Is it possible that two paralleled, identical drivers in one sealed box influence each other heavily?
I have put two Scan 21W8555-01 together into one 70 l sealed box. The box does contains close to no damping material because I will be using a 100 Hz 3rd order cutoff. This may not be the optimal enclosure for these drivers, but there were some SAF constraints to be met. And I was going to use a Linkwitz transform to obtain another octave.
Anyway, I measured them with Speaker Workshop, using a test rig with an LM3886 amp and a 7.5 R resistor for impedance measurements.
On my first try, although I had set the input level correctly, the input was overdriven. I diconnected the speaker, chose more input attenuation, recalibrated the rig and remeasured the speaker. This time, the signal was not very loud. I got no overdrive, but a double resonance. I figured one of the cable pairs was not connected correctly, so I was in fact measuring a BR configuration with a passive radiator. The only thing that still keeps puzzling me was that the impedance at 5 Hz, which is usually within 20% of the DC value I measure with a multimeter, was at 4 R, about the same as in the following measurement, where it should have been 6 R.
Ok, I carefully reinserted both cable pairs into the connectors. This time, the signal was as loud as in the first experiment, and I checked that both drivers were doing their job. This time I got a single hump at 55 Hz, with Q_m=6 and Q_t=0.23, roughly as calculated. But the left channel (voltage divider 7.5 R to speaker) was overdriven, and the impedance curve had nasty spikes, especially around resonance.
I figured that HF parts of the MLS signal bounce around uninhibited inside the enclosure, hit the other driver and hence generate some voltage on the 7.5 R reference resistor. The spikes are higher around resonance because the other driver is also at high impedance there, whereas at other frequencies, it will shunt the spikes to ground more efficiently.
Now my questions:
- The spikes seem to fool the fit algorithm. When comparing measured and fit curves, I would place the impedance maximum a few Hz lower and choose a higher Q_m. Now, how do I get a decent measurement?
- Is the effect really relevant soundwise? I will be using a 100 Hz low pass, and both drivers will be connected in parallel to an amp with low damping factor.
I have put two Scan 21W8555-01 together into one 70 l sealed box. The box does contains close to no damping material because I will be using a 100 Hz 3rd order cutoff. This may not be the optimal enclosure for these drivers, but there were some SAF constraints to be met. And I was going to use a Linkwitz transform to obtain another octave.
Anyway, I measured them with Speaker Workshop, using a test rig with an LM3886 amp and a 7.5 R resistor for impedance measurements.
On my first try, although I had set the input level correctly, the input was overdriven. I diconnected the speaker, chose more input attenuation, recalibrated the rig and remeasured the speaker. This time, the signal was not very loud. I got no overdrive, but a double resonance. I figured one of the cable pairs was not connected correctly, so I was in fact measuring a BR configuration with a passive radiator. The only thing that still keeps puzzling me was that the impedance at 5 Hz, which is usually within 20% of the DC value I measure with a multimeter, was at 4 R, about the same as in the following measurement, where it should have been 6 R.
Ok, I carefully reinserted both cable pairs into the connectors. This time, the signal was as loud as in the first experiment, and I checked that both drivers were doing their job. This time I got a single hump at 55 Hz, with Q_m=6 and Q_t=0.23, roughly as calculated. But the left channel (voltage divider 7.5 R to speaker) was overdriven, and the impedance curve had nasty spikes, especially around resonance.
I figured that HF parts of the MLS signal bounce around uninhibited inside the enclosure, hit the other driver and hence generate some voltage on the 7.5 R reference resistor. The spikes are higher around resonance because the other driver is also at high impedance there, whereas at other frequencies, it will shunt the spikes to ground more efficiently.
Now my questions:
- The spikes seem to fool the fit algorithm. When comparing measured and fit curves, I would place the impedance maximum a few Hz lower and choose a higher Q_m. Now, how do I get a decent measurement?
- Is the effect really relevant soundwise? I will be using a 100 Hz low pass, and both drivers will be connected in parallel to an amp with low damping factor.
First of all: It is never bad to use separate compartments for each driver, since they are never exactly equal.
Did I get it right that you didn't connect the two drivers in parallel but drive them both with an own "reference" amplifier and series resistor, but take the measurement voltage only from one of them ?
Regards
Charles
Did I get it right that you didn't connect the two drivers in parallel but drive them both with an own "reference" amplifier and series resistor, but take the measurement voltage only from one of them ?
