're actively ignoring suggestions on how to this the right way.

 

[1096bimu]

Crossovers aren't just about frequency response on axis. You have to consider phase which you can't get right without measuring the drivers individually and import the data into design software. Considering you have a measurement mic one has to wonder why you're actively ignoring suggestions on how to this the right way.

because even if I find out the exact phase mismatch there's no way for me to adjust the phase except for reversing the poles. And I don't need an exact measurement to know whether reverse is better than not reversing.

Your suggestions are welcome in that they help me recognize what might be wrong or missing, but I'm not looking to become a speaker engineer or start competing with KEF, I don't have the time, energy or right testing environments to make precise engineering measurements. Your standards for the "right way" are simply too high and unnecessary for me.

 

[Ktacos]

because even if I find out the exact phase mismatch there's no way for me to adjust the phase except for reversing the poles.

You adjust phase relationship between the drivers with the crossover filter. Polarity is not phase.

but I'm not looking to become a speaker engineer or start competing with KEF, I don't have the time, energy or right testing environments to make precise engineering measurements. Your standards for the "right way" are simply too high and unnecessary for me.

Seems you have the time since you've spent so much on the project already, but working effectively backwards. I'm not sure why people think taking two measurements of FR and two impedance sweeps is creating "precise engineering measurements". It's like the absolute bare minimum and it'd take like 10 minutes. To not really try seems like far more of a waste of time and effort. Whatever.