Some Interesting Drivers, a New 3-way Project

We were talking about the Kms(x), but the Kms(x) is not responsible for the IMD.
See Klippel information of link shared above.

The Kms(x) is the limiting factor the SB drivers.

Hello,

No, you were speaking of Kms(x). I was speaking of increasing bass distortion and increasing mid-range IMD with distortion reaching 10% with X-Max only reaching +/- 2.8mm.

If you must get into the weeds:

You said that it must be Bl(x) because Kms(x) does not generate IMD. The referenced Klippel paper only speaks of Kms(x) and Bl(x) being the mechanisms of bass distortion. By inference Force Factor must be the Bad Boy.

If you look at the SB17 Bl(x) measurements, between +/- 2.8mm the Bl(x) curve is largely flat it does not account for the measured high levels of bass distortion or the mid-range IMD.

The Klippel paper does not account for: Surround Radiation Distortion (SRD).​



https://purifi-audio.com/blog/tech-notes-1/distortion-the-sound-that-dare-not-speak-its-name-8

Thanks DT
 
  • Like
Reactions: tktran303
Not more expensive compared to a well made passive filter these days.
Perhaps you could elaborate on that a little... What is a reasonable cost estimate for a DIY 3-channel DSP? Let's assume it needs to handle both analog and digital inputs, but no bluetooth streaming. Let's assume it needs a minimum resolution of 96k / 24 bit, and process all the normal formats (FLAC etc). And let's assume that FIR is not required, but if it is available, that is a nice to have feature.

I expect that the passive network for a 3 way speaker to be about $250 for a pair. This is paying full retail price and shipping for individual components, not bulk pricing.
 
What do you constitute as 'very high mH'?! Out of interest. Thanks
My guess would be 6-12mH, depending on your goals.
As cleverly mentioned, 6-12mH is exactly the range.
For an 'average sized' speaker with bass down to 35 Hz > 10mH / 3ohm DCR is the 'ballpark'.
To clarify my description, the "Actively Assisted Rear Radiator" is not in series with any other driver.
Also, just to keep things 'clean', a Zobel network of roughly 47uF / 4.7ohms across the A.A.R.R. is useful.
 
  • Like
Reactions: jimk04
Perhaps you could elaborate on that a little... What is a reasonable cost estimate for a DIY 3-channel DSP? Let's assume it needs to handle both analog and digital inputs, but no bluetooth streaming. Let's assume it needs a minimum resolution of 96k / 24 bit, and process all the normal formats (FLAC etc). And let's assume that FIR is not required, but if it is available, that is a nice to have feature.

I expect that the passive network for a 3 way speaker to be about $250 for a pair. This is paying full retail price and shipping for individual components, not bulk pricing.
That's a bit hard to answer, because there are billions of options and possibilities.
So I am just gonna sketch a scenario or two.

I have been using a HTPC (home theater PC) for about 13 years.
Used to be incl DVD/CD drive.

With a good quality multi channel soundcard you can do all the filtering, incl FIR on the PC.
Most of those cards do miles better compared to your average CD or DVD player.
Even most onboard sound cards aren't that bad anymore.

So I assume the only analog input needs to be from a record player.
The RIAA correction can also be done in DSP.

2nd hand surround receivers costs here about 10-90 bucks or so.
You can also do it with multiple stereo amplifiers.

Even a 5-10 year old PC will be plenty fast for this purpose.

Another idea could be an hybrid idea.
Meaning using active as well as passive filtering.
This works well, because the hardest and most expensive part is always the crossover between woofer and midrange.
In that case we only need two stereo amplifiers or just 4 channels.

Some of those simple Class-D, LM3886 or Class-AB board diy kits work great and cost nothing

Filtering can be done again with a DSP or otherwise with analog active filters.

Or just go full active analog.

There are also some dedicated DSP boards available. 96kHz and 24bit can be done with a simple ADAU1701.
If you can live with 100dB SNR and limited distortion performance.

There are better variants available, but not many perform great. (Something still on my to-do list to make and produce).

Or just go for a 2nd hand dedicated DSP unit.

Maybe 2nd hand Hypex Fusion amp.
Although really keep an eye on the very limited and much lower continuous power.

Now I stop writing.
Let your creativity just look around and discover the endless amount of possibilities.

I personally prefer any type of active filtering for crossovers that filter around <500Hz.
Those are just not easy to do with passive filtering and quickly become expensive if you want to do it well.
 
Hello,

No, you were speaking of Kms(x). I was speaking of increasing bass distortion and increasing mid-range IMD with distortion reaching 10% with X-Max only reaching +/- 2.8mm.

If you must get into the weeds:

You said that it must be Bl(x) because Kms(x) does not generate IMD. The referenced Klippel paper only speaks of Kms(x) and Bl(x) being the mechanisms of bass distortion. By inference Force Factor must be the Bad Boy.

