Improving IPL M4s

[S-Man]

I have had a pair of (old version) IPL M4s in my kitchen for quite a few years. They sound OK but IMO don't meet the claimed "The M4s will compete with the very best of small speakers giving a smooth and very natural sound". Every now and again I try them against my Murphy CAOW1s, which I think can rightly be placed amongst the very best of small speakers, and the M4s sound a bit flat - tonally and in terms of dynamics.

The M4s use Morel MW144 and MDT39 drivers, which ought to be pretty decent.
IPL's specs and FR plot look very impressive. +/-1.5dB from 250Hz to 20KHz.

I do seem to remember having a chat with Ivan about the "flat" sound and IIRC he said that's how the drivers sound.

So what is going on?


I decided to try to find out using a spectrum analyser app on my iPhone and pink noise. The results were not helpful due to lack of resolution of the app and the room adding way too much, however it was very obvious that the pink noise sounded duller/flatter than on the CAOW1s.

So I started looking for a way to turn the mids up a bit. I fouhd this thread which describes the crossover topology and values, post #43 (can't see any issue with this as Ivan has retired and this is an obsolete design):
https://www.diyaudio.com/community/threads/help-with-crossover-for-mw144-tl.295584/page-3

After a lot of playing with XSim, compromised by lack of my own measurements and only having an MDT30 FRD file I decided to try a different crossover. It uses the topology shown here with slightly different values in the tweeter branch:
https://ldsound.info/2-x-polosnaya-as-na-morel-mw144-i-morel-mdt30/
C2 = 4uF, L2 = 0.5mH, R1 = 5R6 and R2 = 12R

The effect of these changes is to lower the crossover point from IPL's claimed 4KHz to around 3KHz. TBH I don't think the original is as high as 4KHz.
For quite a large tweeter this seems like a good move.

I have this crossover sat on the top of one M4 and can confirm that pink noise now sounds much more similar to the CAOW1, if anything it's slightly lighter/brighter sounding.

Initial impressions using the one modded speaker are that it does sound better. It sounds more tuneful and less congested and gritty in the upper mids.

There were so many assumptions in my XSim sims that I am sort of surprised that I was able to make a worthwhile improvement over a veteran designer's efforts.


Any opinions on what direction to take next? Are these drivers good enough to delve deeper?

 

[S-Man]

Not sure how much interest there is in this?

Anyway, I've bought a measurent microphone so that I don't have to keep borrowing one.

Just been having a go with it and initial in-room measurements show that my new crossover is an improvement. The original has a significant dip around 4KHz and reversing the polarity of the tweeter seems to help fill this a bit. My XO has a lesser dip and actually has something of a reverse null at 3KHz, which indicates better phase alignment through the XO region.
This all needs further investigation with proper gated impulse measurements to be more sure though!

Further listening and comparison with the CAOW1s confirms my initial opinion, the main difference seems to be in hf dynamics. The Morel tweeter is some way behind the OW1 at portraying the transient attack and decay of cymbals, although so are most tweeters!

To answer my own questions:

Measure and tweak (for education and fun). Maybe consider a better tweeter, although as it stands the M4s are good enough for the kitchen.

I'm still left wondering how Ivan come up with his XO .

 

[a.palfreyman]

The original has a significant dip around 4KHz and reversing the polarity of the tweeter seems to help fill this a bit.

We're they s/hand? Wonder if the tweeter was switched by previous owner, or accidentally connected reverse polarity...does seem odd though.

 

[S-Man]

No, they were build from scratch by me, to the plans provided with the kit.
The XO diagram clearly shows that the tweeter should be connected with the same polarity as the bass unit.

 

[S-Man]

I am more or less ready to start measuring...

I bought myself a low cost Behringer ECM8000 microphone and have spent considerable time trying to calibrate it against a borrowed B&K 4133 (+ energiser/preamp). Having looked at the cal curve for the 4133 I've come to the conclusion that the Behringer most likely has a flatter FR!!
There are horror stories of ECM8000 responses being "all over the place", maybe I got a good one?

My expectations were met in respect of just how difficult it is to get reliable acoustic data.

 

[S-Man]

Starting to get somewhere. It's been a steep learning curve!

