Hi Davey,
It's easier for me to run a capture of the electrical repsonses (when my two year old tearaway can't interefere!), and then post-process the file at my leasure. That way I can also run a different analysis using the same input data.
The amplitude scale was set like that to ensure a gridline occured at -3dB. - For the benefit you guys mainly!
I'll try playing with the FFT size as you suggest.
Thanks for all the input so far - this is educational and fascinating to me...
Steve
It's easier for me to run a capture of the electrical repsonses (when my two year old tearaway can't interefere!), and then post-process the file at my leasure. That way I can also run a different analysis using the same input data.
The amplitude scale was set like that to ensure a gridline occured at -3dB. - For the benefit you guys mainly!
I'll try playing with the FFT size as you suggest.
Thanks for all the input so far - this is educational and fascinating to me...
Steve
Davey said:
I'm interested to see how you work out the transfer function of the crossover from those results. It looks very complicated to me, when you consider the impedence of the speakers.
If you need to know what the individual networks are doing, in order to replicate them actively, why not measure the acoustic response of the drive unit alone, then together with its crossover. Subtracting the two would surely give the response required of the active crossover for that drive unit. Repeat for the other drive units. You have to open the box and seperate the signal paths, but it looks like you will have to do that anyway.
If you need to know what the individual networks are doing, in order to replicate them actively, why not measure the acoustic response of the drive unit alone, then together with its crossover. Subtracting the two would surely give the response required of the active crossover for that drive unit. Repeat for the other drive units. You have to open the box and seperate the signal paths, but it looks like you will have to do that anyway.
johnnyx said:I'm interested to see how you work out the transfer function of the crossover from those results. It looks very complicated to me, when you consider the impedence of the speakers.
If you need to know what the individual networks are doing, in order to replicate them actively, why not measure the acoustic response of the drive unit alone, then together with its crossover. Subtracting the two would surely give the response required of the active crossover for that drive unit. Repeat for the other drive units. You have to open the box and seperate the signal paths, but it looks like you will have to do that anyway.
Hi Johnny,
I agree with you that at this point we need to know what the drivers' acoustic response is. The only problem is that I don't have mic. with a flat response (or even known response) for taking acoustic measurements.
It's now a high priority on my shopping list though! I think BrianGT might be doing a Group Buy for pre-built wands with a mic pre-amp kit, so I might try and get in on that.
Steve
Johnny/Steve,
You're thinking this is more complicated than it really is...it's very simple. The result you see IS the transfer function. The drivers are still connected to the filters so the impedance effects are included in the measurement. In order to replicate them actively all that is necessary is to match those curves for amplitude/phase and you're done.
In an active setup the drivers are connected directly to the power amps which are essentially true voltage sources so they're not effected by nominal impedance swings of the raw drivers.
There are no acoustic measurements required or needed to "reverse engineer" a crossover in this way. If you want to go further and improve or change the basic design then, yes, you would need to start taking some acoustic measurements.
Cheers,
Davey.
You're thinking this is more complicated than it really is...it's very simple.
The result you see IS the transfer function. The drivers are still connected to the filters so the impedance effects are included in the measurement. In order to replicate them actively all that is necessary is to match those curves for amplitude/phase and you're done.In an active setup the drivers are connected directly to the power amps which are essentially true voltage sources so they're not effected by nominal impedance swings of the raw drivers.
There are no acoustic measurements required or needed to "reverse engineer" a crossover in this way. If you want to go further and improve or change the basic design then, yes, you would need to start taking some acoustic measurements.
Cheers,
Davey.
Hi Davey,
Of course you're right, I was just getting ahead of myself!
I'm planning upon implementing the filters in the DSP of a PC soundcard, and have had a rough stab at reproducing the mid's slopes. (Image attached)
It's not safe to leave CD's around near my two year old (for the CD's, not for him!), so the whole lot's been converted to extra high quality MP3s, on a server in my workshop. My first stage is to built a test rig with an old P3 500, and use an inexpensive soundcard. If I can justify the full project, it'll end up as a near silent PCwith a better soundcard. I already built a 6 channel Gainclone Amp - it's setup for HT at the moment though.
Steve
Of course you're right, I was just getting ahead of myself!
I'm planning upon implementing the filters in the DSP of a PC soundcard, and have had a rough stab at reproducing the mid's slopes. (Image attached)
It's not safe to leave CD's around near my two year old (for the CD's, not for him!), so the whole lot's been converted to extra high quality MP3s, on a server in my workshop. My first stage is to built a test rig with an old P3 500, and use an inexpensive soundcard. If I can justify the full project, it'll end up as a near silent PCwith a better soundcard. I already built a 6 channel Gainclone Amp - it's setup for HT at the moment though.