Regards
Charles
Well, the raw drivers measured similarly enough.
In measurements #1 and #3 (both of which were spiky and overdriven), the drivers were in parallel, connected to one amp via one reference resistor.
In #2, it seems only one driver was connected, even if still I don't understand why the impedance was roughly 3.8 Ohm at 5 Hz in both cases.
Yes, I could use a separator, but due to mechanical constraints, the volumes would not be equal. Ok ok, I can use separate amps and separate Linkwitz shifters, but I would like to avoid this if it is not absolutely necessary.
Regards,
Eric
In measurements #1 and #3 (both of which were spiky and overdriven), the drivers were in parallel, connected to one amp via one reference resistor.
In #2, it seems only one driver was connected, even if still I don't understand why the impedance was roughly 3.8 Ohm at 5 Hz in both cases.
Yes, I could use a separator, but due to mechanical constraints, the volumes would not be equal. Ok ok, I can use separate amps and separate Linkwitz shifters, but I would like to avoid this if it is not absolutely necessary.
Regards,
Eric
You can of course use one LTF and one amp.
I was just wondering whether you made the measurements with two amps and two series resistors (which wouldn't be suggested) because of your measurement problems.
Do you intend to run the LF part with no stuffing at all? I would strongly advise to stuff since no crossover is sufficiently steep to keep the higher frequencies, that can cause trouble, completely off the drivers.
Regards
Charles
I was just wondering whether you made the measurements with two amps and two series resistors (which wouldn't be suggested) because of your measurement problems.
Do you intend to run the LF part with no stuffing at all? I would strongly advise to stuff since no crossover is sufficiently steep to keep the higher frequencies, that can cause trouble, completely off the drivers.
Regards
Charles
Well, I have stuffed the lower 10 cm of the box (below the last brace) with mineral wool. This would attenuate longitudal modes if I ever were to use it that high (or the filter is indeed not steep enough). I could probably also glue some wood felt onto a brace that is in the center, between the two drivers.
Any idea how I could arrive at a decent measurement?
What about sound effects? After all, there are even highly regarded commercial designs with a passive XO (i.e. lower damping factor) that use the woofers in parallel and without a divider.
Cheers,
Eric
Any idea how I could arrive at a decent measurement?
What about sound effects? After all, there are even highly regarded commercial designs with a passive XO (i.e. lower damping factor) that use the woofers in parallel and without a divider.
Cheers,
Eric
Surely the test resistor should be on the inputs to the enclosure, if you are trying to measure the whole system, rather than just one driver?
Both drivers were paralleled and connected to one amp via ONE test resistor. Just the regular test setup except that two drivers were connected to the output of the test rig.
Regards,
Eric
Regards,
Eric
Hi Eric,
I have no explination for the spikey signal except for possible a bad connection. You should expect driver impedence to be close to Re at very low frequencies. You might want to check eack speaker individually to be sure they are all ok.
I ran your speakers using Bass Box Pro. I used a 70l box with the drivers in parallel. This should be a very well behaved arrangement. Qt is ~.5 with an f3 of 63Hz and is flat to about 1kHz where it rolls off at 12db/oct.
The difference in my simulation is that my box has heavy stuffing. I use dual JBL LE10A's in 60l closed boxes running in parallel. They work great. I would stuff the boxes.
I have no explination for the spikey signal except for possible a bad connection. You should expect driver impedence to be close to Re at very low frequencies. You might want to check eack speaker individually to be sure they are all ok.
I ran your speakers using Bass Box Pro. I used a 70l box with the drivers in parallel. This should be a very well behaved arrangement. Qt is ~.5 with an f3 of 63Hz and is flat to about 1kHz where it rolls off at 12db/oct.
The difference in my simulation is that my box has heavy stuffing. I use dual JBL LE10A's in 60l closed boxes running in parallel. They work great. I would stuff the boxes.
>Is it possible that two paralleled, identical drivers in one sealed box influence each other heavily?
====
Of course! In a sealed design, if one draws more current it can be dominant enough to turn the other into a highly reactive 'passive' radiator. In vented, I've watched one driver over-excurse while the other was barely moving due to one being mistuned. These cases were either in prosound apps or being pushed to their limits playing 16Hz organ pipes or special effects so isn't an issue with typical HIFI apps beyond increased distortion, which probably isn't enough to be distracting.
====
>Ok, I carefully reinserted both cable pairs into the connectors. This time, the signal was as loud as in the first experiment, and I checked that both drivers were doing their job. This time I got a single hump at 55 Hz, with Q_m=6 and Q_t=0.23, roughly as calculated. But the left channel (voltage divider 7.5 R to speaker) was overdriven, and the impedance curve had nasty spikes, especially around resonance.
====
Hmm, if I understand you correctly, this must be a huge cab to get a Qt = 0.23. Also, if the correct voltage divider for a single driver is 7.5ohms, then it seems reasonable to me it should be 3.75ohms for parallel.
====
>Is the effect really relevant soundwise? I will be using a 100 Hz low pass, and both drivers will be connected in parallel to an amp with low damping factor
====
The primary goal of any XO is to get the driver's HF BW down to -35dB where we can no longer sense its effect on the rest of the signal's BW, so the name of the game is to do wide BW measurements to find out how well the XO point/slope is doing its job, not to mention the effects of the XO on the speaker's resonant performance if passive.
GM
====
Of course! In a sealed design, if one draws more current it can be dominant enough to turn the other into a highly reactive 'passive' radiator. In vented, I've watched one driver over-excurse while the other was barely moving due to one being mistuned. These cases were either in prosound apps or being pushed to their limits playing 16Hz organ pipes or special effects so isn't an issue with typical HIFI apps beyond increased distortion, which probably isn't enough to be distracting.
====
>Ok, I carefully reinserted both cable pairs into the connectors. This time, the signal was as loud as in the first experiment, and I checked that both drivers were doing their job. This time I got a single hump at 55 Hz, with Q_m=6 and Q_t=0.23, roughly as calculated. But the left channel (voltage divider 7.5 R to speaker) was overdriven, and the impedance curve had nasty spikes, especially around resonance.
====
Hmm, if I understand you correctly, this must be a huge cab to get a Qt = 0.23. Also, if the correct voltage divider for a single driver is 7.5ohms, then it seems reasonable to me it should be 3.75ohms for parallel.
====
>Is the effect really relevant soundwise? I will be using a 100 Hz low pass, and both drivers will be connected in parallel to an amp with low damping factor
====
The primary goal of any XO is to get the driver's HF BW down to -35dB where we can no longer sense its effect on the rest of the signal's BW, so the name of the game is to do wide BW measurements to find out how well the XO point/slope is doing its job, not to mention the effects of the XO on the speaker's resonant performance if passive.
GM
Thanks for all the answers, guys. I will redo the test with all connections double-checked, but I doubt there is a problem, after all test #2 looked ok.
Yeah, I noticed that one when I went back to my simulation sheet. Fs was where I expected it, Qt should have been around .5 as Rodd pointed out. There is nothing surprising about this low number, though. The fit simply went wrong because of all the spikes.
The reference resistor is simply chosen to be on the same order of magnitude as the impedances I get to measure (which range from 3 to 50 Ohms if the resonance is included). It is perfectly ok for both 4 and 8 Ohm speakers. This arrangement is used by all soundcard based systems I know of.
Regards,
Eric
GM said:
Yeah, I noticed that one when I went back to my simulation sheet. Fs was where I expected it, Qt should have been around .5 as Rodd pointed out. There is nothing surprising about this low number, though. The fit simply went wrong because of all the spikes.
The reference resistor is simply chosen to be on the same order of magnitude as the impedances I get to measure (which range from 3 to 50 Ohms if the resonance is included). It is perfectly ok for both 4 and 8 Ohm speakers. This arrangement is used by all soundcard based systems I know of.
Regards,
Eric
partial solution
Ok, here?s a partial explanation:
1) At 48 kHz sample rate, the spikes were less pronounced. Apparently, the input filter of the soundcard is fixed at > 20 kHz. At 6 kHz sample rate, any kind of noise pickup will cause aliasing.
2) When I recalibrated, I sometimes got strange values for the reference resistor. It turned out that inside the rig box, a ground connection of the input attenuator was kind of bad. When it was open, it would pick up dirt from the output leads.
OK, here are some reproducible measurements:
Both drivers in parallel:
50.965 Hz Q 2.388 / 0.896 / 0.652
Q_m is not really as low as the fit suggest. In fact, it is a double peak, consisting of a major peak at about 58 Hz and a slightly weaker one at 43 Hz. So where does this double peak come from?
