Something stirs in the Undergrowth

[IslandPink]

Tested on Friday Saturday . These curves make more sense in comparison to the listening. Taken at a load of discrete frequencies ( from my test CD ) from 75Hz to 6kHz , using a hand-held Radioshack meter on 'C' weighting . I had the horn in the existing test position for curves 1, 2 and 3, just with the meter in three different locations near the listening position, about a foot or two apart. The fourth curve, I moved the horn unit more central into the bay window area and had the meter around head height somewhat more to the right. All this just to get a mix of room response effects & try & get a useful average to see what the horn is doing .

I think I had a cushion on the floor for one of those to try & do something about floor bounce. Must check notes. Not sure if the 850-900Hz thing is real as it swings around a lot on the 4 measurements. Generally there are two many big dips & bumps which ties in with the poor tone of the finished Mk.1 design as discussed above. The horn is now history as I started chopping bits out and getting rid of the third bend.

[IslandPink]

So, more carpentry last couple of days in between sneezing bouts. Conditions have been challenging on the back porch - daylight hours very restrictive.
I'm cutting some heavy-duty stuff like 25mm MDF and 30mm chipboard & constructing into a new mouth section which extends 34cm out from the previous top-panel, and flares to around 60 x 43cm.
The first two pieces I've made and joined ( with screws and epoxy ) seem to be about 1cm out of square when checked just now
Also finally put a few screw holes through the second side-panel and got it fixed on.
Will try & test later this evening when the epoxy has set on that bit - see if I'm making progress.

[IslandPink]

New mouth nearly complete .
Got a visitor tomorrow for a 2nd opinion.

Was looking for a 1mH choke to adjust low-pass but only had 0.33 and 1.8 to hand. Tried 0.33mH and it cleaned up the vocals a touch, by rolling the folded horn down a bit quicker after 2kHz.
So, I decided to put this in the driver enclosure.
I also included a little bottle of water in case the Fostex gets thirsty.

[IslandPink]

I'll try & get a picture up later - camera/PC aren't talking at the moment.
Anyway, so Nick came round on Saturday.
We had a good listen to the mono system with 'Mk.2' PMVO horn. Trying to characterise where it's working and not working.
The new straightened (non-existent) third bend has not done much to improve the sound in the lower mids . This is annoying but also useful information.
Lower down it's alright. Upper mids where it interfaces to the Yuichi horn are also good. We got nice sounds out on female vocals and higher male voices.

Nick brought my tapped horn ( white coffin ) back and brought a mini-DSP board and the F5 amp.
We were able to get a decent integration for low bass by putting the tapped horn about level with the back of the PMVO and applying a 70Hz low-pass with about an 18db/octave slope. It took a bit of faffing and swapping phase and adjusting the level ( thanks Nick ! ) but it came good - at least as good as could be expected when you have problems with the lower mids say 150-300Hz , and the harmonics are hence not all there.

We also had a quick listen to a loaned Fostex FX120 which I'm keeping for a friend and agreed it's a nice unit - one of the best .

So anyway , back to the horn . The fact having NO third bend doesn't cure it suggests the real problem is in bend 2 . This is a near-180 degree turn with a height starting at 11cm and rising to 14cm on the way out . There could easily be 34cm of path difference arounf the outside compared to the inside here ( I calculated today ) . This would be about 1/4 wavelength at 250Hz so maybe this is the cause. The effect is to have a lack of snap and precision in the lower mids and a generally fuzziness to anything that uses that area, like drums. I will try ( tomorrow perhaps ) to do another set of dB measurements in different positions ( & average ) with the RadioShack meter as shown above - see what's changed.

I have been looking very carefully at pictures of WE13a horns and replicas. I'll throw up a picture in my next posting . I reckon the final bend on the big WE13s could be shallower than my 12cm high 2nd bend - perhaps only 10cm ; and not as tight a curve. I need to learn from this, but also use the current Mk.2 prototype horn as much as possible for experiments.

[MJK]

Is what you have designed a horn or a TL? You need to answer this question to plan your path forward.

If the mouth is too small for it to behave as a horn you are probably getting TL standing waves and a ragged SPL response, stuffing with fiber may help. If it is a TL then the bends probably do not impact the results too much. If you are trying for a horn and using "horn math" then you need to make sure the mouth size is consistent and large enough for the frequency band you are trying to reproduce.

Remember, not all expanding geometries behave like horns.