If you look at the SB17 Bl(x) measurements, between +/- 2.8mm the Bl(x) curve is largely flat it does not account for the measured high levels of bass distortion or the mid-range IMD.

The Klippel paper does not account for: Surround Radiation Distortion (SRD).​



https://purifi-audio.com/blog/tech-notes-1/distortion-the-sound-that-dare-not-speak-its-name-8

Thanks DT
You mean Sd(x).

It's a nice hypothesis from Purifi but the significance of it is never proven with objective data.
So we can also not comment on it accordingly.
I have my doubts about it, as can be read in the topic linked before.


Klippel shows multitone distortion, meaning bass as well as midrange.
They very obviously show and tell that non-linear compliance won't provide and IMD products, only HD distortion.

This was also confirmed by Mr Lisbro from Purifi in the topic I just referred too.

So once again, the IMD products you're referring to can't be from the 2.8mm coming from the non-linearity of the compliance. Since the compliance will never result in IMD to begin with.

I would recommend reading Klippel's paper thoroughly, because it's written very well and completely in line with Purifi's thoughts.
In fact. They even mentioned this briefly with an interview with Erin.

I am fine if you still disagree with that
But in that case you challenge Klippel's years of well written research on these subjects.
 
I’m contemplating how nonlinear Km(x) behavior could lead to IMD. Can’t see or understand which mechanism would allow that to happen… so to me the hypothesis doesn’t even hold an edge.
I was just about to add this as well haha.
But this is what we concluded in the other topic, so it's a bit double.

But yes, there are no physical elements that can cause mid-range intermodulation with a non-linear compliance.
The only things that will modulate are the Fs as well as Vas and a bit of Qms (therefor Qts)
 
I’m contemplating how nonlinear Km(x) behavior could lead to IMD. Can’t see or understand which mechanism would allow that to happen… so to me the hypothesis doesn’t even hold an edge.

Hello,

None of the luminaries make a claim about pure clean suspension stiffness Km(x) causing IMD. Do not take the bait. Do not go there.

For grins take a look at Fig. 8 in the referenced Klippel paper. Over a wide range of voice coil displacement the surround stiffness remains largely unchanged. Over the same range of coil displacement the shape and geometry of the half round surround experiences large changes. Depending on + / - displacements the effective radiating surface of the driver can change up to 20%. This up to 20% change in Sd(x) causes Amplitude Modulation that we can hear and IMD that is seen on the FFT plots. This is not caused by Force Modulation or by pure clean suspension stiffness either.

Now that is a mechanism.

Thanks DT

Note:
In the models, Km(x) and Sd(x) are often "lumped" in together. See Fig. 7 in the Klippel paper.
 
Last edited:
  • Like
Reactions: Hörnli
Depending on + / - displacements the effective radiating surface of the driver can change up to 20%. This 20% change in Sd(x) causes Amplitude Modulation that we can hear and IMD that is seen on the FFT plots
It's very close but not quite that.
The effective Sd stays constant, since the change in Sd value is the opposite for max inner vs outer excursion.

Which is fundamentally different compared to Kms(x) or BL(x), where the value change in the same direction on both ends.

Btw, we don't need to go into whole papers to know and understand what modulates what.
Just look at the basic formulas which parameters influence certain sections.
The compliance (where this argument started) has no further influence.

I have never seen any appels to appels comparison with a linear Sd vs a non-linear Sd.
So again, showing actual practical data would give a final conclusion to this hypothesis.

I am fine whichever way it goes, but the fact is that without any true practical data, we are not able and allowed to draw any final conclusions.

But I would suggest discussing this in the other topic, instead overtaking hifijim's thread.
 
I expect that the passive network for a 3 way speaker to be about $250 for a pair. This is paying full retail price and shipping for individual components, not bulk pricing.
In my experience that only works if (and only if) you keep a no nonsense approach. And use core coils and bipolars at the right places. Don’t dream of air coils and film caps all the way. You’ll probably need a correction circuit on the (about) 100Hz resonance if you want to cross below 300Hz. And the highpass components for the midrange often need inductors too. For the BSC either a 3d order highpass for the mid or a LCR correction are the usual suspects.

Btw are you planning to use a laddered approach for the mid-high section?
 
  • Like
Reactions: b_force
Perhaps you could elaborate on that a little... What is a reasonable cost estimate for a DIY 3-channel DSP? Let's assume it needs to handle both analog and digital inputs, but no bluetooth streaming. Let's assume it needs a minimum resolution of 96k / 24 bit, and process all the normal formats (FLAC etc). And let's assume that FIR is not required, but if it is available, that is a nice to have feature.

I expect that the passive network for a 3 way speaker to be about $250 for a pair. This is paying full retail price and shipping for individual components, not bulk pricing.

For an off-the-shelf solution I certainly can't find to purchase in unit quantities (or even 2-4) that is US$250 a pair.

