Deluxe series ©Vandermill 2022
Deluxe series ©Vandermill 2022
After the simple but effective correction that TNT suggested, I do think it was worth it to create this Deluxe series ©Vandermill 2022.
They do seem to tell a compelling story that would otherwise take a lot of words to express. So I figured: what the heck,
I'll take the time and create these, I can upload them to my website to try and create a better understanding of these arrays.
I think these three graphs do tell a convincing story of what an array like this is all about. The headroom for EQ, the way it interacts
with the room... it's all there... these texts sure make it easier to understand, good call @TNT!
So, after all these new graphs, I just might have to introduce yet another exclusive series. First installment:
I think it's pretty self explanatory, don't you? I've created a set of 360 degree measurements for the Scan Speak 10F 8414G10 driver.
It has a few small differences, like the impedance curve, so it's going to react a bit different too. I made this one in the diffraction tool
of Vituixcad, just because I had not done that before. I've used driver properties that nc535 made and ABEC simulations from fluid.
I figured it was time to learn how to trace an SPL graph and make a set on my own. I did use my baffle dimensions with round-overs
so it's bound to be a bit different from the TC9 model. It's close but not quite the same as the ABEC model. Close enough for this.
It's clear that this driver does look quite a bit different in Frequency response. Maybe I should compare the TC9 and 10F graphs as well
.
I think it's pretty self explanatory, don't you? I've created a set of 360 degree measurements for the Scan Speak 10F 8414G10 driver.
It has a few small differences, like the impedance curve, so it's going to react a bit different too. I made this one in the diffraction tool
of Vituixcad, just because I had not done that before. I've used driver properties that nc535 made and ABEC simulations from fluid.
I figured it was time to learn how to trace an SPL graph and make a set on my own. I did use my baffle dimensions with round-overs
so it's bound to be a bit different from the TC9 model. It's close but not quite the same as the ABEC model. Close enough for this.
It's clear that this driver does look quite a bit different in Frequency response. Maybe I should compare the TC9 and 10F graphs as well
.
I was guessing that I can add basic information as a foot note, is that acceptable? The footnote in this case should read something like this:
Simulation of the speaker at 2.7m distance. No EQ was applied, microphone set at 1m height. Single driver elevated to 1m.
Indeed it is, as I'm not about to open up my speakers yet again to remove all of the filters, just to get an unshaded graph.
And I don't have an option for anechoic measurements anyway, living in the center of a town.
Even in that case it's interesting
//
I hope so. For me, it was an exercise if I could make the diffraction model within Vituixcad. I had not done that before, using a factory provided
SPL sheet.
So I traced the SPL graph in Vituixcad and used the diffraction tool to get a set of 360 degree measurements. Pure curiosity to find out what it takes.
I don't need these, but it was fun to be able to walk trough the entire proces like this. I must say, Vituixcad is very versatile and complete.
I'm glad we have these generous people like Kimmo providing us with tools like this. It's awesome, really.
Some more simulations, basically to show why I moved forward by adding the filters to the straight array.
First up, vertical movement at listening distance:
Vertical set of predicted in-room curves spanning from -100 mm up to + 100 mm around the sweetspot. No EQ applied.
After creating filters that had the most even vertical response I was happy to see that the predicted improvement went
all the way down to 200 Hz. The frequency variation is notably improved due to the more bundled vertical output of the
array. The combing is highly reduced, almost eliminated completely. The variation left even on the top end is no more
than plus/minus 2 dB. Below 5K it is reduced to plus/minus 1 dB.
Results when moving closer or further away:
A set of in-room predictions spanning a distance of 2.5 to 3.5 m from the array. No EQ. What can be seen here that moving closer
or further from the array has less variation for the shaded array. When I saw this result (after trying countless versions) I figured
I had nothing to loose by going forward with these plans to add the filters. The filtered array was acting more well behaved under all
use-cases. The cost of it all is about 2 dB less sensitivity that gets burned off in the filters itself.
Under angle:
In-room prediction of the angles 0 to 15 degree at a listening distance of 2.7 m. No EQ applied.
Here we see similar behavior between both arrays. As there's no vertical change and the horizontal span is all
that has changed. The filtered array still shows a benefit of a more linear drop/octave with less swings.
