Adventures In Transmission Line Loudspeakers

The answer here would seem to be a baffle width which gives edge diffraction where needed (400Hz, nearly 12") and then to roll-off above it. Pretty much what the 7.5" baffle step correction does at present. However, this would result in a square front which might have its own problems?

I have made a start in measurement using the Audio Precision acoustic response measurement suite.

I don't have a calibrated microphone - it's an Audio Technica ATR3M unidirectional mic intended for studio talkback so isn't ideal for the purpose - but will suffice for now. It has a 150Hz - 10,000Hz frequency response, but as my main area of interest is within that range (at present) that's OK.

I am comparing the prototype Alpair 7.3 cabinets with a pair of Castle Trent loudspeakers. I finf these have the right balance for my room, and I'm trying to tune the Alpair 7.3's to perform similarly.

If I were driving the amps with the AP's analogue outputs I could use the EQ and invert functions to load the response of the Castle Trent's as a "flat" reference, and in that way see how the Alpair 7.3 cabinet tuning deviates.

Unfortunately these functions don't work using the digital outputs, and so I now need more balanced mic cables to run the amps from the analogue signal generator, rather than the digital signal generator via the Majestic DAC.

However, I have printed plots of both frequency responses which help me do a visual comparison.

I have reverted to acoustic baffle-step "compensation" by extending the cabinet fronts to the left and right, making the baffles wider. I have tuned the width to 400Hz, which is where the deep trough is on the Alpair 7.3 data sheet plot.

This makes the width 12" (305mm), which is same as their height, and so we have a square baffle. This isn't considered good, but the driver is quite well offset from centre, and doesn't seem to be causing too much trouble.

The baffle width has improved the sound considerably, and the plot shows a bit of a hump around 400 Hz, which the Trent's also exhibit. The Trent's do this due to crossover baffle step compensation, or I assume they do.

The prototypes exhibit quite a bit more ripple in the 500 - 2000 Hz band than the Trent's which might be due to driver positioning, which could be a little too close to the top of the cabinet. The ripple is about 5dB (max 7.5dB) between 400Hz and just over 2kHz, above which the output grows by an average 2.5dB (4dB peak) centered on 3.3kHz. The Trent's do similarly but around 2.7kHz. The mids sound quite similar.

However, it can be seen that the Alpair 7.3 spike considerably at approx. 7kHz, and again at 11kHz, although I guess the mic response is 3dB down here, so the peak is quite a bit higher than indicated.

Compared with the Trent's these peaks are about 3dB higher - the Trent's also exhibiting some spikyness (if there is such a word).

Obviously some of the ripples and spikes are being contributed by the microphone response, so I am not being hyper critical.

It is possible that internal reflection is contributing to these response spikes, and I did find earlier by sticking some "egg-box" foam to the inside rear panel that a lot of harshness was removed. The type of foam is packing grade and its egg-box pattern is quite a bit closer than regular egg-box foam. It will still reflect somewhat and perhaps a different grade of foam is required.

I also damped the insides of the sides with pieces of car panel deadening material, which also helped a lot.

I noted the Trent cabinets are fully foam lined, and this could help the prototypes.

In the next few days I will print the plots to PDF so I can get images from them and post them on this topic. You should be able to see by a "broad brush" that the two speakers have similarities, which shows you can get close tuning by ear, even though it is very time consuming.

If the high frequency spikes can be tamed by padding then it will make sense to make new cabinets based on my findings. For anybody wanting to have a go, the approximate external size will be 12"W x 12"H x 6"D. This assuming a 1" baffle and 15 - 18mm board (apologies for mixing measures).

The driver centre is 3.75" down from the top and is on the vertical centre line. A 40mm dia hole for porting experiments is located 2.5" up from the bottom, again on the vertical centre line.

Now for a bit of a conundrum: the port was fitted with a 6" 1.5" waste pipe tube for the tests, because I'm using a deeper cabinet with side extensions. A shallower 6" deep (full-width) cabinet won't take these, so the diameter of the tube will need to be smaller to get its length down. I guess a 1.25" waste pipe 3.6" long will do much the same. According to LDC7 design suite spreadsheet calculations the tuning frequency is approx. 45Hz for each.

If anybody would care to try I'd be eager to hear their thoughts on the sound.

Richardl60 wrote: I must admit I find this thread very interesting and unlike ,any of th electronic hardware threads I am not lost by it all after the first few lines of technicalities and jargon. Going back a week or so I understood the room/walls were basically a 'pig' and couldn't get anything to work there. Any ideas why the Castles appear to do so or have I missed a posting?

Thanks Richard, makes it worthwhile writing this knowing someone is reading it

Yes, the rooms/walls are still the same, but stumbling upon the Castle Trent's which were in the loft (a gift from our old lovely neighbour in MB a few years ago) made a considerable difference.

My thought is that cabinet resonances in both the LS35A (known for it) and the prototype cabinets - and their tuning - was somehow at, or beating with, the room resonance, making a very audible 63Hz boom. The Alpair 7.3 driver resonance is around 68-70 Hz (from memory).

Right now the port tuning should be 45Hz (ish) which is the "magic" 0.7 times the room resonance, which might have helped, and also pulling them out from the concentrated boundary area has helped. Added to that the anti-resonance measures to the cabinet have helped too.

Going back quite a few years I remember a conversation with an acoustic engineer who was of the opinion that nothing above 8kHz really mattered in a venue, and it should be at the -3dB point after which the frequency response was to roll off. I argued that it would lose the harmonic structure, but he simply laughed.

Having tried some top cut to the prototypes I might be in partial agreement. I feel that the higher harmonics are masking the musical harmonics and fundamental notes. There is an increase in acoustical output from the drivers above 6kH - and an unfortunate 9kHz deep trough - and then the high harmonics really take off above 10kHz.

So, next up is a trial of a wide notch filter which will give some 6dB of cut around 8kHz where the driver's output is peaking. This cut deepens a little more in the 10-30kHz region to bring down the large response peaks. If the drivers produce any supersonic output then there will still be useful output to drive them above 30kHz. I note that some say super-tweeters which go very high make for incredible imaging. This filter rolls back on at those frequencies.

We will see...