If I develop a prototype and decide to go to market, then yes, getting one for under $250 is not problem.

Here me out. In most commercial speakers with a ported/PR woofer, below Fb excursion just takes off. And so the power handling is not usually thermally limited, it's excursion limited. And so it doesn't benefit from a lot of power, like you one might used a sealed woofer + EQ/LT etc.
You work out the max SPL based on excursion limits and then look at the power. Well it turns out it's not all that much.

Typically typically, 40-80W for the woofer will be enough for a 8" ported woofer. And so with that max SPL, you can see what the voltage swing needed for your midrange, and tweeter. So it turns out the mid and tweeter doesn't really need much more than 20-40W. (Who needs the tweeter to play louder than the woofer)

So now you're looking at a plate amp with 80 + 40 + 40 W, a DSP and some ADC.
If you market for an audience of 500 buyers, for MOQ of ~1000, you can certainly get one in the $125 range.

I'm not talking from a theoretical standpoint. I've seen inside a Kali Audio IN-8 V2 and my goodness! Likewise with the Edifier S1000W.
For a product that is $350 / $225 for a complete speaker- well you can guess BOM can't be that much. But they both include DSP and the appropriate number of amplifier channels and analog inputs. The Edifier has digital and wireless too and chunky cabinet that most DIYers would be proud to have built.
 
  • Like
Reactions: Juhazi and stv
So again, showing actual practical data would give a final conclusion to this hypothesis.

There are many papers published with practical data for Force Factor modulation.

There are few to no papers addressing surround generated distortion. The Purifi site had Tech Notes about the development of their low distortion surround. That material is all now removed from their site. Even their AES publication about Force Factor modulation is now behind a paywall at AES. The old existing models of Force Factor Modulation have been updated to better match the measurement data.

I suppose that a driver manufacture is inclined to keep their technology as IP. You know, secret sauce.
 
There are many papers published with practical data for Force Factor modulation.

There are few to no papers addressing surround generated distortion. The Purifi site had Tech Notes about the development of their low distortion surround. That material is all now removed from their site. Even their AES publication about Force Factor modulation is now behind a paywall at AES. The old existing models of Force Factor Modulation have been updated to better match the measurement data.
Yes the BL(x) story is and was already quite evident?


I suppose that a driver manufacture is inclined to keep their technology as IP. You know, secret sauce.
That makes absolutely no sense.
There is something called patents to protect IP.

Fact is that you can't claim certain things without backing them up with actual numbers.
I am more than happy to believe them, but again without any practical data we simply can't draw final conclusions.

They seem to be very open with data about most things.
So it naturally makes me suspicious for leaving data out.
It could be that those feelings are totally incorrect, but I just don't understand why providing the right numbers has to be so mysterious?

Again, there are patents and also other things to protect IP. So that can't be the reason.
In fact they have a design patent on the surround.
 
For an off-the-shelf solution I certainly can't find to purchase in unit quantities (or even 2-4) that is US$250 a pair.
An active analog filter can be made for 20 bucks at most.

Passive filters are also not off-the-shelf, need to be manually soldered and are always custom made.
So I don't understand why comparing off-the-shelf solutions to custom made solutions?
 
Lars bares all in his interview with Erin:

Erin:
"Everybody asks. Every time I mention that drive unit, or Purifi drive unit, they always ask about the surround"

Lars (laughing)
"Hehe, yeah I anticipated that... that question come up.

So, I should say, 3 years ago we finished the smallest perfect motor, and shortly before Christmas we were measuring it and we thought it was time for champagne and everything. But guess what. Measurements, they sucked. And they barely budged from an ordinary driver. A lot of disappointment so it almost spoiled the Christmas.

So, again, we had to carefully analyse and find what is the source of the problem Well it turns out that a very old a very old problem is that err... ( @lrisbo showing a Peerless woofer to the camera) we have the surround which typically something like a hemispherical profile piece of rubber here. If I push the cone in, then you can see that the crest, the top point of the row, that moves upwards, which means when my cone moves in the negative position, then the effective radiating area is higher than the resting position. Conversely if I do the opposite you can see that the crest of the row is having a smaller radius, actually the radiating area, it now shrinks. It turns out actually that the area of the surround, that can be 10, 20 even 30% of the total radiating area. And you have all this focus on pristine cones and geometry and everybody is using the same flabby piece of rubber. And this thing has been going under the radar. So after discovering this problem err, umm my partner Carsten said Oh wait a minute, there was a paper in the 90s. Turns out there some researchers at the Danish Technical University who discovered this problem, wrote a paper, err. that basically no-one had referenced and.. they were so disappointed that no one had picked it up. But uh, we accidentally discovered that because we now had the perfect motor and we said "Why is the distortion still so high". Turned out that was simply the problem."

Watch on to understand how Lars and Carsten reinvented the wheel (surround)!
 
Last edited: