Let me ask a fundamental question around speaker distortion-
If you have high distortion and are not the driver designer what opportunities/strategies are there for reducing distortion? I know multiple drivers can reduce distortion, but a DIY'er can't modify the voice coil or magnet. Playing games with the cone has substantial risks and not usually reversable. What else?
WRT mike distortion B&K presented a paper on reducing the distortion at the interface. It was a capacitive feed back to compensate for something. I cannot find the paper with a quick search. However I have measured the distortion of a number of mike "preamps". While they are all low they are not that low.Typically around .005% to .01% at 50mV in. Some are worse. The HP preamp had the widest dynamic range but it uses a 200V supply for the circuit and very low distortion.
If you have high distortion and are not the driver designer what opportunities/strategies are there for reducing distortion? I know multiple drivers can reduce distortion, but a DIY'er can't modify the voice coil or magnet. Playing games with the cone has substantial risks and not usually reversable. What else?
WRT mike distortion B&K presented a paper on reducing the distortion at the interface. It was a capacitive feed back to compensate for something. I cannot find the paper with a quick search. However I have measured the distortion of a number of mike "preamps". While they are all low they are not that low.Typically around .005% to .01% at 50mV in. Some are worse. The HP preamp had the widest dynamic range but it uses a 200V supply for the circuit and very low distortion.
Great question,
As far as I know, not much!
It has been shown that incorporating current drive for mid-woofers can lower distortion. How much they can improve SOTA drivers with -65dB (~0.05%) distortion @94dB, I do not have that data. I have been advised that current drive that it doesn't seem to have an effect on HF units.
In addition to using multiple driver units, another option is to use woofers in push-pull, as the Sd(x) variation in each driver cancels the other out in the far field. This has been demonstrated in LF units in regards to lower 2nd order distortion. I'm not sure if it lower others eg. H5 or more.
For midwoofers, replacing an existing surround Purifi's neutral surround may help. This will need some kind of group buy with a MOQ from Purifi, whom seem very supportive of DIYers/startups. Dali use one for their in-wall (IW Sub 100) subwoofer.
the "Constant Surface Surround (CSS) technology, developed from the ground up by DALI's engineering team in Denmark." looks conspicuously like Purifi's neutral surround, so I would like to think they have a license to use it.
Some have argued that there's little point to doing these measurements, since they are already readily available. But this what started my investigations in the first place, when I noticed discrepancies between Klippel's data and Voice Coil and Hificompass.com
eg. EAC Klippel's measurement of an 8" midwoofer @HiFiCompass, measuring driver at 96dB@1m:
@HiFiCompass' hand tuned method with B&K 4133+2639+2804
For people not familiar with these charts, there is at least a 10dB difference @ 1KHz.
So if the person with Klippel distortion measurement system isn't within 1-2dB error of another measurement setup, who is (more) right?
And what a person with a basic microphone to do?
Is there an alternative?
In my preliminary testing ,a low sensitivity, high SPL capable microphone with low noise pre-amp like the Line Audio Omni1, AT 2020, Rode NT1-A or upcoming Omni-mic 40K microphone, or the dynamic microphone eg. Sennheiser MD21U or MD42 in a carefully considered setup/process closer to Yevgeniy's hand tuned method than a Klippel machine.
The other alternative to not measure them at all, and use a proxy that is easier to measure, like maximum SPL, with a given THD limit, like 3%.
If your speaker can hit 115dB@1m at 3% THD and mine can hit 100dB@1m... it's arguable that at 85-105dB it's cleaner than mine. But Earl Geddes has shown that THD is not consistent with user's listening preferences...perhaps I should start using his Gedlee metric...
And of course, all of this microphone madness doesn't even take into the account our individual listening environments, or the psychoacoustics of hearing eg. nonlinearity of SPL, masking of frequencies (and thus harmonics).
If anyone has read the patent on MPEG 1 Layer 3, please send me a link. It would be interesting to understand the decisions on how the algorithm determines how/when to discard bits, such that a 16bit/44KHz 1,411 kilobits per second file can be shrunk to 320 Kbp/s file, yet maintain studio-level transparency for 95% of the population.
It is not lost of me that I haven't been able to answer "is this even important, in the grand scheme of things" a great question by people like @b_force .
It may well be that absolute distortion is unimportant in loudspeaker design. Perhaps just relatively distortion (ie. when the driver start to rise in distortion is a one consideration for the XO point).
One must first seek to understand...
As far as I know, not much!
