THAT 4315 question

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
Post Reply
motoindo
Posts: 4
Joined: Sat May 09, 2015 9:04 am

THAT 4315 question

Post by motoindo » Sat May 09, 2015 10:31 am

Hi there,

I've been using Design Note 04 - Improving Loudspeaker Signal Handling Capability (217k) with a THAT4315 to build a dynamic subsonic filter for usage with a Motional feedback loudspeaker to safeguard the bass driver from excessive excursion when fed modern dubstep/chillout/deephouse material 8-) I would like to learn if the mods I made are ok - would it be ok if I post the modified schematic here ? The prototypes are already working fine but I wonder if it makes sense for instance to change the timing cap to a NLC variant. Hereby some images of the test module :

Image
Image
Thanks - Chris
where solid state amplification is like photography, tubes are for painting. sometimes the best solution is to paint over a photograph.

User avatar
JR.
Posts: 3657
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: THAT 4315 question

Post by JR. » Sat May 09, 2015 11:08 am

Discussing schematics is kind of what we do here...

The THAT app note sounds like it was intended for small full range (computer) speakers.

Is your motional feedback bass driver operated full range? Or with a crossover? If there is a crossover you could consider applying the limiting to just the low bass bandpass.

Loudspeaker failure generally falls into two categories, over-excursion and over heating. These are subtly different... Thermal protection wants to take a slower look at average power, and excursion is more short term, but not simple. At least it isn't simple for conventional speakers, if the motional feedback delivers a 100% servo of the bass driver cone motion, excursion limiting will be simpler.

There are professional speaker management DSP systems that already deal with some of these considerations, for convention drivers.

JR

PS: While it's true that the derived dividing network in the THAT app note sums to unity when summed at 100%, in use the limited side will be level reduced so not sum perfectly. FWIW
Cancel the "cancel culture", do not support mob hatred.

User avatar
mediatechnology
Posts: 4906
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: THAT 4315 question

Post by mediatechnology » Sat May 09, 2015 11:12 am

Wow Chris. What a beautiful piece of work. Nice build!

Please post your schematic.
For it to fit as an embedded attachment you might want to limit it to 800-1200 pix or so horizontally.
Much larger than that and the forum sets it up with scroll bars.
Or an alternative is to provide a small embed with a link to a bigger one.

I'm curious about the Motional Feedback aspect since I recently posted about my Panasonic MF-800.

I think generally you'll find a NLC will reduce LF distortion from detector ripple control modulation.
https://ka-electronics.com

Covidland (The Movie) Part 1: Lockdown https://banned.video/watch?id=615f96b3aa816336dbd21792

A voice-over with Orson Welles: "What Is It That You Want?": https://waynekirkwood.com/content/Orson ... u_Want.mp3
Americas Frontline Doctors Lawsuit Motion For Preliminary Injunction Of Covid-19 Vaccine Emergency Use Authorization: https://waynekirkwood.com/images/pdf/Am ... zation.pdf

ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: THAT 4315 question

Post by ricardo » Sat May 09, 2015 6:18 pm

motoindo wrote:I've been using Design Note 04 - Improving Loudspeaker Signal Handling Capability (217k) ...
I'm pleased THAT conducted listening tests on this device. However, there are a few items that show they are electricians rather than horny handed speaker designers. ;)

Their tests show 'better' results for their derived HiPass filter. Some comments
  • The 'detector' needs to be fed with a simulacrum of the cone excursion. A 12dB/8ve LP filter set at the resonance of the (sealed) box is such a beast. Their 'derived HiPass filter' version sorta provides this. The threshold & frequency really need to be adjusted to fit the speaker.
  • However, the 'xover' to the uncompressed signal chain should be a separate circuit block. In their circuit, the amount of 'compression' / protection provided is limited by this.
  • When you are trying to avoid >>10% THD due to excursion limits, agonising over 0.1% isn't productive. You just need to ensure the 'clean' chain is untouched.
  • There are better & simpler methods to get around the the attack & release problems in non-linear LF stuff like this.
I'm not a fan of 'derived' filter topologies. Their theoretical advantages .. perfect transient response, perfect summing bla bla .. NEVER show up in DBLTs. However, their faults .. assymetrical responses in the simple versions like THAT, peaking on both sides for the more complex symmetrical versions .. DO show up when playing with stuff like this.