I hope that illustrates the differences in behavior that can be seen between a filtered and unfiltered straight array of 25 drivers.
The array is optimized to work at a single distance and listening height. But care has been put in that all other distances and
heights of interest function at least as well or better. Modeled with the Scan Speak 10F 8414G10 drivers I started using after
the last array surgery.
I'm hoping my next graphs will be as measured instead of more sims, but it depends on my luck to get the room to myself for
a few hours. I actually need way more than that, but a start would be to measure the arrays. The rest will hopefully follow soon
after that. The moments of having the house to myself have dropped significantly. Working from home has become the new
standard for my girlfriend. How I wish I had way more time to spend on this all.
First up, vertical movement at listening distance:
Vertical set of predicted in-room curves spanning from -100 mm up to + 100 mm around the sweetspot. No EQ applied.
After creating filters that had the most even vertical response I was happy to see that the predicted improvement went
all the way down to 200 Hz. The frequency variation is notably improved due to the more bundled vertical output of the
array. The combing is highly reduced, almost eliminated completely. The variation left even on the top end is no more
than plus/minus 2 dB. Below 5K it is reduced to plus/minus 1 dB.
Results when moving closer or further away:
A set of in-room predictions spanning a distance of 2.5 to 3.5 m from the array. No EQ. What can be seen here that moving closer
or further from the array has less variation for the shaded array. When I saw this result (after trying countless versions
) I figuredI had nothing to loose by going forward with these plans to add the filters. The filtered array was acting more well behaved under all
use-cases. The cost of it all is about 2 dB less sensitivity that gets burned off in the filters itself.
Under angle:
In-room prediction of the angles 0 to 15 degree at a listening distance of 2.7 m. No EQ applied.
Here we see similar behavior between both arrays. As there's no vertical change and the horizontal span is all
that has changed. The filtered array still shows a benefit of a more linear drop/octave with less swings.
I hope that illustrates the differences in behavior that can be seen between a filtered and unfiltered straight array of 25 drivers.
The array is optimized to work at a single distance and listening height. But care has been put in that all other distances and
heights of interest function at least as well or better. Modeled with the Scan Speak 10F 8414G10 drivers I started using after
the last array surgery.
I'm hoping my next graphs will be as measured instead of more sims, but it depends on my luck to get the room to myself for
a few hours. I actually need way more than that, but a start would be to measure the arrays. The rest will hopefully follow soon
after that. The moments of having the house to myself have dropped significantly. Working from home has become the new
standard for my girlfriend. How I wish I had way more time to spend on this all.
Last edited:
I've been on a quest to find out if adding the coils to the arrays had a pleasant by-product. For a long time I had been
following Joe Rasmussen's Elsinore thread, and his vision on lowering what he called current distortion. The theory is
understandable enough, if you look at a few of his posts. The distortion originates from the fact that drivers vary in
impedance with the movement of the cone. For instance, here's the impedance change per cone position for the Scan
Speak 10F 8414G10:
Joe's suggestion is adding an inductor in series with the "woofer" as this makes the swing in impedance seem smaller
as seen by the amplifier. When playing two notes, one lower note that makes the cone swing and the one higher up
will be modulated by the reaction from the amp, as the inductance will vary due to playing that lower note. Much
saver to just read what Joe says here...
I've added a lot of coils to my series/parallel schematics, so I figured to check if it would lower my current distortion as
a pleasant by-product. If I read the above graph, and use the 2.3 mm x-max of this driver, the swing or variation of
the impedance in my array (will be similar to that of a single driver) is just over 30%.
As the added inductance is in the parallel legs of my circuitry, I had to figure out how that changes the numbers. If I
did it right, there will be about a 7% improvement in the induction swing. Not spectacular, but an improvement non the
less. Should I add a single 0.5 mH coil with the (total) array, that improvement would grow to 20+%, but it would shave
off 7 to 8 dB on the top end SPL (gradually). See:
Dark blue and light gray are without 0.5 mH inductor, black and gray are with the series inductor.