It has been shown that incorporating current drive for mid-woofers can lower distortion. How much they can improve SOTA drivers with -65dB (~0.05%) distortion @94dB, I do not have that data. I have been advised that current drive that it doesn't seem to have an effect on HF units.
In addition to using multiple driver units, another option is to use woofers in push-pull, as the Sd(x) variation in each driver cancels the other out in the far field. This has been demonstrated in LF units in regards to lower 2nd order distortion. I'm not sure if it lower others eg. H5 or more.
For midwoofers, replacing an existing surround Purifi's neutral surround may help. This will need some kind of group buy with a MOQ from Purifi, whom seem very supportive of DIYers/startups. Dali use one for their in-wall (IW Sub 100) subwoofer.
the "Constant Surface Surround (CSS) technology, developed from the ground up by DALI's engineering team in Denmark." looks conspicuously like Purifi's neutral surround, so I would like to think they have a license to use it.
Some have argued that there's little point to doing these measurements, since they are already readily available. But this what started my investigations in the first place, when I noticed discrepancies between Klippel's data and Voice Coil and Hificompass.com
eg. EAC Klippel's measurement of an 8" midwoofer @HiFiCompass, measuring driver at 96dB@1m:
@HiFiCompass' hand tuned method with B&K 4133+2639+2804
For people not familiar with these charts, there is at least a 10dB difference @ 1KHz.
So if the person with Klippel distortion measurement system isn't within 1-2dB error of another measurement setup, who is (more) right?
And what a person with a basic microphone to do?
Is there an alternative?
In my preliminary testing ,a low sensitivity, high SPL capable microphone with low noise pre-amp like the Line Audio Omni1, AT 2020, Rode NT1-A or upcoming Omni-mic 40K microphone, or the dynamic microphone eg. Sennheiser MD21U or MD42 in a carefully considered setup/process closer to Yevgeniy's hand tuned method than a Klippel machine.
The other alternative to not measure them at all, and use a proxy that is easier to measure, like maximum SPL, with a given THD limit, like 3%.
If your speaker can hit 115dB@1m at 3% THD and mine can hit 100dB@1m... it's arguable that at 85-105dB it's cleaner than mine. But Earl Geddes has shown that THD is not consistent with user's listening preferences...perhaps I should start using his Gedlee metric...
And of course, all of this microphone madness doesn't even take into the account our individual listening environments, or the psychoacoustics of hearing eg. nonlinearity of SPL, masking of frequencies (and thus harmonics).
If anyone has read the patent on MPEG 1 Layer 3, please send me a link. It would be interesting to understand the decisions on how the algorithm determines how/when to discard bits, such that a 16bit/44KHz 1,411 kilobits per second file can be shrunk to 320 Kbp/s file, yet maintain studio-level transparency for 95% of the population.
It is not lost of me that I haven't been able to answer "is this even important, in the grand scheme of things" a great question by people like @b_force .
It may well be that absolute distortion is unimportant in loudspeaker design. Perhaps just relatively distortion (ie. when the driver start to rise in distortion is a one consideration for the XO point).
One must first seek to understand...
Last edited:
When the topic is harmonic distortion in loudspeakers the problem is not the lower orders (2nd, 3rd) but the higher ones. As an example I give you Erin Hardison's measurements of the MOFI Sourcepoint 8. This is a great sounding loudspeaker (based on limited listening to them in person). But the amount of 2nd order distortion below 200Hz is actually quite high. If Andrew Jones thought this sounded bad he would not release this design:
https://www.erinsaudiocorner.com/loudspeakers/mofi_8/
NOTE: scroll down to you see plots labeled "Harmonic Distortion". For the 96dB measurement the distortion reaches almost -20dB (10%) at 80Hz...
Higher orders distortions are not really reduced by current drive or motional feedback. Instead the driver itself must have low high-order distortion. IMD is another problem that will not be solved using current drive IIRC.
Vinyl lovers are basking in circa 1% 2nd+3rd order distortion all the time, for example, and often say that ultra-low distortion digital sources are "sterile". OTOH most people agree that too much 5th and higher order distortion is bad. If you look at the waveform distortion from HD 4th and higher orders it definitely looks pretty ugly!
It has also been shown that using series impedance will reduce 3rd order distortion, just like with current drive. See:
https://purifi-audio.com/blog/app-notes-2/low-distortion-filter-for-ptt6-5x04-naa-11
https://www.erinsaudiocorner.com/loudspeakers/mofi_8/
NOTE: scroll down to you see plots labeled "Harmonic Distortion". For the 96dB measurement the distortion reaches almost -20dB (10%) at 80Hz...
