Go deep!

In my previous post, I mentioned a discussion about this project on the diyAudio site. One comment referenced an analysis by Martin J. King (http://www.quarter-wave.com/), widely regarded to have the best transmission line modeling software available. He stated that Dr. King had found “Dalines” (what G. L. Augspurger calls “Pipe with Coupling Chamber” ?) inferior to designs. I did find this reference on diyAudio that supports his statement.

Without further ado, here is my model of the Tang Band W8Q-1071F in a RSTL line. Could there be something better? Maybe. Does this strike me as a configuration that will have deep bass? Yes. Am I worried about the dip at 100Hz? Not really – right now I expect I’ll have an active crossover at or below 100Hz, so I’m not at all concerned. Should I go full passive, I’ll have to review my stuffing plans.

Image of TangBand response graph in RSTL

Simulation of TangBand behavior in the RSTL line


Standing on the shoulders of a giant

In trying to work out an imaginary crossover, I decided the best way to start was to take a similar design and see what CALSOD might do to optimize it. The closest design I could find to my plan (I keep changing my thoughts on drivers) is the Zaph Audio SB12.3. In particular it uses two SB Acoustics SB15NRXC30-8 5″ mid-woofers in an MTM arrangement along with a woofer. Pretty similar to the RSTLM (well, as close as anything gets these days). So I started by attempting to match the SB12.3 mid-woofer modeled response by using the crossover John Krutke specified. I then let CALSOD optimize it and it came pretty darn close to hitting the target:

Graph of predicted response vs target response

Predicted response vs. target response of CALSOD-optimized crossover


It is very interesting to see what CALSOD does to the component values. Very large or very small values are often a hint to remove a component altogether or change the topology. But if you let CALSOD optimize all parameters, it can come up with some pretty crazy values. Here’s the comparison of the original SB12.3 values to the CALSOD values:

Component values for mid-bass

The CALSOD optimizer chose an unrealistic value for C17.

In this case, it chose a value of over 5000 μF for C17. Ouch. In this case you must mark one or more values FIXED in the optimization so CALSOD won’t try to vary this value. This will likelybe a less optimal solution, but there is not much to be done unless we want to have the huge C17 value.

Also, after the optimizer finishes its magic, we must go back and select actual values that are available to purchase. Of course you can wind (or unwind) your own inductors, parallel resistors and capacitors and so on, but you have to decide just how important it is to have exactly the right value. When I completed this exercise, the final response was:

Graph of frequency response

CALSOD-optimized crossover response using standard parts

This is clearly more ragged at 300 Hz and 1.4 kHz than the original optimized design. It requires tweaking to get just the right tuning, but to a first approximation this is a reasonable approach.

I’m not going to bore everyone with a blow-by-blow analysis of each driver tuning and each change I make along the way. Remember all I’m looking for is proof of concept for my driver choices that I could create a passive crossover with acceptable impedance if I should choose. So this doesn’t have to be perfect – just reasonable.

Thank you

I’d like to thank John Krutke  at Zaph Audio and Madisound for creating and sharing this design. Frankly I don’t expect the RSTLM to ever come close to it in pure sonic terms. I’ve learned a great deal from reading the write-up for it. It is a shining example of what makes this such a great hobby.

Using CALSOD for crossover modeling

I own two programs for crossover modeling – CALSOD by Audiosoft and SoundEasy by Bodzio Software. Interestingly, they are both located in Australia. There really is no comparison between these two programs – SoundEasy is several magnitudes more sophisticated than CALSOD. However it is also painfully hard to learn and understand. For now, I’m not trying to achieve any sort of accurate models. Mostly I want to get a feel for how low the impedance might get with various drivers used in parallel or how low the efficiency might get with various drivers used in series. I know, I know… I’ve already picked the drivers, right? This 4 ohm world we live in is driving me nuts. TDL had the advantage of being able to create any driver they needed – hence the 16 ohm drivers they had available to them. All I want to do for now is make sure I have a reasonable chance to end up with a speaker that will stay mostly above say 3 ohms while reaching say 87 dB efficiency.

So I’m going to use CALSOD, so what? The version I have – 1.30N – was written in 1994. So it is a 16-bit DOS program that uses the character graphics of the original IBM PC (think EGA and VGA). So those were the two problems I had to solve – 16-bit DOS and OEM 437 font.

The first I tried to solve the compatibility mode in Windows 7 Ultimate. To make a long story short – no go. It couldn’t draw the graphs. Otherwise it looked pretty good with Windows 95 compatibility mode.

DOSBox to the rescue. A DOS emulator that runs on Windows, Linux and Mac, DOSBox worked, well, right out of the box. Mostly developed for gamers, it ran CALSOD without a hitch. Great job!

As for the font, I was able to find something called MS Line Draw (linedraw.tff) at this site. Using that along with Wordpad and a font size of 9, I was able to get the user’s guide to format properly. Yippee!

After all that, what does it look like? An old DOS program. A cool old DOS program though:


Simple graph ouput from CALSOD

Screen to configure CALSOD plots

Screen to configure CALSOD plots