Mike's MTM- Initial Crossover Ideas

[Mike C]

Having chosen the SEAS Er18RNX and 27TDFC, I'm now considering the crossover. New territory for me - but I've already had great support here.

Even if I had a good crossover design, it would probably need reworking to suit my 300B SET amp which has a signficant Zout of 2 ohms.

Anyway, the immediate objective is to get some confidence, to enable me to start building the cabinets.
So here's where I am.

I have FRD and ZMA files for the 27TDFC but not for the ER18. So I started with the makers published graphs, and 'massaged' them. First, the FR is measured at 0.5 metre .... but they do give a sensitivity figure so I moved the line to this. Then the FR is measured in a 12 litre closed box so I added the difference due to my enclosure. Next, the FR shows baffle loss, but they do give an ideal curve without baffle loss. After much messing about I came up with a curve and measured about 40 points on it. I also measured Z at about 40 points.

Then I converted the data for my parallel woofers by adding 6dB to FRD values, and halving the ZMA figures.

Next I ran the baffle loss simulator, getting these figures:
Relative to 1kHz ...
-1 dB at 2kHz; -1.5 dB at 500Hz; steady ramp to -4dB at 200Hz; -6dB at 100Hz and below.
Not so much a step as a ramp - must be due to two woofers, one near the top of the box.
Jay_WJ suggests starting with 5dB of bsc so I set my target at:
0 at 1kHz; +1.5 at 500Hz; +3.5 at 200Hz; +5 at 100Hz.

I first tried Isaac MCN's simulator and quite liked it. Since then I've used Jeff Bagsby's - wow!

Obviously I've looked at Zaph's crossover for his XG18/27TDFC MTM, but I've done a lot of playing around myself .... this simulator is pure magic! I love it, it's real engineering.

Here's where I get to.
Woofer, 2nd order; 1.5 mH and 25uF.
Zobel on the woofer, 20uF and 25 ohms (mentioned below).

Tweeter, second order; 8uF and 0.75 mH. Dropper resistor 3.4 ohms before the crossover.

I have (I hope) attached a picture of the resulting FR. Note the green line where I have taken off my BSC 'target' values above, to give an expected FR. It looks reasonable to me - but hey, I'm no expert here. Even so, it seems to be close enough to be worth building and then fine tuning.
In which case, I've met my immediate need (to get enough confidence to go ahead with the cabinet). Further crossover modelling, perhaps with better data, may follow.

Comments:
I put a series R in ahead of the crossover, 2.1 ohms, to simulate the amp and cable. This has the effect of lowering the BSC - which perhaps explains why my series inductor on the woofer is a size higher than you might expect.

The zobel isn't of normal values - I just found that a zobel type RC seems to adjust the point where BSC comes in. By trial and error I adjusted this to give 0 BSC at 1kHz but coming in at a bit below.

You can see slight peaks at 5kHz and somewhat above 10kHz due to peaks in driver response. However, it seems that listening at about 15 deg off axis may smooth these out, so I plan to do nothing else about them.

I have a question about the tweeter cct. L-R calculators give a lowish C of 8uF, and a high L of 1mH. Whereas Zaph, Newfinish and Jay_WJ use higher C and less L, roughly 11uF and 0.25mH and evidently get good results. I don't quite see why.
(If there's a problem here it's not with the ccts, it's with my understanding).
I seem to get good results with 8uF and 0.75 mH; though 10uF and 0.4mH also look pretty good.
What am I missing here? Should I go for the lower L and higher C? (Cost is not an issue).

The hump between 70Hz and 110Hz approx is not a function of the crossover, but results from my reflex cabinet. If this gives a *slight* touch of warmth, it may be OK (it's only 1dB or so max). If it gives boom then I'll play with port tuning, port stuffing or convert to sealed with scanvent.

I'm encouraged .... am I living in a fool's paradise? :-)
PS: planning to buy materials tomorrow.

[sreten]

Hi,

If you want the best efficiency midrange / treble consider line level
BSC - this will ~ double effective amplifier output for near free.
(more true for classical music than bass heavy popular music)

The LR crossover calculators assume zero source impedance, in your
case the source impedance is ~ 5 ohms, so for L/R electrical values
should not anywhere near the calculator values.
Of course what matters is the acoustic roll-off of the tweeter, the
electrical filter function typically is not any of the classic functions.

also checkout : http://www.troelsgravesen.dk/C17.htm

/sreten.