Each driver measured separately, the other one was shorted (this is in fact a highly damped passive radiator):
top 33.008 Hz 1.745/ 0.661 / 0.479
bottom 34.469 Hz 1.777 / 0.611 / 0.455
The TSP are still in good agreement, albeit slightly less than the free air measurements. Maybe this is because the top driver is about 3 cm away from the top whereas the lower driver has about 20 cm to the bottom.
Measured separately, other driver open (passive radiator):
top 24.000 Hz 9.953 / 1.004 / 0.912
bottom 23.988 Hz 9.559 / 0.964 / 0.876
This is fit is not entirely correct because there is the secondary bass reflex maximum, but agreement between simulated fit curve and measured curve is pretty goog around the main maximum..
OK, now for the remaining questions:
- why do I get a double maximum?
- Will I still have this double maximum when the paralleled drivers are driven from 10 mOhm rather than 7.5 Ohms?
- If not, it is ok not to insert a separator. But how do I determine the Q for the Linkwitz circuit?
Ok, here?s a partial explanation:
1) At 48 kHz sample rate, the spikes were less pronounced. Apparently, the input filter of the soundcard is fixed at > 20 kHz. At 6 kHz sample rate, any kind of noise pickup will cause aliasing.
2) When I recalibrated, I sometimes got strange values for the reference resistor. It turned out that inside the rig box, a ground connection of the input attenuator was kind of bad. When it was open, it would pick up dirt from the output leads.
OK, here are some reproducible measurements:
Both drivers in parallel:
50.965 Hz Q 2.388 / 0.896 / 0.652
Q_m is not really as low as the fit suggest. In fact, it is a double peak, consisting of a major peak at about 58 Hz and a slightly weaker one at 43 Hz. So where does this double peak come from?
Each driver measured separately, the other one was shorted (this is in fact a highly damped passive radiator):
top 33.008 Hz 1.745/ 0.661 / 0.479
bottom 34.469 Hz 1.777 / 0.611 / 0.455
The TSP are still in good agreement, albeit slightly less than the free air measurements. Maybe this is because the top driver is about 3 cm away from the top whereas the lower driver has about 20 cm to the bottom.
Measured separately, other driver open (passive radiator):
top 24.000 Hz 9.953 / 1.004 / 0.912
bottom 23.988 Hz 9.559 / 0.964 / 0.876
This is fit is not entirely correct because there is the secondary bass reflex maximum, but agreement between simulated fit curve and measured curve is pretty goog around the main maximum..
OK, now for the remaining questions:
- why do I get a double maximum?
- Will I still have this double maximum when the paralleled drivers are driven from 10 mOhm rather than 7.5 Ohms?
- If not, it is ok not to insert a separator. But how do I determine the Q for the Linkwitz circuit?
>The reference resistor is simply chosen to be on the same order of magnitude as the impedances I get to measure (which range from 3 to 50 Ohms if the resonance is included). It is perfectly ok for both 4 and 8 Ohm speakers. This arrangement is used by all soundcard based systems I know of.
====
OK, thanks for the info, never used a soundcard based system.
====
>- why do I get a double maximum?
====
Well, the driver/boundary proximity certainly is an issue. You can sim each driver location in MJK's sealed program to find out approximately how much effect there is.
====
>- Will I still have this double maximum when the paralleled drivers are driven from 10 mOhm rather than 7.5 Ohms?
====
I don't see why not.
====
>- If not, it is ok not to insert a separator. But how do I determine the Q for the Linkwitz circuit?
====
Good question, I would be prone to take the average and skip the divider if drivers share the load fairly evenly, but might want to ask 'the man' himself.
GM
====
OK, thanks for the info, never used a soundcard based system.
====
>- why do I get a double maximum?
====
Well, the driver/boundary proximity certainly is an issue. You can sim each driver location in MJK's sealed program to find out approximately how much effect there is.
====
>- Will I still have this double maximum when the paralleled drivers are driven from 10 mOhm rather than 7.5 Ohms?
====
I don't see why not.
====
>- If not, it is ok not to insert a separator. But how do I determine the Q for the Linkwitz circuit?
====
Good question, I would be prone to take the average and skip the divider if drivers share the load fairly evenly, but might want to ask 'the man' himself.
GM
- Status
- Not open for further replies.