Just wanted to follow up on my earlier post. I had some more conversation with Lars Risbo and I have a better understanding of what is going on with his tech note.
According to that tech note and another published at Purifi, and what Lars told me, the voice coil inductance gets modulated during cone movement and this mechanism gives rise to a "distortion voltage" that is generated within the motor. This is happening all across the passband and gives rise to the distortion we measure. This tech note was written about the PTT8.0X which has a metal cone and some tall breakup peaks. You can think of the breakup peak as a narrow frequency region in which the efficiency of the driver at converting current into sound is 10+dB higher than the nominal passband level, and we see this in the FR of the driver. Let's say the peak (for easy math) is located at 6kHz. If we excite the driver with current at 2kHz, the 3rd order distortion product is occurring at 6kHz, which is where we have the breakup peak. The distortion voltage is causing current to flow at 6kHz in addition to the drive current we are supplying and thus there is a peak of 3rd order distortion at 2kHz with a shape that follows the frequency response shape of the 6kHz breakup peak. So, when you see these "echoes" of breakup peaks in the distortion signature at lower frequency, this is what is happening. Supposedly this is mainly 3rd order and higher odd order products, but I am not sure why that is at the moment (e.g. why not even orders as well).
Now, what about the network that Lars is using to reduce these distortion echoes? How does that work? Well remember I said that there is a "distortion voltage" generated in the motor. This voltage will induce a current to flow along whatever path is available to it. Current must flow in a loop, thru the driver, out one end of the voice coil, thru some path via the crossover network or the amplifier, and then back into the other end of the voice coil. The "trick" to reducing the distortion is to use a network element in series with and directly connected to the driver that has a high impedance at the breakup peak, in my example that is at 6kHz. The higher this impedance the lower the current flow at 6kHz (by Ohm's Law). Since the driver is converting current into sound pressure, less current means less sound produced. So the 3rd order product from our 2kHz distortion voltage source meets a high impedance path and relatively little current can flow, and therefore the 6kHz output from it is suppressed. Lars chose a network in series with the driver (the parallel RLC notch) for exactly this reason - it puts a high impedance in series with the driver at 6kHz. He compared that to notch filters that were placed in parallel to the driver (series RLC notch networks) and showed that in contrast these actually provided a low impedance path around the driver at 6kHz, and this allowed lots of current to flow and did not have any effect on the 2kHz 3rd order distortion product.
In general, resistance in series with the driver will suppress distortion products at the frequency where the resistance is present. So for example a passive, series low pass network can be beneficial in this regard and will suppress the distortion spectrum at 1/3 the frequency broadly. You can see this in the first figure in the tech article in how the distortion level starts to decrease as low as 1kHz compared to the second figure (the raw driver).
This is all new to me, and I think it is pretty cool. Makes me think about the possibilities of hybrid DSP and passive.
Okay, I understand this better now too. I agree with the 3rd order reduction having to go through the driver in this case, as that is the distortion being targeted.
However, other HD can still be reduced by using a series notch if the issues never make it to the driver.
In general, resistance in series with the driver will suppress distortion products at the frequency where the resistance is present. So for example a passive, series low pass network can be beneficial in this regard and will suppress the distortion spectrum at 1/3 the frequency broadly. You can see this in the first figure in the tech article in how the distortion level starts to decrease as low as 1kHz compared to the second figure (the raw driver).
This is all new to me, and I think it is pretty cool. Makes me think about the possibilities of hybrid DSP and passive.
Charlie - you did a spectacular job of explaining this phenomenon with your post - congrats and thanks for your effort!!
I assume this link is the one you reference with your latest post?
Wolf - Thanks for pushing the subject concerning Fs/Breakup compensation, Sir!
The discussion brought a subjective perspective in to alignment with an objective-one.
I would never have understood what I didn't understand unless you had done so.
There is a good reason why you received a [well deserved] Life Achievement Award from your peers and followers in the DIY Speaker Building/Design community, dude.
@Steve_Lee said:
...I think the real answer lies in a hybrid approach to sound shaping of speaker drivers using both passive and active methodologies... a MT passively XO'd and EQ'd by DSP while the W~MT XO handled by DSP may just be a superior approach.