[Mike C]

Hi,
Thanks for the info.
Line level BSC does sound interesting .... a filter between preamp and power amp. I'm not very good at filters (though I have done a couple of phono stages) but I can read up about this. If it needs an active filter it'll be a problem ... no room in the hifi cabinet. That's one reason I'm not looking at active speakers.

I have studied the Troelsgravesen C17 MTM very carefully already. In fact it influenced my design a lot. If the driver were available I might have built that.

The ER18RNX is somewhat similar to the C17; well, Qts and Vas are very similar. I'm hoping that overall the ER18 may be a better performer (though less sensitive); fingers crossed here!

Anyway, I've bought the wood and I'm going for it .....

Regards
Mike

[sreten]

Hi,

line level BSC does not need an active filter see :

http://sound.westhost.com/bafflestep.htm

A simple fixed BSC can be built into the amplifier input.

Another way of doing it is modding the amplifier feedback loop.

/sreten.

[Jay_WJ]

Nice work, Mike.

As for the Zobel, it is only a theoretical, textbook idea. If you design a crossover relying only on textbook formulas with limited T/S parameters, a Zobel may be useful. But with CAD available, it is not necessary in most cases---waste of components. Here's a good thread discussing the Zobel: http://www.pesupport.com/cgi-bin/config.pl?read=379703

The use of low C and high L in the tweeter net is possible, too. The only reason why I tend to use high C and low L is that I want to avoid a large impedance swing. If you compare your crossover and mine (without impedance flattening circuit), you'll see what I mean.

I'm almost ready to post my MTM (floorstander) crossover at my website---probably this weekend. Unlike the previous result I showed you, this one will reflect different baffle diffraction effects for bottom and top woofers. Before I post it, here are some results:

This crossover is a version with an impedance flattening circuit to be used with a tube amp.







Mike, if you're not planning to use your own in-box measurements for this project, I recommend to build my crossover. I'm pretty sure that a little amount of tweaking will make the speakers sound right to your ear. I've done a lot of simulations using the same, consistent method to create my designs and replicate many existing designs. Also, I have a good "feel" of how my (not others') simulated responses will sound in reality because I have voiced my designs which were simulated in exactly the same way as the above crossover was done. Of course, it is always a good idea for you to play with your own simulations to make everything sure.

Best,

[Mike C]

Many thanks for this ... I'm amazed by the great support and expertise on this forum.

I will definitely model your crossover and have a play around. Also I'll check how the tweeter cct affects impedance - though my playing around with different speakers and weird parallel combinations seems to suggest my amp is moderately tolerant of varying load impedance.

Building the boxes may start tomorrow. It'll take a while as I'm going for curved sides laminated from MDF and plywood, and they will take a long time to build up.

I'm getting excited and that's what it's all about .....

[sreten]

Hi,

FWIW impedance flattening circuits are anything but valve amplifier
friendly, what they do do is reduce the frequency effect of a valve
amplifiers output impedance by heavily loading it.

A valve amplifier friendly speaker should be designed with significant
source impedance and balanced with such whilst maintaining low
phase angles at low impedance and a high as possible impedance
midrange / treble, phase angles at high impedance matter less.

Unlike transistor amplifiers valve amplifiers can and do give higher
output voltage (not power, but dBV) into higher impedance loads
which the constant impedance load effectively slugs.

Consider dBV output of a valve amplifier at say 10KHz into a variable
load. Often for say 1% distortion output is pretty constant, e.g. 10W
for a 25W (at 1khz) amplifier. the higher the treble impedance load
the more clean voltage swing you get.

Valve amplifier watts / voltage swing is expensive, seems a shame
to waste any capability, which why line level BSC is a good idea.

/sreten.

[Jay_WJ]

Mike,

FYI, my preliminary crossover page has been updated with the Seas ER18RNX / 27TDFC MTM design. The crossover has been changed a little bit from the previous version I showed you. The change is a little more BSC.

I think this design is really great for those who look for low-distortion, high-sensitivity speakers to be used with a valve amp.