I do doubt if dropping the top end by another 8 dB and making up for that with EQ would be a way forward to
lower total IMD distortion. Limiting the cone movement needed on the bottom end would be a quicker way to
do that. But doing that introduces yet another compromise, one that got me in an argument a few times. Not
going there again. I already lowered the low end movement with the introduction of the subs. Once I find the
time to dial in everything once more, I'll try and maximize what's possible without creating new compromises.
Anyway, I did get my answer to this question: yes, the induction of the coils does bring an ever so slight
advantage. An IMD test could tell us, but I did not run any prior test to compare with... Anyone with an array,(*)
try adding a series inductor, value of about 0.5 uH and enough wire thickness, and run an IMD test to find out!
Run it with and without the inductor in series, EQ it back into shape. One should see up to 20+% reduction of
distortion if the added EQ doesn't steal most of it away again. I don't have spare coils and not about to get
them for this test just yet. Maybe some day I'll be curious enough.
(*) sadly, I've never seen an 'Le' graph of a TC9. The graph for the Scan Speak 10F comes from audioXpress.
following Joe Rasmussen's Elsinore thread, and his vision on lowering what he called current distortion. The theory is
understandable enough, if you look at a few of his posts. The distortion originates from the fact that drivers vary in
impedance with the movement of the cone. For instance, here's the impedance change per cone position for the Scan
Speak 10F 8414G10:
Joe's suggestion is adding an inductor in series with the "woofer" as this makes the swing in impedance seem smaller
as seen by the amplifier. When playing two notes, one lower note that makes the cone swing and the one higher up
will be modulated by the reaction from the amp, as the inductance will vary due to playing that lower note. Much
saver to just read what Joe says here...
I've added a lot of coils to my series/parallel schematics, so I figured to check if it would lower my current distortion as
a pleasant by-product. If I read the above graph, and use the 2.3 mm x-max of this driver, the swing or variation of
the impedance in my array (will be similar to that of a single driver) is just over 30%.
As the added inductance is in the parallel legs of my circuitry, I had to figure out how that changes the numbers. If I
did it right, there will be about a 7% improvement in the induction swing. Not spectacular, but an improvement non the
less. Should I add a single 0.5 mH coil with the (total) array, that improvement would grow to 20+%, but it would shave
off 7 to 8 dB on the top end SPL (gradually). See:
Dark blue and light gray are without 0.5 mH inductor, black and gray are with the series inductor.
I do doubt if dropping the top end by another 8 dB and making up for that with EQ would be a way forward to
lower total IMD distortion. Limiting the cone movement needed on the bottom end would be a quicker way to
do that. But doing that introduces yet another compromise, one that got me in an argument a few times. Not
going there again. I already lowered the low end movement with the introduction of the subs. Once I find the
time to dial in everything once more, I'll try and maximize what's possible without creating new compromises.
Anyway, I did get my answer to this question: yes, the induction of the coils does bring an ever so slight
advantage. An IMD test could tell us, but I did not run any prior test to compare with... Anyone with an array,(*)
try adding a series inductor, value of about 0.5 uH and enough wire thickness, and run an IMD test to find out!
Run it with and without the inductor in series, EQ it back into shape. One should see up to 20+% reduction of
distortion if the added EQ doesn't steal most of it away again. I don't have spare coils and not about to get
them for this test just yet. Maybe some day I'll be curious enough.
(*) sadly, I've never seen an 'Le' graph of a TC9. The graph for the Scan Speak 10F comes from audioXpress.
Last edited:
Hi all,
mvs0 what mechanism in the amp makes the IMD in this case, or reduces it? Impedance, load, seen by amp can be made flat if one wanted to, and still have coil in series with the driver.
Coil in series, or in otherwords increased impedance seen by driver, reduces distortion demonstrated in Purifi paper, first on the list here
https://purifi-audio.com/tech/
It is good thing if also amp distortion would reduce by stabilizing variation in load impedance. I'm curious how it happens as I'm not familiar with amplifier technology. All I have been able to figure out for now is that low amplifier output impedance lets the driver generated current flow and manifest itself as distortion, and that the low output impedance is result of feedback in the amp. Not sure how current in the amp/driver circuit would change the operation of the amp so that distortion results in the amp? any tips and pointers to material to read would speed the process, thanks
Found some pointers for example this https://www.stereophile.com/reference/60/index.html
mvs0 what mechanism in the amp makes the IMD in this case, or reduces it? Impedance, load, seen by amp can be made flat if one wanted to, and still have coil in series with the driver.
Coil in series, or in otherwords increased impedance seen by driver, reduces distortion demonstrated in Purifi paper, first on the list here
https://purifi-audio.com/tech/
It is good thing if also amp distortion would reduce by stabilizing variation in load impedance. I'm curious how it happens as I'm not familiar with amplifier technology. All I have been able to figure out for now is that low amplifier output impedance lets the driver generated current flow and manifest itself as distortion, and that the low output impedance is result of feedback in the amp. Not sure how current in the amp/driver circuit would change the operation of the amp so that distortion results in the amp? any tips and pointers to material to read would speed the process, thanks
Found some pointers for example this https://www.stereophile.com/reference/60/index.html
Last edited:
Wouldn't that depend on the severity of the distortion created? The amp wouldn't create this kind of distortion if all drivers
were build like the Purify drivers we've seen so far. Think of te graph I showed of Le fluctuating with cone position, now
imagine playing the two tones of an IMD test. The lower tone makes the cone move trough almost it's entire x-max distance
of +/- 2.3 mm, now imagine what the higher tone around 1 KHz gets. If the cone, playing a 50 Hz tone moves to - 1.8 mm
The 1 KHz tone still is moving the cone at a faster rate to make that tone, the amp sees a higher Le. a few moments later
the cone will be at the other end, still playing that 50 Hz tone, + 1.8 mm. What does the amp see now while still playing
that same 1 KHz note. Who's fault is it, the amp or the driver...
This should be an easy test even with a single driver, or especially with a single driver. An array of drivers will limit the
movement needed to play the notes at similar SPL levels.
Sure, amps will act differently when confronted with this kind of varying load. But we've stated already within this thread
that amplifiers do sound different.
Long ago I tried the array with a series resistor. I didn't have the power back then to play at similar levels, nor the cooling
needed for the resistor. So that was a failed and/or flawed test. This one is simpler, but I kind of lack the time to do it.
I'd have to get some coils too, as I don't have that kind of stuff in my drawers. I never build passive crossovers and when
I do, I only order the stuff I need. This graph shows the 10F is a likely candidate for a test like this, as it has a nice and low
Le. So even a relatively small coil in series will have an impact and change the balance as seen by the amplifier.
Obviously, the effect of that same coil on a single driver is going to be a bit different:
Dropping the top end by a large margin...
Anything that drops the SPL level is kind of moving backwards. But I do like this subject and lots of other projects could potentially benefit
from playing around with this theory. For instance an active 2 or 3 way. What would one have to loose by experimenting? If the driver Le
is low, you'd have a likely candidate to experiment. In an active setup, the coil used can be made part of the crossover. Wouldn't it be fun
to find out that a hybrid setup like this produces less distortion than a full active setup?
Last edited:
I realize I didn't get a lot of support when I build my conjugation network. "For a capable amplifier it wouldn't make any difference"
Sure... but I was interested in why one would make such a compensation network, meanwhile knowing way to little about what it
does. And I still fall short on that end. I do know that amplifiers sound different though, and there has to be a reason.
If there are simple tricks that can help us create better results... I'm all for it... Not all of it will work, but writing it off before trying
(and even measuring) it doesn't feel quite right to me.
But sacrificing SPL to gain a bit... that doesn't work either.
Sure... but I was interested in why one would make such a compensation network, meanwhile knowing way to little about what it
does. And I still fall short on that end. I do know that amplifiers sound different though, and there has to be a reason.
If there are simple tricks that can help us create better results... I'm all for it... Not all of it will work, but writing it off before trying
(and even measuring) it doesn't feel quite right to me.
But sacrificing SPL to gain a bit... that doesn't work either.
Last edited:
I tried to briefly look for evidence that amplifiers would make more distortion with varying load and it doesn't seem to be raised in papers / discussion much at all, at least its hard to find. For example extensive article by Douglas Self about amplifier distortions does mention effect of load only about few times saying that distortion is greater to lower impedance loads, but not much about varying impedance https://www.edn.com/distortion-in-power-amplifiers-part-i-the-sources-of-distortion/ other than "usually there is enough margin designed in for it not to be issue", which kind of says that some amplifier can drive loads perhaps from 2-8ohms within specs provided in the specsheet.