I now have reasonable correlation between simulations and measurements for the M4 with IPL XO:
(Below 250Hz is not accurate due to the gating used)

The top curve in the measured reponse and the bottom on is the sim. I'm not sure why there's a bump at 4.5KHz in the sim and not in actual measurement. It does seem to be in the tweeter-on-it's own measurement at a lower level though.
Anyway, the large dip in the upper mids and presence region explains why the M4 sounds so flat!?!

Now that I have the sim under control, I can (re)start crossover tweaking.

 

[Fatmarley]

Starting to get somewhere. It's been a steep learning curve!

I now have reasonable correlation between simulations and measurements for the M4 with IPL XO:
(Below 250Hz is not accurate due to the gating used)

The top curve in the measured reponse and the bottom on is the sim. I'm not sure why there's a bump at 4.5KHz in the sim and not in actual measurement. It does seem to be in the tweeter-on-it's own measurement at a lower level though.
Anyway, the large dip in the upper mids and presence region explains why the M4 sounds so flat!?!

Now that I have the sim under control, I can (re)start crossover tweaking.

That looks pretty bad. They should sound much better if you can fill the dip.

 

[S-Man]

After quite bit of fiddling with XSim, I have come up with a new crossover. It uses 2 more components and partially fills the gap.
The lower plot is the original and the upper is my (hopefully) improved XO (the green plot if the raw tweeter - which seems to have plenty of peaks and troughs on my baffle ).

 

[a.palfreyman]

Be interested to know how your improved crossover in the OP measures as you said it sounded much better...

 

[PigletsDad]

You could try some thick felt round the tweeter to eat up diffraction peaks and troughs, see if that gives something that doesn't need so much fixing up in the XO. One person's experiments at Felt Rings verses Felt Blocks for Baffle Diffraction Reduction (speakerdesign.net)

Some other measurements Diffraction Doesn’t Have to Be a Problem | audioXpress

 

[S-Man]

Interested to try the diffraction reduction thing.

Felt seems to come in various qualities. The real stuff is very expensive and seemingly not available in small quantities.
I wonder if this would work?:
https://www.ebay.co.uk/itm/38422289...d=link&campid=5338728743&toolid=20001&mkevt=1

 

[Fatmarley]

Interested to try the diffraction reduction thing.

Felt seems to come in various qualities. The real stuff is very expensive and seemingly not available in small quantities.
I wonder if this would work?:
https://www.ebay.co.uk/itm/38422289...d=link&campid=5338728743&toolid=20001&mkevt=1

Good find.

I used some Hapla 100% wool felt sheet for testing once. I don't know where the measurements are, but it definitely improved the frequency response.

 

[S-Man]

Be interested to know how your improved crossover in the OP measures as you said it sounded much better...

Unfortunately I have dismantled that version, so I don't have any measurements. In the meantime I have gained a deeper (or less shallow) understanding of what matters in crossver design, especially regarding phase alignment. I also managed to get measured results that align much better with the sims.
After much measuring and aligning of curves I have a better estimate of the distance between the acoustic centres of the drivers. I believe the correct figure is 0.75", but it could be anywhere between 0.6" and 0.9". This figure is vital when optimising the crossover and it seems like I had a wrong value in my earlier XSims.

Actually it seems I didn't actually fill in any of the upper mid hole. It looks like all I did was to knock the 600Hz to 3KHz region down a dB or 2 whilst worsening the phase alignment :
(Green = original. Blue = XO version 1)

I built up XO version 2 and measured it today:

This one does have good phase tracking and looks like it fills the upper mids to some extend, BUT I don't like the sound. It's thin and bright .

I have another to build. I am learning that crossover design is not easy!!

 

[S-Man]

So, on to XO3. The Sim shows nice phase tracking (deep reverse null), flatter upper mids and the XO point has moved been down to 2.1KHz. The MW144 is rolled off a bit faster, that and the lower XO point means that the 7-9KHz resonances of this driver are significantly lower in amplitude.

Here's the measured M4 with XO3 (in red) , the grille and some felt diffraction reduction versus the original, with grille (green):

Here's the measurements for the XO3 (red) version versus my CAOW1 (blue), both with grilles:

The M4 will always be used with the grille in place. THe CAOW1 measures better without the grille, but I listen with them in place because I don't like looking at naked drivers.