[IslandPink]

Hi Martin -
I spent a fair bit of time getting the basic params OK for this one, after having failed with a shorter lower-mid Tractrix horn a few years ago.
Length is 1.7m, so 1/2-wavelength at 100Hz.
Mouth size is 2600 cm^2 , so , trying to remember just from notes I wrote about 3 or 4 months ago, that's 1/2-size ( 2pi) at 110 or 120Hz and 1/4-size at 85 Hz. My target was to get it horn-loaded to around 100Hz and it seems to do that.
The best sound I had from it was here :
http://www.audio-talk.co.uk/phpBB2/view ... start=1740
..and my comments from 13th October. This has been referred to since as the 'legendary'' (my words) 'Lashup #2' . Since then it's been downhill or pros/cons. At this stage it had an incomplete third bend that must have somehow been compensating for problems earlier in the line. At that stage it had most of what I wanted from a lower-mid ( & upper-bass ) horn and exceeded anything I'd heard elsewhere in a few key areas. It's what is keeping me going at the moment. There was a tone and dynamic precision on lower piano notes that I've not heard with anything else I've built.
The basic idea for this beast is based on Kevin Scott's quest at Living Voice that started with the Vitavox Thunderbolt, through the 'Air Partner' and culminated in the 'Vox Olympian' , hence the PMVO name which stands for 'Poor Man's Vox Olympian' .

So I think it's problem with implementation ( folding ) rather than the basic design which meets the general ( Keele ) rules and sims very nicely. The literature on how to get folded horns to work is very sparse. I'm trying to extract what I can from the WE designs and other archive material from Dinsdale, postings by Bjorn Kolbrek with old-school patents, and cross-sections of things like the Tannoy Westminster.

[IslandPink]

Forgot to mention John Sheerin - this collection is ACE :
http://www.jhsaudio.com/images/roundbend/
..and from over 10 years ago. Where is speaker simulation going ... ?
Great shot of a 13a replica here. I reckon the internal height through the final bend may be only 10cm or so. :
http://www.itishifi.com/2012/04/western ... ction.html

[IslandPink]

More test data here from last night. Measurements from the 'Mk.1b' horn ( straightened third bend ) at a typical listening position, with the Radioshack. Some moving of the horn and meter as before .
Five different sets of data at 30 frequencies -

Average of all 5 to get a better idea what the speaker is doing.

Not great but different to the 3-bend version .... after a couple of hours this morning I had an insight into what might be the problem and how the Mk.2 would be built differently.
More on that later.

[IslandPink]

So later that morning after creating the curves, I had an idea about what's going wrong.

I reckon some fraction of the wavefronts are seeing the outer curve of the 2nd bend as a reflector , and there are closed-pipe resonances then generated between the horn mouth and the outside of that bend.

This distance is about 114cm on the current Mk.1b 'bodged' prototype ; this would result in a fundamental ( peak ) at 75.4Hz , then a dip at 151Hz, a peak at 226Hz, a dip at 302Hz , etc . This is near enough exactly what I’m getting, if you look at the plot above.

Also if I look back at the previous curves from the unit as-built at first, see further up this page, I get a completely different set of peaks and dips – but they correspond rather nicely to the distance between the 2nd and 3rd bend, which in that case was about 70cm ; so we get peaks at 120Hz, 300Hz etc .

So for a better design we need to avoid anything like bend 2 - it's is too broad , especially coming out, and too abrupt ( around 160 degrees turn ).

Edit - just to make this a little clearer I've drawn a picture below. This illustrates the original folded horn path ( dotted at mouth end ) , the Mk.1B bodge to straighten the third bend, and the crucial distance of 114cm I'm referring to, above, that may cause the peaks & dips.
On the right is a modification I did yesterday to see if the second bend can be improved or at least changed. More on that once I get a bit more listening.

[IslandPink]

This is a very interesting thread :
http://www.analog-forum.de/wbboard/boar ... 4f31ddcf78
Which leads to this, also extremely interesting - the discovery that a $30 police siren driver ( with alnico magnet ) can work as a superb bass/mid compression driver a bit like the WE555. Look at the measured curves half-way down this page !
http://www.diyaudio.com/forums/multi-wa ... rs-14.html

[IslandPink]

Going back to the fold mod in the drawing above. I can say there are some useful ( educational ) differences in the sound, with the diagonal fold piece, but it's not suddenly corrected the problems & made it sound great.
The results are somewhat as I expected. There's a bit more coherence in the upper part of the horn range, some more snap on things like snare drums ; and the female vocals are probably slightly better - but they were pretty good before. Unfortunately the lower end of the range has dropped a bit in level - I think the 'pinch point' caused by the diagonal piece has affected the loading and extension of the horn below 200Hz. Useful experiment though.