But there is no doubt that this type of technology, even crudely implemented gives substantial musical advantages. I've done a few commercial products of this nature and experimented with even zanier versions as part of my Powered Integrated Super Sub technology. :geek:

I'm not sure how your motional feedback affects this but note the THAT caveats about where to put the EQ etc. Looking back on my BS, I apologise that not much of this is of practical use to you .. except perhaps for ..
  • make the frequency & threshold adjustable or matched to your speaker
  • have a separate 'xover', at least 12dB/8ve on the HiPass side.
I think generally you'll find a NLC will reduce LF distortion from detector ripple control modulation.
Waz is das? (NLC)

motoindo
Posts: 4
Joined: Sat May 09, 2015 9:04 am

Re: THAT 4315 question

Post by motoindo » Sun May 10, 2015 5:37 am

Image
large
JR. wrote:Is your motional feedback bass driver operated full range? Or with a crossover? If there is a crossover you could consider applying the limiting to just the low bass bandpass.
Thanks for your reply JR :P

Most of your Q's will be answered by the circuit above, it's a 2 way system with a 4th order LR filter in a Sallen Key setup. The subsonic filter is build around a modified substractive filter, I've added an allpass filter in the substractive branch - IC2A - to mirror the phase shift from the primary 70hz lpf - IC1B - so the lpf output at IC2B should display the same 12db/oct slope as the hpf output at IC1B.
JR. wrote: if the motional feedback delivers a 100% servo of the bass driver cone motion, excursion limiting will be simpler.
Tried that exactly, using the MFB signal as input for the RMS detector - output IC5A - but found this to work less optimal then feedforwarding the IC1B output, rationale is that the latter allows the limiter to act before the signal actually reaches the driver.
JR. wrote:There are professional speaker management DSP systems that already deal with some of these considerations, for convention drivers.
Correct, I understand that Klippel is working on a single chip correction chip - sounds promising.
JR. wrote:PS: While it's true that the derived dividing network in the THAT app note sums to unity when summed at 100%, in use the limited side will be level reduced so not sum perfectly. FWIW
Actually I didn't come up with the allpass trick, just borrowed it from Douglas Self his filter bible 8-)
mediatechnology wrote:Wow Chris. What a beautiful piece of work. Nice build!
Like yours too, very neat !!!
mediatechnology wrote:I'm curious about the Motional Feedback aspect since I recently posted about my Panasonic MF-800.
Cool ! please check out this doc ... 1968. I was age 11 when MFB bit me - sounds like we have something in common :shock:
mediatechnology wrote:I think generally you'll find a NLC will reduce LF distortion from detector ripple control modulation.
Definitely worth a try then, thanks.
ricardo wrote:horny handed speaker designers. ;)
:mrgreen: :mrgreen: :mrgreen:

ricardo wrote:The threshold & frequency really need to be adjusted to fit the speaker.
Right now I'm adjusting the 4315 loop so it limits excursion @ 30 hz and 35 watts, the graphs below are taken from a faital-pro 5 inch driver, this is what the FR looks like without MFB:

Image

And with correction on, the little bump @ 70hz was gone after I increased R16 from 4702 to 8202

Image
ricardo wrote:Waz is das? (NLC)
non linear capacitor - capacitance varies with freq.

love to talk but have to run, sunday, mothersday, you know 8-)
Thanks - Chris
where solid state amplification is like photography, tubes are for painting. sometimes the best solution is to paint over a photograph.

User avatar
JR.
Posts: 3657
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: THAT 4315 question

Post by JR. » Mon May 11, 2015 1:06 pm

motoindo wrote:

Most of your Q's will be answered by the circuit above, it's a 2 way system with a 4th order LR filter in a Sallen Key setup. The subsonic filter is build around a modified substractive filter, I've added an allpass filter in the substractive branch - IC2A - to mirror the phase shift from the primary 70hz lpf - IC1B - so the lpf output at IC2B should display the same 12db/oct slope as the hpf output at IC1B.
Looks like you have already invested some time into this project. All crossovers involve trade-offs. L-R is popular because of steep skirts and so much phase rotation that it is back in phase at crossover. This is attractive to professional high power applications to protect drivers, but often for Hifi applications the gentler phase response of lower order crossovers are preferred, when drivers are robust enough to handle the out of band power.
JR. wrote: if the motional feedback delivers a 100% servo of the bass driver cone motion, excursion limiting will be simpler.
Tried that exactly, using the MFB signal as input for the RMS detector - output IC5A - but found this to work less optimal then feedforwarding the IC1B output, rationale is that the latter allows the limiter to act before the signal actually reaches the driver.
Another trade-off, if the limiter is responding to the wrong bandpass errors can occur. I have done tricks for wide-band processing where I used a mix of fast and slow energy input. By cap coupling the fast component, and DC coupling the true bandpass component you might be able to get the best of both worlds, faster response but longer term accuracy. .
JR. wrote:There are professional speaker management DSP systems that already deal with some of these considerations, for convention drivers.
Correct, I understand that Klippel is working on a single chip correction chip - sounds promising.
JR. wrote:PS: While it's true that the derived dividing network in the THAT app note sums to unity when summed at 100%, in use the limited side will be level reduced so not sum perfectly. FWIW
Actually I didn't come up with the allpass trick, just borrowed it from Douglas Self his filter bible 8-)
Crossovers are relatively mature and easier to get right on the bench than in the room, loudspeakers off axis response is not simple and hard to get smooth.