Interesting presentations / conclusions from Meadowlark, whom in the past made some very nice passive modified-transmission line speakers, and the approach they have taken with their new line active speakers.
While promotional (what company’s white papers aren’t) I find it interesting that while they could have expanded with their transmission line ‘brand’ in an active / powered speaker, they opted for a DSP sealed woofer design (over transmission, ported and/or PRs) for signal extension, resolution, and integrity.
IMO once you get below 500Hz it makes a lot of sense to use an active crossover. I also like to have the ability to adjust the shape and extension of a closed box subwoofer using an LT. The M-T crossover can be done passively and there are fewer reasons to use DSP up there.
@Steve_Lee said:
...I think the real answer lies in a hybrid approach to sound shaping of speaker drivers using both passive and active methodologies... a MT passively XO'd and EQ'd by DSP while the W~MT XO handled by DSP may just be a superior approach.
Interesting presentations / conclusions from Meadowlark, whom in the past made some very nice passive modified-transmission line speakers, and the approach they have taken with their new line active speakers.
While promotional (what company’s white papers aren’t) I find it interesting that while they could have expanded with their transmission line ‘brand’ in an active / powered speaker, they opted for a DSP sealed woofer design (over transmission, ported and/or PRs) for signal extension, resolution, and integrity.
Food for thought with hybrid active W / passive MT… it's all about the bass (quality)
The Meadowlark technical presentation looks great, but one problem I have with it. All the graphs are simulated. If their tech is so great, why not use real world measurements? Another problem is that I really like the sound of good transmission line alignments. I haven't been able to prove any difference from ported, but TLs just sound better to me - more controlled bass for one.
But Chahly - Stahkist don't want speakers that look good, Stahkist wants speakers that sound good!
rjj45
I too like/prefer TL's base (and eq'd open backs) over ported/PRs. But they don't pass the WAF with the highest marks.
(and sealed won't 'pass gas').
I'd expect their presentations are drawn from their RWMs, and done so for presentation simplification.
I couldn't get the bass to sound clean across all music choices until I stuffed the ports in the sub to seal them and used high sensitivity 10" midwoofers XO'd to the sub at 95 Hz.
The DSP just makes dialing in the XO and EQ a breeze and then touch it up with minor adjustments to the gain of the sub, MW & T.
I also find that a 2nd order (12db/oct) roll-off curve to sound the best btween the Sub and MW:
Comments
Just wanted to follow up on my earlier post. I had some more conversation with Lars Risbo and I have a better understanding of what is going on with his tech note.
According to that tech note and another published at Purifi, and what Lars told me, the voice coil inductance gets modulated during cone movement and this mechanism gives rise to a "distortion voltage" that is generated within the motor. This is happening all across the passband and gives rise to the distortion we measure. This tech note was written about the PTT8.0X which has a metal cone and some tall breakup peaks. You can think of the breakup peak as a narrow frequency region in which the efficiency of the driver at converting current into sound is 10+dB higher than the nominal passband level, and we see this in the FR of the driver. Let's say the peak (for easy math) is located at 6kHz. If we excite the driver with current at 2kHz, the 3rd order distortion product is occurring at 6kHz, which is where we have the breakup peak. The distortion voltage is causing current to flow at 6kHz in addition to the drive current we are supplying and thus there is a peak of 3rd order distortion at 2kHz with a shape that follows the frequency response shape of the 6kHz breakup peak. So, when you see these "echoes" of breakup peaks in the distortion signature at lower frequency, this is what is happening. Supposedly this is mainly 3rd order and higher odd order products, but I am not sure why that is at the moment (e.g. why not even orders as well).
Now, what about the network that Lars is using to reduce these distortion echoes? How does that work? Well remember I said that there is a "distortion voltage" generated in the motor. This voltage will induce a current to flow along whatever path is available to it. Current must flow in a loop, thru the driver, out one end of the voice coil, thru some path via the crossover network or the amplifier, and then back into the other end of the voice coil. The "trick" to reducing the distortion is to use a network element in series with and directly connected to the driver that has a high impedance at the breakup peak, in my example that is at 6kHz. The higher this impedance the lower the current flow at 6kHz (by Ohm's Law). Since the driver is converting current into sound pressure, less current means less sound produced. So the 3rd order product from our 2kHz distortion voltage source meets a high impedance path and relatively little current can flow, and therefore the 6kHz output from it is suppressed. Lars chose a network in series with the driver (the parallel RLC notch) for exactly this reason - it puts a high impedance in series with the driver at 6kHz. He compared that to notch filters that were placed in parallel to the driver (series RLC notch networks) and showed that in contrast these actually provided a low impedance path around the driver at 6kHz, and this allowed lots of current to flow and did not have any effect on the 2kHz 3rd order distortion product.
