Using FRD Tools: Impedance response

[DcibeL]

Hello,
I am using the FRD tools to design a set of speakers. I have followed the walk-through on RJBAudio, but I have a question.

RJB states that the measured impedance response be used when designing the crossover. So my question is what use is the box impedance response generated by Unibox? Does this impedance graph have any effect when designing a crossover that should be considered?

Thanks!

[Jay_WJ]

Hi DcibeL,

Unibox models impedance that depends on the box type you're using. For a two-way design, this doesn't affect your xo simulation. Perhaps this is for your Peerless 830884 8" project? If you can't find a good impedance plot with a high resolution you want to trace, try to model it using Jeff Bagby's Response Modeler. By matching it to the manufacturer's and Zaph's impedance plots, you can get smooth impedance curve without bothering to trace it.

BTW, in addition to Roman's guide, my notes may also be of some help: http://www.geocities.com/woove99/Spk...esigningXO.htm

-jAy

[DcibeL]

Thanks Jay. Yes, I've read over your notes as well. I've been re-educating myself on speaker design for the past few months, actually. So if I understand you correctly, the box impedance doesn't need to be taken into account unless the impedance peaks in it are in a frequency range that will affect the crossover. This is what I thought, but I just wanted to make sure since the box impedance at higher frequencies was different from the measured plot.

The link to the minimum phase response modeller on the FRD Consortium is dead.

Yes, this is for my Peerless 830884/810921 project. I have everything worked out (for the most part) except the crossover (save the best for last ). The cabinet will be a 2 cu.ft tower tuned to 31Hz. It will be fairly low sensitivity at only 84dB@1W (this is assuming 4dB of BSC). Maximum acoustic output should be around 98dB, since I run out of cone excursion at 35Hz with 22W input power.

[Jay_WJ]

Yes, I found Unibox impedance modeling not very accurate above bass range---impedance lower than measured. Use measured impedance (by either tracing or modeling) for your xo simulation.

Every link in my page seems to be working now. (It wasn't yesterday.)

[69stingray]

Similar question, and Jay I think it was on your website you mention combining the Unibox impedance with the SPL trace impedance, how do you do that? Just find a common point in the two point and combine in an excel file? Use Freq. Response Combiner? If so, how? Thanks again.



P.S.- Jay, do you have any comparison of a system designed 100% with software/freeware? Predicted vs. actual? I have always wondered how actual a system would be using SPL Trace, Unibox, BDS and PCD.

[Jay_WJ]

Originally posted by 69stingray
Similar question, and Jay I think it was on your website you mention combining the Unibox impedance with the SPL trace impedance, how do you do that? Just find a common point in the two point and combine in an excel file? Use Freq. Response Combiner? If so, how? Thanks again.

I scale Unibox-modeled impedance by adding constant ohms so that impedance (at the freq whre the curve slope becomes zero) is matched to free-air, measured impedance. You can use Excel for the scaling. And then splice it to the measured impedance using Notepad. You don't need these multiple steps if you use Jeff Bagby's Frequency Response Modeler to model a driver's impedance over the full frequency range.

P.S.- Jay, do you have any comparison of a system designed 100% with software/freeware? Predicted vs. actual? I have always wondered how actual a system would be using SPL Trace, Unibox, BDS and PCD.

Go to the appendix of my xo design notes.

Also, see this post: http://www.diyaudio.com/forums/showt...84#post1312484

[DcibeL]

Originally posted by Jay_WJ
Unibox models impedance that depends on the box type you're using. For a two-way design, this doesn't affect your xo simulation. Perhaps this is for your Peerless 830884 8" project? If you can't find a good impedance plot with a high resolution you want to trace, try to model it using Jeff Bagby's Response Modeler. By matching it to the manufacturer's and Zaph's impedance plots, you can get smooth impedance curve without bothering to trace it.

Hi Jay_WJ, sorry to bother you again but maybe you can help me with this. I am trying to use the response modeler to match the manufacturer's/Zaph's impedance plots (Zaph has a zoomed in impedance plot for the 830884), however when I enter the Le of 0.7mH (what Zaph measured), the impedance only matches up to 1kHz. Above 1kHz, they are unmatched. If I use the manufacturer's Le of 1.4mH the situation only gets worse.

See attached image of the impedance model overlayed on top of a quick trace I did of the manufacturer's impedance plot (note I also compared Zaph's plot which is very similar). The modeled impedance is blue, the trace is grey.

[Jay_WJ]

I forgot to add this. The use of only Re and Le values cannot model a driver's impedance accurately enough to match the measured impedance curve. You have to vary the other parameters under "Driver Voice Coil Impedance Parameters" to match the modeled curve with the measured curve. The reason is that a driver's impedance profile is the result of its unique motor design, which cannot be predicted by Re and Le alone. I edited my notes accordingly.

I'll model the 830884's impedance. Let's compare ours after I'm done. I'll post it quickly.

[Jay_WJ]

Here's my modeling result of the 830884's impedance. Take a look at the voice coil parameter values I used. The 830884's impedance is very linear thanks to its good motor design. That's why we need some extreme parameter values to model it.

P.S. Previously attached figure was incorrect. Now uploaded again.

[DcibeL]

The reason I asked is that I was having trouble coming up with a combination of parts that worked well for impedance equalization (zobel). The standard calculation, which relys on Le and Re doesn't work well in this case, because as you say the impedance rise is dependant on the motor design and not just Le alone.

So crossover design has become mostly trial and error. I was using Speaker Workshop, but have found Passive Crossover Designer to work very well for quickly seeing the result of changing certain parts in the circuit.

I am designing two crossovers. One will consist of many parts to get near perfect response. I then try to reproduce the same response using as few components as possible. Then I'll have to decide whether the extra cost of the complex crossover is worth it.

Thanks for your help. Wish me luck