Higher orders distortions are not really reduced by current drive or motional feedback. Instead the driver itself must have low high-order distortion. IMD is another problem that will not be solved using current drive IIRC.
Vinyl lovers are basking in circa 1% 2nd+3rd order distortion all the time, for example, and often say that ultra-low distortion digital sources are "sterile". OTOH most people agree that too much 5th and higher order distortion is bad. If you look at the waveform distortion from HD 4th and higher orders it definitely looks pretty ugly!
It has also been shown that using series impedance will reduce 3rd order distortion, just like with current drive. See:
https://purifi-audio.com/blog/app-notes-2/low-distortion-filter-for-ptt6-5x04-naa-11
Last edited:
Any idea about search words for that B&K paper? Capacitive feedback sounds a lot like bootstrapping. To the best of my knowledge, most of their preamps use a bootstrapped FET. Not the worst way to do it but not the best. You get pretty low distortion when you keep the output swing low enough compared to the operating voltage of 120 V. Another drawback of their amps it the lowish input impedance of a few 100 MOhms. This means you loose charge on the membrane as it swings close to the back plate (similar to a low impedance foil in an electrostatic loudspeaker).
This here is pretty next level and not overly complicated: https://www.by-rutgers.nl/ME6211-PRO37R.html
Why shouldn't current drive reduce some higher order distortions, i.e. those that result from Bxl nonlinearity or inductance modulation? Of course, it cannot do anything about those that arise from mechanical hysteresis of the soft moving parts.
MFB with insufficient loop gain, however, may just shift nonlinearities from low to higher orders.
The inductance modulation effect produces IIRC primarily 3rd order distortion.
I do not recall seeing that current drive was all that helpful in reducing the high order distortion products. If anyone knows about such data please post a link or info since I would like to learn about it.
I do not recall seeing that current drive was all that helpful in reducing the high order distortion products. If anyone knows about such data please post a link or info since I would like to learn about it.
hi Charlie,
The hysteresis distortion from iron shows up as a cascade of odd harmonics that only decay slowly in amplitude vs harmonic number. Current slashes this distortion.
Cheers and happy Xmax
Lars
The hysteresis distortion from iron shows up as a cascade of odd harmonics that only decay slowly in amplitude vs harmonic number. Current slashes this distortion.
Cheers and happy Xmax
Lars
Charlie, this thread has interesting data:
Unfortunately, the op (who later deleted his account) only measured 2nd and 3rd harmonic. When you do the math, you get a dependence on the square of Bxl with voltage drive, but only on Bxl with current drive. This should result in more benign higher order distortion.
Lars, I must admit I didn't read all 15 pages of that thread, so my apologies if it was explained there. Why would current drive completely do away with hysteresis distortion? I get the part about inductance modulation, but would hysteresis also modulate B and hence result in a distortion contribution that even current drive cannot get rid of.
Unfortunately, the op (who later deleted his account) only measured 2nd and 3rd harmonic. When you do the math, you get a dependence on the square of Bxl with voltage drive, but only on Bxl with current drive. This should result in more benign higher order distortion.
Lars, I must admit I didn't read all 15 pages of that thread, so my apologies if it was explained there. Why would current drive completely do away with hysteresis distortion? I get the part about inductance modulation, but would hysteresis also modulate B and hence result in a distortion contribution that even current drive cannot get rid of.
the magnetic hysteresis show up as sudden steps in the slope of the B-H traces. this amounts to the apparent inductance jumping up or down at discrete time instants. this adds odd order harmonics.
Thanks, Lars. This much was clear to me. But if B(H) changes, this will change the force factor of the transducer also, and current drive cannot completely eliminate this part.
There are a few plots showing harmonics above H3 in this thread (most only show H2 and H3 though). One good example is post #136, which shows H5 for voltage (red) and current (yellow) drive.
I think it's likely that the effect is smaller on average vs a mid-woofer, but the assertion that there's no effect seems to contradict measurements I've done on a couple compression drivers. I recently posted some measurements showing the effect of passive filter networks on the distortion of a two-way loudspeaker with a compression driver (Celestion CDX1-1747). Increasing the source impedance from near-zero to 10-15Ω dropped H3 and H5 by ~5dB over a good chunk of the tweeter's passband.
