What crossover type ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I know I spliced the nearfield in about the right place, but I didn't know what levels to choose.


That, in a nutshell, is the problem with nearfield measurements!

There is a theoretical approach that says, if you know the driver diameter (and port diameter if present). then you can calculate the relationship between the nearfield level and the farfield level. Most people do what you have done which is to shift the curve until it looks about right, or compares well enough to the in-room curve. Nearfield splicing is an approximate approach anyhow.

You also have a fair amount of error between your model and your measured response. Debugging that usually means comparing the simulated sections to measurements of woofer + network and tweeter + network. Also comparing simulated network voltage and measured network voltage. Reversing polarity, both of your simulation and your built up system, will also give some clues to the discrepancies.

Good Luck,
David S.
 
This has all been a bit vaguely described right from the start! :)

The bass unit is the weird foam-surround 8" Vifa M21WG-09-08 with an abominable Qts of 0.7 and 0.9mH inductance, Re of 5.6 ohms, SPL 90dB used in the Heybrook HB1. Sreten doesn't usually say nice things about anything, but this one HE ABSOLUTELY HATES! And he's right! :D

An externally hosted image should be here but it was not working when we last tested it.


Vifa M21WG-09-08

Combined with the wave-guided and ferrofluid-stripped Monacor DT-300 tweeter (Why?), I'd say this is a marriage made in Hell, especially on a simple crossover. How on earth can you fix this? I really don't know. The woofer SPL hump of 5dB between 1kHz and 4kHz is enough to put me off.

Monacor DT-300
 
Last edited:
That, in a nutshell, is the problem with nearfield measurements!

There is a theoretical approach that says, if you know the driver diameter (and port diameter if present). then you can calculate the relationship between the nearfield level and the farfield level. Most people do what you have done which is to shift the curve until it looks about right, or compares well enough to the in-room curve. Nearfield splicing is an approximate approach anyhow..

Thanks for relying Dave. It's good to hear my splicing method is considered ok'ish.

You also have a fair amount of error between your model and your measured response. Debugging that usually means comparing the simulated sections to measurements of woofer + network and tweeter + network. Also comparing simulated network voltage and measured network voltage. Reversing polarity, both of your simulation and your built up system, will also give some clues to the discrepancies.

Good Luck,
David S.

The difference between the measured and simulated response has been bugging me. Do you know if it's a problem with LspCAD or something i've done wrong?

So if I understand you correctly, I need to measure the woofer + crossover, without the tweeter connected and vice-versa with the tweeter + crossover. Do I measure the DCR of the tweeter (and woofer when measuring the tweeter + crossover) and replace it with a fixed resistor or something? (remember i'm using a series crossover).
 
Oops, I forgot about the series network issue. Replacing a driver with its resistance wouldn't really be accurate enough for determining if you have errors in a section. Put the woofer in another room to measure the tweeter (and vice versa)?

As a start, can you measure what is usually called terminal volts? This is the electrical curve of the network as loaded by the driver. I assume your simulation software would allow you to see that as a curve to compare too. Note that you might have to change your amplifier connections between measurements because you will need to measure between each side in the middle and you will probably need to ground each side to take the measurement. (Tread carefully or your amplifier will be unhappy!)

I haven't used LspCAD, but most simulators are accurate enough that errors are usually from driver measurements or network values that are off.

David S.
 
This has all been a bit vaguely described right from the start! :)

That's me all over. My brains wired up wrong :)

The bass unit is the weird foam-surround 8" Vifa M21WG-09-08 with an abominable Qts of 0.7 and 0.9mH inductance, Re of 5.6 ohms, SPL 90dB used in the Heybrook HB1. Sreten doesn't usually say nice things about anything, but this one HE ABSOLUTELY HATES! And he's right! :D

An externally hosted image should be here but it was not working when we last tested it.


Vifa M21WG-09-08

Combined with the wave-guided and ferrofluid-stripped Monacor DT-300 tweeter (Why?), I'd say this is a marriage made in Hell, especially on a simple crossover. How on earth can you fix this? I really don't know. The woofer SPL hump of 5dB between 1kHz and 4kHz is enough to put me off.

Monacor DT-300

I've noticed Streten can be a touch negative, he seems to know his stuff though.

If you look at the raw measurements (post 17), I don't have a big hump between 1k and 4k.

"Marriage made in Hell" I think you'd change your mind if you heard them :)
 
Last edited:
Oops, I forgot about the series network issue. Replacing a driver with its resistance wouldn't really be accurate enough for determining if you have errors in a section. Put the woofer in another room to measure the tweeter (and vice versa)?

That's possible I suppose - It would be a right pain in the **** though.

As a start, can you measure what is usually called terminal volts? This is the electrical curve of the network as loaded by the driver. I assume your simulation software would allow you to see that as a curve to compare too. Note that you might have to change your amplifier connections between measurements because you will need to measure between each side in the middle and you will probably need to ground each side to take the measurement. (Tread carefully or your amplifier will be unhappy!)

This ones new to me. I don't suppose you have any links I could look at, or should I just google "Terminal volts" or something?


I haven't used LspCAD, but most simulators are accurate enough that errors are usually from driver measurements or network values that are off.

My tweeter measurement looks very similar to ones i've seem online, so I think that's ok. Im not sure about the woofer though. Although I think it looks quite similar to the manufacterers measurements. I use a WT3 to measure impedance - I don't know how accurate it is though.
 
diyAudio Moderator
Joined 2008
Paid Member
What do you think? Also has anyone got any good tips for getting the measurements right?
Have you considered taking your measurements without a crossover so you can use the data in your simulator? These multi angle measurements could be used to select the best crossover frequency based on the driver directivity, and could then be 'averaged' for a plot of total sound into the room. This would also help you chose the best listening axis.