Steve
Attachments
BlackDog said:
Are you familiar with baffle step compensation? Here's a discussion, see figure 1 where the response falls by about 6 dB from 2 kHz down to 100 Hz:
http://sound.westhost.com/bafflestep.htm
Steve,
Notice that the woofer filter response slopes down from 100 Hz, and the mid slopes up from 2 kHz, also notice where the mid and woofer cross, if it is an in phase network they'll sum by 6 dB not just 3. If you move up 6 dB from the point where they cross the summed point falls on a line connecting the 100 Hz response and the slope of the midrange. It looks about right for a crossover that compensates for the baffle step.
The mid to tweeter is less clear, there are probably some acoustic effects going on and the tweeter doesn't seem to have any flat passband. Some tilt up the high end to make up for the increasing directivity of the dome and provide for more of a flat power response than on axis frequency response but it's hard to tell without seeing the acoustic response of the drivers and system. That's a steep slope on the tweeter.
Did you put the drivers back in the box so that they were acoustically loaded by the box when you took the measurements?
If you did this right then you do have the transfer functions as loaded by the drivers you only have to take acoustical measurements if you want to change these transfer functions.
Pete B.
PB2 said:
Hi Pete,
I see what you mean. Yes, I had heard of baffle step comp., but didn't fully understand what's required. Now I do. (I think!).
I didn't screw the drivers back in to take measurements. I just had them resting on the edge of the hole in the baffle, with the weight of the magnets keeping them in. I'll have to make up a test cable with VERY thin wire if I need to measure with the drivers screwed back!
Yep, I also noticed the steep tweeter plot!
I think I'm just starting to realise how involved designing a crossover can be! Doing things digitallly should speed things up in this case - I hope....
Steve
BlackDog said:
Hi Steve,
Digital XO filters make it easy to quickly experiment and a measurement system will help if you decide to completely redo the crossover. There's another possible reason for the tipped up bass and high end which is to take into account the Equal Level Loudness contours or Fletcher Munsen curves.
http://www.aist.go.jp/aist_e/latest_research/2003/20031114/20031114.html
http://www.webervst.com/fm.htm
You'll hear many speak of how important flat frequency response is in a speaker, yet if you don't listen at the same levels as the original event (few home systems can come close to even an acousitc event) there is no way it is going to sound accurate. It's no wonder that subjectivists tend to select speakers with non flat frequency response, they're probably correcting for the difference in listening level. Many rave about ProAc speakers. I've never heard them by the way just offering them as an example of non flat response, see figures 3 and 4:
http://stereophile.com/loudspeakerreviews/297/index6.html
You should also keep in mind what's been called the BBC dip, see section H:
http://www.linkwitzlab.com/models.htm
Perhaps the industry should take a fresh look at line level loudness compensation.
Regards,
Pete B.
Now the funs begins
Hi all,
I thought I'd ask you crossover experts once again, as you were so helpful with my last questions.
I've just pulled the passive crossovers out of my speakers. If you can remember, I will be be tri-amping them, and using a sound card DSP to run digital crossovers.
I've heard that it's very easy to blow tweeters by feeding them low frequencies or DC, so I thought I would use a cap. to protect the tweeters at least for now. - which is where my problem lies.
How do I calculate what is a sensible value cap. to protect the tweeters?? -- The DC resistance of the tweeter is approx. 7ohms, but I'm sure that's not really relevant!
thanks in advance,
Steve
Hi all,
I thought I'd ask you crossover experts once again, as you were so helpful with my last questions.
I've just pulled the passive crossovers out of my speakers. If you can remember, I will be be tri-amping them, and using a sound card DSP to run digital crossovers.
I've heard that it's very easy to blow tweeters by feeding them low frequencies or DC, so I thought I would use a cap. to protect the tweeters at least for now. - which is where my problem lies.
How do I calculate what is a sensible value cap. to protect the tweeters?? -- The DC resistance of the tweeter is approx. 7ohms, but I'm sure that's not really relevant!
thanks in advance,
Steve
Steve,
I think your tweeter is probably not in much danger for normal usage, but during testing it's possible to heat/damage it with MLS bursts, etc. I would make the capacitor value something that corresponds roughly to your crossover frequency and then let the DSP crossover perform the rest of the filtering duties. For a typical dome tweeter I think a value of 15-20uF would do the job fine. You could optimize the value later.
Ultimately, if you feel your power amps are well behaved, you could eliminate the capacitor entirely.
Davey.
I think your tweeter is probably not in much danger for normal usage, but during testing it's possible to heat/damage it with MLS bursts, etc. I would make the capacitor value something that corresponds roughly to your crossover frequency and then let the DSP crossover perform the rest of the filtering duties. For a typical dome tweeter I think a value of 15-20uF would do the job fine. You could optimize the value later.
Ultimately, if you feel your power amps are well behaved, you could eliminate the capacitor entirely.
Davey.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.