I also measured a Celestion CDX1-1446 in a somewhat less-well-controlled manner. The frequency response looks horrendous in part because I used a cheap cardioid dynamic mic without any correction. No EQ compensation was used in this case, but the responses aren't that far off above 2.5kHz:
The amp I used was an LM1875-based circuit with variable output impedance that I originally made for electric guitar use. Here's the measured distortion in "low-Z" mode (about 9Ω):
And in "high-Z" mode (about 70Ω at 2kHz, falling to 60Ω at 20kHz):
H3 drops almost 10dB above 2.5kHz.
Hi Lars, thank you for the nice explanation! Happy Xmax to you, too. 🙂
@bmc0 Also thanks for posting that data. Very interesting, as I never expected such a dramatic change for e.g. H5.
All interesting so far. One datapoint I will provide and remember correlation is not causation BUT if you wade through the measurements here : https://www.soundstagenetwork.com/index.php?option=com_content&view=article&id=16&Itemid=140 the speakers that are reasonable flat and neutral in response are not the ones with the best magazine reviews. However some really messy response speakers are rated highly in the magazines and they tend to have lower distortion curves. But the higher ratings also correlate with higher prices (more that better objective performance).
Creating a current drive for a speaker and getting reasonably flat response is not a trivial exercize. I will explore it on some upcomming project. However it may be beyond typical DIY speaker stuff.
I am still looking for the paper on microphone distortion reduction from B&K. In the mean time this was an interesting paper I ran across in my collection:
I found the article: Reduction of Non-linear Distortion inCondenser Microphones by UsingNegative Load Capacitance
https://www.bksv.com/media/doc/bv0048.pdf Pg. 19. Its pretty heavy going.
I think the distortion number are really quite low to start with but still an interesting concept. To my knowledge its not used in any of the B&K preamps commercially available.
https://www.bksv.com/media/doc/bv0048.pdf Pg. 19. Its pretty heavy going.
I think the distortion number are really quite low to start with but still an interesting concept. To my knowledge its not used in any of the B&K preamps commercially available.
I just rechecked my collection of mike preamp types. Its a lot of not much to process. The noise floor is the limiting element. They were all tested at 50 mV. 50 mV is 94 dB SPL with a 1" B&K mike. Rarely was there identifiable distortion at 50 mV. However the GR handled 500 mV well. The HP's 5V max is pretty unique and a fallout of a 200V internal supply. If I get energetic I'll check upper limits on the B&K with the two supply voltage standards. The 2609 supplies 130V to the preamp's electronics. The 2804 is only 28V. All the options get to 100 KHz except the 2804 measuring amp which has a 20 KHz low pass filter and was set for 2V out for 50 mV in. It doesn't have a direct mike output.
If I were starting now I would get the GR preamp. It works well, runs on a reasonable voltage (battery operation is easy) and they are not expensive on eBay now. This note may ruin that. . .
If I were starting now I would get the GR preamp. It works well, runs on a reasonable voltage (battery operation is easy) and they are not expensive on eBay now. This note may ruin that. . .
This one, 1560-P42. 2 9V batteries and you are ready to go. Works with 1/2" B&K capsule or 1" or 1/4" with adapters and the bias is switchable so you can use it with prepolarized mikes. They do pop up on eBay. the 1560-p40 does NOT have a mike bias capability but is mostly the same.
I picked up 10 of these several years back when someone was selling off a bunch of them - you don't want to know how cheap I bought them for!
But 4 of them needed some repair due to blown bipolar transistors or open electrolytic capacitors or flaky switch contacts, to work at all reasonably. They all came without connectorized cables and use an unobtainium connector so they needed to have a cable hardwired onto them. I made 3D printed plastic pieces to hold it all together afterwards.
Thee Preamps were made with several wet slug tantalum capacitors from a batch made then which had a short life and dried up. Those should be replaced with solid tantalum or ceramic, which also need to be chosen to be small enough to be kludged onto the boards (microscope required).
So not low effort really, but if anyone needs details, I wrote up mod instructions a while back if needed. Will post in another several replies.
But 4 of them needed some repair due to blown bipolar transistors or open electrolytic capacitors or flaky switch contacts, to work at all reasonably. They all came without connectorized cables and use an unobtainium connector so they needed to have a cable hardwired onto them. I made 3D printed plastic pieces to hold it all together afterwards.
Thee Preamps were made with several wet slug tantalum capacitors from a batch made then which had a short life and dried up. Those should be replaced with solid tantalum or ceramic, which also need to be chosen to be small enough to be kludged onto the boards (microscope required).
So not low effort really, but if anyone needs details, I wrote up mod instructions a while back if needed. Will post in another several replies.
Last edited:
Here;s a manual of the earlier version of the 1560-P42, see page 72 in particular.