[IslandPink]

The stuff on the German forum about compression drivers for the Sato horn has really got me thinking hard this afternoon about possibilities for the Mk.2 horn.
The SD370 siren driver turns out to be a bit limited at the upper end of the FR - goes to around 1.5kHz flattish then down sharply beyond 2kHz. The other driver Dietmar worked with is a modified version of the Klipsch K-55 driver for the Klispchhorn, LaScala etc. I haven't got the full picture of what the mods are ( a spacer on the surround to help lower the Fs , I think, a Peavey magnet , not sure if this adds or is a replacement to the standard magnet ) but these drivers can be picked up for about £150 a pair . This in sharp contrast to any WE555's or WE555 remakes which are £7500 and upwards.
Here's the frequency responses of the SD370 and the K-55 Mod , on the 100Hz Sato horn, thanks to Dietmar on the analog-forum ; I believe he's from Munich area :

The modified K-55 is just stunning. Alnico too !
On a plane-wave tube you can see the K-55 can go below 100Hz :

The disadvantage of using a driver like this is the extra 1.4m of length ( to around 3m ) added to the horn. The exit aperture is only 20mm for these drivers. The advantage is you get the entire vocal range ( & harmonics) from 100Hz to 5kHz with one unit, in the Western Electric style. You may be able to get away with not time-aligning the tweeter which comes in at 5kHz , which would avoid DSP.
For the SD370 siren driver ( same company as makes the later K-55 drivers actually ) you would really need DSP, as you'd have to cross at 1kHz or so to another horn & driver, with a potential path length difference of ~2.5m at 1kHz. DSP would be used to delay the big horn signal by around 7msec. Not sure I'd want to do this unless someone could demo a properly transparent DSP system to me.

Anyway, so..... do I have to make the MK.2 horn 'convertible' to take a FE208Ez with 90cm^2 throat plus an extra 1.4m 'coil' of horn expansion for the 3.3cm^2 throat of the K55 ?
Mulling over it now !

[IslandPink]

Here's some efforts I've made towards a Mk.2 horn , not so much a PMVO , but a 'Mini 12A' based on the WE12a, but also with some specifics in terms of bend radius and height ( aspect ratio ) from the 13a. Since it's only 1.7m long, it doesn't easily join up if you want to get the final bend 'gentle' . No.2 below just about joins up into a circle, when you add the back-volume around the driver. However, No.1 looks more elegant and likely to be more successful acoustically as regards the bends.
The expansion areas vs. length are exactly as the Mk.1, but it's a little wider and the overall aspect ratio is different, being 20-30cm deeper. It has only 90deg bends - nothing like the 160deg or so on bend 2 of the MK.1 .
For the potential compression driver option , I'm not sure yet where to go. If I built No.1 below, then I could make the throat end 'modular' and just have a 1.4m long 'stalk' of the initlal horn expansion feeding into the Fostex position, but extending out into the room, for trials with the K-55V . Not sure how eventually this would work if it went ahead with the compression driver. It would really need to be 'coiled' inside or somehow come out sideways after one 'rev' ...
Decision, decisions ....

[Toppsy]

Mark,
have you considered CNC routing for cutting all the panels for your Mk.2 horn?
Certainly would make life a lot easier for you and you can build both versions you have sketched in double quick time.
I do believe one of the Simon's on this forum (can't remember which one but not Simon the camera guru) has access to a University /shared CNC router for just the cost of the material, provided you can supply him the CAD files.
I'd be prepared to do the CAD drawings for you if it helps any?

[IslandPink]

This is a jolly good idea Colin, I had vaguely had this in the back of my mind.
I think doing all the panels/pieces is optimistic, but the side pieces as shown in the cross-sections above might be possible, along with a few other parts.
There are compound angles along the edges, of course, due to the width-taper, which must be considered, especially around the bends - and to be honest these keep me awake at night sometimes, trying to visualise ! - but the width taper could be fairly slow in the Mk.2 if done correctly, which would make the compound angles along the edges pretty small, at a level that a small amount of planing could correct ( or of course angling of the piece during CNC routing ).
The top and bottom would be best made from pieces down to about 1" or 2" size along the lines of the Western Electric horns ( 12a, 13a ) .
My plan currently is to make the sides from 25mm birch, and the curved out/inner curves from 1" thick ash or maple for rigidity. 2" wide slats might be enough to follow the curves, given the shallower radii on the Mk.2 - and these could be planed smooth on the outside to give a nice smooth look.

I have a tame mechanical engineer at work who could help, but your understanding of CAD in the context of speakers could certainly be a bonus.
Definitely worth finding out which Simon we are talking about to have some initial discussions

Of course at present the plot could thicken, if the K-55V compression driver and the 3m overall length have to be part of the plan ....