If you can't get your measurements to work below 400Hz, it may not matter. If you've considered your baffle conditions and chosen a target level, then you can optimise 400Hz and up for that level and it should be workable.
 
Have you considered taking your measurements without a crossover so you can use the data in your simulator? These multi angle measurements could be used to select the best crossover frequency based on the driver directivity, and could then be 'averaged' for a plot of total sound into the room. This would also help you chose the best listening axis.

If you can't get your measurements to work below 400Hz, it may not matter. If you've considered your baffle conditions and chosen a target level, then you can optimise 400Hz and up for that level and it should be workable.

Sounds like a good idea, but i'm not sure how i'd go about doing it without a step by step guide or something.
 
diyAudio Moderator
Joined 2008
Paid Member
If you take a run of plots for each driver, we can do some of the legwork and probably set up a kind of best fit scenario. It may help if you include a few details about your cabinet as well.

Be careful with the tweeter. Some are best done with a capacitor large enough not to affect anything. If you can get your measurements out of the noise floor without straining the tweeter, the woofer should be taken at the same levels but they are often less sensitive, so just do the best you can.
 
If you take a run of plots for each driver, we can do some of the legwork and probably set up a kind of best fit scenario. It may help if you include a few details about your cabinet as well.

Be careful with the tweeter. Some are best done with a capacitor large enough not to affect anything. If you can get your measurements out of the noise floor without straining the tweeter, the woofer should be taken at the same levels but they are often less sensitive, so just do the best you can.

This is how I do my measurements (I do mine indoors though).

Would you like me to take horizontal, 0, 15, 30, and 45 degree off-axis measurements of the woofer and tweeter, with the mic at the same height as the tweeter at all times? and at the same level?

Would you like any other measurements or would that be enough?
 
diyAudio Moderator
Joined 2008
Paid Member
If you are going to use the ground plane technique you'll need to keep your mic on the ground (or on the reflective, non-resonant surface you've placed there). There will be some difference between the height of the tweeter and woofer and therefore from each to the mic. It is better if they are the same at each measurement, and simpler if you didn't need to move anything between driver measurements. If it comes to tilting the baffle to achieve that, there may be some ambiguity with the effect of the baffle as you rotate the speaker. Your mileage may vary, but raising both the driver of interest and your mic to half the room height and gating your measurement electronically to eliminate the reflections is a second option.

Running your measurements out to 60 or even 90 degrees could have benefits. They may be farther from the listening axis but statistically they cover more of the sound field, and help to outline the directional demarcation between the drivers. To what extent you want to take this is ultimately your decision and whatever you can offer will be useful.
 
I'd like to do the ground plane technique, but I don't have anywhere to do it.

I've got to admit, i'm not sure I can be bothered to take any more measurements of this design. When I start my new 3-way project, i'll post details here - Perhaps I could upload the measurements for someone to play with? Is that possible?
 
fatmarley, I have been giving this some thought. I really think you need second order crossover here. You need a Zobel on the woofer, which was Le=0.9mH IIRC at whatever Re your woofer has. Le/Cz = Re squared for the Zobel which is Re in series with Cz.

These Rod Elliott series values are for butterworth slopes at 1000Hz and 8 ohms so need scaling for your values which are probably 3000Hz and 6 ohms, and you can include bafflestep as a simple RL circuit in series I'd reckon, due to the constant impedance property of butterworth.

An externally hosted image should be here but it was not working when we last tested it.


What have you got to lose? :)
 
Would this be any good?

Parallelcrossover150oldmeasurements23mH.jpg


I made a mistake with the crossover diagram I uploaded before. The inductor should have been 2.3mH not 2.7 (forgot I unwound it)
 
diyAudio Moderator
Joined 2008
Paid Member
Flattening the impedance just for your amp 'might' be a good thing ie: it shouldn't matter and it only sometimes does. I may occasionally throw something over it like you have here, then forget about it for several iterations of the crossover before I'd tweak it back into shape.

So is your proposed crossover better? I don't know, please tell us how you think it is better. Did you use better data than before and how is it better?
 
Flattening the impedance just for your amp 'might' be a good thing ie: it shouldn't matter and it only sometimes does. I may occasionally throw something over it like you have here, then forget about it for several iterations of the crossover before I'd tweak it back into shape.

I guess the sensible thing to do, would be to try it with and without to see what sounds better.

So is your proposed crossover better? I don't know, please tell us how you think it is better. Did you use better data than before and how is it better?

When I first tried a second-order (electrical) crossover, I was trying to get the frequency response as flat as possible and assumed by getting the response flat, I was adding BSC. I tried loads of different crossovers, but they all sounded terrible.

After lots of messing about with the AR-XO I got the BSC right (for my room) and a flat response through the mid and treble (using measurements plus my ears).
To get where I am with the second-order (electrical) crossover - All i've done is copy the frequency response magnitude of the AR-XO to the second order and made sure the phase tracking was aligned through the crossover.

I'm quite tempted to try it and report back with measurements and a subjective opinion. Assuming anyone's interested?

EDIT: I used the same data as before because I didn't want to confuse matters with new measurements. Although the predicted frequency response in LspCAD doesn't match with what i'm measuring, the reverse null measures as predicted, so it seems not everything is wrong.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.