So for any sensible passive crossover/driver load combination that has relatively high minimum impedance that falls withing amplifier specs I don't think amplifiers suffers much increased distortion. After all the driver impedance varies but not that much that it would reduce amplifier load significantly from the nominal, unless silly things in passive network.
Although there is high possibility I'm wrong, because I don't have enough information about the subject, but to me it seems to be wrong conclusion currently that amplifier distortion would get affected by the varying load as long as its within reason. The perspective to look the issue is wrong, its the loudspeaker driver distortion that is affected by amplifier / passive network impedance and not the amplifier by loudspeaker / passive network impedance. They both are in the same circuit together and at least partially same current flows through both depending if there is any shunt between them then the current through both differs. If there is shunt both currents can be affected somewhat independently. Amplifier is just another impedance for the loudspeaker in the same circuit, like loudspeaker is for the amplifier and one can look and analyze the system from either perspective. One can think the loudspeaker work as amplifier for itself as the back-EMF in the driver, due to fluctuating impedance, is another voltage source within the circuit just like the amplifier is. For this there is plenty of resources found in comparison. Perhaps I'm not using correct search terms to find stuff on amps.
So, you've had some mixed advice and even opinions like mine (as I'm not experienced enough to really give advice) so I guess you need to do some research to determine how to approach the issue and come to a conclusion on your own. This has been the case for me as well and just going through it slowly, figuring out whats going on. Soon to order some crossover components to conduct tests within perhaps few weeks hopefully. Anyway, be it what ever inductance added in series outside the driver would make fluctuation of driver inductance make less current fluctuation in the voltage controlled circuit.
So for any sensible passive crossover/driver load combination that has relatively high minimum impedance that falls withing amplifier specs I don't think amplifiers suffers much increased distortion. After all the driver impedance varies but not that much that it would reduce amplifier load significantly from the nominal, unless silly things in passive network.
Although there is high possibility I'm wrong, because I don't have enough information about the subject, but to me it seems to be wrong conclusion currently that amplifier distortion would get affected by the varying load as long as its within reason. The perspective to look the issue is wrong, its the loudspeaker driver distortion that is affected by amplifier / passive network impedance and not the amplifier by loudspeaker / passive network impedance. They both are in the same circuit together and at least partially same current flows through both depending if there is any shunt between them then the current through both differs. If there is shunt both currents can be affected somewhat independently. Amplifier is just another impedance for the loudspeaker in the same circuit, like loudspeaker is for the amplifier and one can look and analyze the system from either perspective. One can think the loudspeaker work as amplifier for itself as the back-EMF in the driver, due to fluctuating impedance, is another voltage source within the circuit just like the amplifier is. For this there is plenty of resources found in comparison. Perhaps I'm not using correct search terms to find stuff on amps.
So, you've had some mixed advice and even opinions like mine (as I'm not experienced enough to really give advice) so I guess you need to do some research to determine how to approach the issue and come to a conclusion on your own. This has been the case for me as well and just going through it slowly, figuring out whats going on. Soon to order some crossover components to conduct tests within perhaps few weeks hopefully. Anyway, be it what ever inductance added in series outside the driver would make fluctuation of driver inductance make less current fluctuation in the voltage controlled circuit.
Last edited:
And just now found something, amplifier distortion measurement using reactive load and distortion seems to vary between amplifiers, although this is lacking the back-EMF stuff as there is no moving coil? https://www.audiosciencereview.com/...and-fr-into-complex-speaker-dummy-load.21682/
On graphs there distortion of amplifiers vary with frequency, and is different between the amplifiers so perhaps there is something to this. Although the load impedance could be made flat with conjugation network, I'm not sure what the measurements would look like then as they are not provided (to ~resistive load) at least not on the thread linked.
On graphs there distortion of amplifiers vary with frequency, and is different between the amplifiers so perhaps there is something to this. Although the load impedance could be made flat with conjugation network, I'm not sure what the measurements would look like then as they are not provided (to ~resistive load) at least not on the thread linked.