JR
Cancel the "cancel culture", do not support mob hatred.

motoindo
Posts: 4
Joined: Sat May 09, 2015 9:04 am

Re: THAT 4315 question

Post by motoindo » Mon May 11, 2015 2:45 pm

ricardo wrote:Waz is das? (NLC)
checkout page #3 of http://www.waynekirkwood.com/images/pdf ... ressor.pdf
JR. wrote:but often for Hifi applications the gentler phase response of lower order crossovers are preferred, when drivers are robust enough to handle the out of band power.
Most quality drivers are but some of the cheaper models (<100$) are prone to nasty breakup effects which are better controllable with steep filters.
JR. wrote:By cap coupling the fast component, and DC coupling the true bandpass component you might be able to get the best of both worlds, faster response but longer term accuracy.
Hmm, like combining the mfb signal (feedback) with the lpf 70hz output (feedforward) ... that would require 2 separate rms detectors to avoid both signals from interfering with eachother right?
JR. wrote:loudspeakers off axis response is not simple and hard to get smooth.
Absolutely, it was one of the reasons for choosing a small 5" driver:

Image
Thanks - Chris
where solid state amplification is like photography, tubes are for painting. sometimes the best solution is to paint over a photograph.

User avatar
JR.
Posts: 3657
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: THAT 4315 question

Post by JR. » Mon May 11, 2015 3:42 pm

motoindo wrote:
JR. wrote:By cap coupling the fast component, and DC coupling the true bandpass component you might be able to get the best of both worlds, faster response but longer term accuracy.
Hmm, like combining the mfb signal (feedback) with the lpf 70hz output (feedforward) ... that would require 2 separate rms detectors to avoid both signals from interfering with eachother right?
That might work... I haven't done much with the THAT rms engines, I used to roll my own.

At least two rectifiers, and if the outputs are not same polarity flip one. Basically HP the fast part of the envelope, and low pass the important DC content. You still need a final C to ground to filter out ripple...
JR. wrote:loudspeakers off axis response is not simple and hard to get smooth.
Absolutely, it was one of the reasons for choosing a small 5" driver:
A 5" driver in a small box won't make much SPL but probably enough for a near field monitor. Protection is useful if pushing the bass a bunch,, The few times I experiments with small driver LF tricks (like negative impedance drive) the voice coils usually released their smoke before too long. :lol:

JR
Cancel the "cancel culture", do not support mob hatred.

ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

Re: THAT 4315 question

Post by ricardo » Mon May 11, 2015 10:20 pm

You need to do something like this.

You can't do an electronic xover between bass & treble with textbook filters. It's the ACOUSTIC response that has be LR4, Arthur-Smythe, Butterworth bla bla. Hence no values shown around U6a & U3a.

But the xover between 'clean' bass & 'controlled' bass can be exact electronically. These Q=0.5 2nd order xovers are in fact the simplest and most useful of a whole range of xovers of which LR4 is only the second member. The units are connected out of phase with each other .. hence the differential amp used as 'mixer'.

Note the separate xover for 'clean' & 'controlled'. The excursion sensor is the output of U3b. Its cutoff is matched to the speaker/box.

Any EQ (including the MFB) is around the mixer and should give whatever LF target response you plan. ie feeding either R27 or R28 should give the target response at low levels.

R28 may need to be changed due to the VCA requirements, in which case, so do the other resistors around the mixer.

There are other (perhaps better & simpler) ways of doing this properly but this is the closest fit to what you have presently.
________________________
Tried that exactly, using the MFB signal as input for the RMS detector - output IC5A - but found this to work less optimal then feedforwarding the IC1B output, rationale is that the latter allows the limiter to act before the signal actually reaches the driver.
Nothing to do with feedforward/feedback.

The output of IC5A is Acceleration with a bit of Velocity. This will seriously limit higher frequencies well before it protect the unit from LF. You want Excursion which means at least 2 integrators before feeding the RMS detector. My U3b output is a more sophisticated version matched to the speaker.
Attachments
SpkrExcursionLimiter.gif
SpkrExcursionLimiter.gif (24.71 KiB) Viewed 21988 times

motoindo
Posts: 4
Joined: Sat May 09, 2015 9:04 am

Re: THAT 4315 question

Post by motoindo » Wed May 27, 2015 2:03 am

Just a quick reply - busy building prototypes - thanks for your thoughts Ricard, much appreciated !
Thanks - Chris
where solid state amplification is like photography, tubes are for painting. sometimes the best solution is to paint over a photograph.

Post Reply