In general, resistance in series with the driver will suppress distortion products at the frequency where the resistance is present. So for example a passive, series low pass network can be beneficial in this regard and will suppress the distortion spectrum at 1/3 the frequency broadly. You can see this in the first figure in the tech article in how the distortion level starts to decrease as low as 1kHz compared to the second figure (the raw driver).
This is all new to me, and I think it is pretty cool. Makes me think about the possibilities of hybrid DSP and passive.
Okay, I understand this better now too. I agree with the 3rd order reduction having to go through the driver in this case, as that is the distortion being targeted.
However, other HD can still be reduced by using a series notch if the issues never make it to the driver.
InDIYana Event Website
Wolfy's Photobucket
Charlie - you did a spectacular job of explaining this phenomenon with your post - congrats and thanks for your effort!!
I assume this link is the one you reference with your latest post?
https://purifi-audio.com/wp-content/uploads/2022/03/220211_R05-Notchfilter.pdf
Yes, that is the Purifi Tech Note related to the distortion issue containing the figures I mentioned. Thanks.
pi keeps popping in my head since it is close to 3 but I doubt that has anything to do with it.
mmm pie
Wolf - Thanks for pushing the subject concerning Fs/Breakup compensation, Sir!
The discussion brought a subjective perspective in to alignment with an objective-one.
I would never have understood what I didn't understand unless you had done so.
I'm glad it was beneficial to you!
InDIYana Event Website
Wolfy's Photobucket
There is a good reason why you received a [well deserved] Life Achievement Award from your peers and followers in the DIY Speaker Building/Design community, dude.
Interesting presentations / conclusions from Meadowlark, whom in the past made some very nice passive modified-transmission line speakers, and the approach they have taken with their new line active speakers.
While promotional (what company’s white papers aren’t) I find it interesting that while they could have expanded with their transmission line ‘brand’ in an active / powered speaker, they opted for a DSP sealed woofer design (over transmission, ported and/or PRs) for signal extension, resolution, and integrity.
http://www.meadowlarksings.com/technical discussion subpage 3.htm
Food for thought with hybrid active W / passive MT… it's all about the bass (quality)
IMO once you get below 500Hz it makes a lot of sense to use an active crossover. I also like to have the ability to adjust the shape and extension of a closed box subwoofer using an LT. The M-T crossover can be done passively and there are fewer reasons to use DSP up there.
The Meadowlark technical presentation looks great, but one problem I have with it. All the graphs are simulated. If their tech is so great, why not use real world measurements? Another problem is that I really like the sound of good transmission line alignments. I haven't been able to prove any difference from ported, but TLs just sound better to me - more controlled bass for one.
rjj45
I too like/prefer TL's base (and eq'd open backs) over ported/PRs. But they don't pass the WAF with the highest marks.
(and sealed won't 'pass gas').
I'd expect their presentations are drawn from their RWMs, and done so for presentation simplification.
I couldn't get the bass to sound clean across all music choices until I stuffed the ports in the sub to seal them and used high sensitivity 10" midwoofers XO'd to the sub at 95 Hz.
The DSP just makes dialing in the XO and EQ a breeze and then touch it up with minor adjustments to the gain of the sub, MW & T.
I also find that a 2nd order (12db/oct) roll-off curve to sound the best btween the Sub and MW:
Those 10" MW's (Eminence CX10) are published as having an Fs of 49Hz where DATS found them to be 95 Hz.
Using the USB mic and ear adjustments I wound up with the XO above.
You CANNOT design around published driver specs - you gotta measure them.
More evidence supporting Hybrid (passive + Active/DSP) Xovers with a Series twist:
https://www.diyaudio.com/community/threads/advantages-of-series-crossover-vs-parallel-for-subs-open-baffle.404680/
Very interesting . . .