After listening to one M4 for an evening I would say it is very much improved. The hollow and flat presentation has gone, it now sounds in-tune and on time. Maybe it's not the most dynamic speakar around but it's very pleasing on the ear - which is exactly what I want for the echoey acoustic of our kitchen.

The ECM800 mic spec shows a few bumps of about +2dB in the 7-10KHz region.
All my results are for 1V input to the speakers. So just under 9dB needs adding to the plots to get the output figures for the nominal 1W at 1m.

Might have to have a look at my IPL in-wall TV speakers next .

 

[S-Man]

Did some further measurements, tweaks and listening over the weekend.
Sadly it seems that the grille frame does mess up the measurements AND the sound. Without the grilles the modded M4s are actually very good indeed (crossed over at ~100Hz to my active bass extenders), better than some highly regarded mini-monitors I have tried. Night and day compared to the originals! Although the efficiency is not that high, they can take a lot of volts without signs of stress and are easy to drive - never dropping below 6 Ohms with reasonable phase angles. I would say that they have a similar character to the small ATCs and really wake up as the volume increases.

I'm thinking of redesigning the grille frame. The cloth itself just takes a couple of dBs off between 7K and 10Khz, which is probably favourable in this design, the frame is the problem.
I'm also wondering if it's worth building a new pair of speakers with the Morel CAW538 (new and improved MW144) and a really nice tweeter as my main speakers?

Any suggestions for a really nice tweeter?

 

[PigletsDad]

Long shot, but if you some felt inside the frame, behind the cloth, to absorb some of the reflection off the frame, would that help? Easier than building whole new speakers!

 

[James]

This is how hardcore DIY loudspeakering starts. Well done, @S-Man!

Crossover design is both infuriating (especially for a 3 or more-way design) and rewarding. The key is having complementary drivers, good measurements and methodical modelling. Phase is super-critical to coherence, and getting a deep null with reversed polarity on one driver is a good start.

As for a really nice tweeter, my go-to for 19mm is the OW-1. If you desire a bigger tweeter, I really like the classic Scan-speak 2908-9300 or 9500. Not cheap at all, but really natural sounding.

 

[S-Man]

This is how hardcore DIY loudspeakering starts. Well done, @S-Man!

Oh dear!!
I suppose I'm almost done with amps (at least until my SE230 arrives ).

I suppose it was a really bad outcome with the M4s. Like a new gambler getting a decent win on the 1st bet

Crossover design is both infuriating (especially for a 3 or more-way design) and rewarding. The key is having complementary drivers, good measurements and methodical modelling. Phase is super-critical to coherence, and getting a deep null with reversed polarity on one driver is a good start.

As for a really nice tweeter, my go-to for 19mm is the OW-1. If you desire a bigger tweeter, I really like the classic Scan-speak 2908-9300 or 9500. Not cheap at all, but really natural sounding.

I have some slight concerns that the 2nd order acoustic slope pushes the OW1 just a bit too far in the CAOW1s.
I have always liked the sound of the 2905-9300 and 9500. May I ask why you didn't use one of these in the Ergo IX?

 

[James]

I suppose it was a really bad outcome with the M4s. Like a new gambler getting a decent win on the 1st bet

I can think of worse addictions than DIY loudspeakering.

I have some slight concerns that the 2nd order acoustic slope pushes the OW1 just a bit too far in the CAOW1s. I have always liked the sound of the 2905-9300 and 9500. May I ask why you didn't use one of these in the Ergo IX?

I use a 2nd order filter for the OW-1 in the Ergo E-IIIRs. I think these are crossed at 3.5kHz to a Seas M15CH-001 mid. They don't sound distressed to my ears.

I went with the Morel tweeter mainly to better align acoustic centres on a vertical baffle. The other advantage is the horn loading simplified the crossover design (i.e. can get 4th order acoustic cross with a second-order electrical filter), and conveniently reduced the diffractive effects of a sharp-edged baffle.

 

[S-Man]

I think the CAOW1 crosses the OW1 at 2.5Khz. Dennis Murphy says he couldn't hear any stress with 2nd order acoustic. I just wonder if occasionally I can, perhaps because mine can go cleaner and louder due to not having to do deep bass.

I have been thinking exactly the same about the acoustic centre of the CAT378. I must re-read your Ergo IX write-up.