Last edited:
Conjugating for the impedance at the amp terminals creates a resistive load but as you mention, this doesn't concern a Voltage amp.
The distortion in question is said to be connected with the current the driver sees. The conjugate is not in a circuit position to be able to change the current through the driver. In fact the conjugate intervenes and makes the amplifier current different to the driver current so that amplifier current doesn't determine driver current.
Also as you were speculating, larger drivers with low excursions are one example that already tends to perform well in this area without assistance, especially considering the audibility of the kind of harmonic distortions produced by speakers.
Don't get your hopes up too quick, but a first brief measurement session took place. Just trying to get a feel for all of the changes.
The last EQ I did for the 10F's was a quick job, I don't want to do that with these drivers. I made a couple of changes to the pré-EQ to shape it for these new drivers. They produce more bass than the TC9's for the same output of the amp. I still need to tweak the results below 200Hz, as I did for the TC9's, but here's a distortion graph for the 10F array above 200 Hz, taken at listening position (2.7 meter) and at 80 dB:
Pretty sure it will get a bit better, I usually suffer from traffic messing with the results. I've run the first DRC corrections after which I need
to do some manual tweaks. Here's a Stereo result showing lower distortion at 82 Hz:
Nothing to be ashamed about, good to go I'd say. I have limited the SPL levels as not to include various household items into the graphs.
It it's a level of 82 dB at 2.7 meter (I reduce the level for Stereo measurements as to not upset any room accessories). I moved some to the
back of the room but I can hear them sing along at times.
The room has changed after our living room renovation, no doubt. So a few new challenges to make it all work as before. Hope I get the time
to do it, I'm actually exited about it. I like that H2 is above H3 over the entire midrange.
The last EQ I did for the 10F's was a quick job, I don't want to do that with these drivers. I made a couple of changes to the pré-EQ to shape it for these new drivers. They produce more bass than the TC9's for the same output of the amp. I still need to tweak the results below 200Hz, as I did for the TC9's, but here's a distortion graph for the 10F array above 200 Hz, taken at listening position (2.7 meter) and at 80 dB:
Pretty sure it will get a bit better, I usually suffer from traffic messing with the results. I've run the first DRC corrections after which I need
to do some manual tweaks. Here's a Stereo result showing lower distortion at 82 Hz:
Nothing to be ashamed about, good to go I'd say. I have limited the SPL levels as not to include various household items into the graphs.
It it's a level of 82 dB at 2.7 meter (I reduce the level for Stereo measurements as to not upset any room accessories). I moved some to the
back of the room but I can hear them sing along at times.
The room has changed after our living room renovation, no doubt. So a few new challenges to make it all work as before. Hope I get the time
to do it, I'm actually exited about it. I like that H2 is above H3 over the entire midrange.
That's a nice result! Do you have a comparable measurement with the TC9's ?Don't get your hopes up too quick, but a first brief measurement session took place. Just trying to get a feel for all of the changes.
The last EQ I did for the 10F's was a quick job, I don't want to do that with these drivers. I made a couple of changes to the pré-EQ to shape it for these new drivers. They produce more bass than the TC9's for the same output of the amp. I still need to tweak the results below 200Hz, as I did for the TC9's, but here's a distortion graph for the 10F array above 200 Hz, taken at listening position (2.7 meter) and at 80 dB:
View attachment 1084812
Pretty sure it will get a bit better, I usually suffer from traffic messing with the results. I've run the first DRC corrections after which I need
to do some manual tweaks. Here's a Stereo result showing lower distortion at 82 Hz:
View attachment 1084813
Nothing to be ashamed about, good to go I'd say. I have limited the SPL levels as not to include various household items into the graphs.
It it's a level of 82 dB at 2.7 meter (I reduce the level for Stereo measurements as to not upset any room accessories). I moved some to the
back of the room but I can hear them sing along at times.
The room has changed after our living room renovation, no doubt. So a few new challenges to make it all work as before. Hope I get the time
to do it, I'm actually exited about it. I like that H2 is above H3 over the entire midrange.
Can you also sweep from 20Hz?