Please review the site Rules, Terms of Service, and Privacy Policy at your convenience. Rules, TOS, Privacy
Get familiar with the reaction system: Introducing the Reaction System

on the reduction of Harmonic Distortion of AMT drivers in lower frequencies

edited April 2019 in Hard Data
Testing out a modification to an AMT diaphragm to reduce harmonic distortion 

The problem:
  • The harmonic distortion (measured as a percentage to the fundamental) in one of my AMT tweeters rises as frequency goes down, hits a peak, but then lowers it's magnitude after passing the peak.
My hypothesis was: 
  • on the lower-frequency range of the AMT driver, physical resonances end up propagating/traversing the diaphragm folds (in ways I have not yet visualized).
  • To reduce the amplitude of these resonances, a viscoelastic compound could be applied to the diaphragm in small amounts, across the tops of the folds, to help absorb the energy of the resonances, thus damping them and the harmonic distortion associated with them.
  • This leaves the pleats largely unaffected, and free to contract/expand normally.
  • use a brand new diaphragm and apply two thin threads to the back of the diaphragm, thinly coated with silicone sealant, across the tops of the folds, in an X-shape.  Re-install and test under same conditions.

Before and after frequency response.  Same drive level (approx 2.82V RMS), same setup, mic distance, location,placement.  Just from one day to the next after installing the modified diaphragm into the same AMT tweeter body.  Tweeter is Hygeia RT-5002.  Response is pretty much identical for all practical purposes.  On the after (red) I forgot to raise the HPF on the dbx 234xs to 1000Hz though.  I believe that is why it is showing just a hair more output below 2k.

Here is the harmonic distortion data, before & after diaphragm modification.  I used slightly more drive level than in the FR graphs, but the same for both HD tests.

Doesn't  look like much, but here is the quantitative data.

  • I "used up" one spare diaphragm in the process, but I do not feel it was a waste. 
  • Further tests will include different orientations of the silicone treated thread across the diaphragm folds, both on the rear and also on the front.
  • For my purposes, it works out well since I am looking to mate an AMT tweeter (not necessarily this one) to a 10" midbass driver, using a large horn, so obtaining the best possible low-end extension and performance is a primary objective.  I used this diaphragm/tweeter as the test piece because it was relatively cheap.  
  • I suspect other larger [rectangular] diaphragms also suffer from un-necessary harmonic distortion content as a consequence of physical resonances traveling along the folds, and treating them in a similar fashion would also decrease distortion as a whole. 
  • I also slightly suspect that square diaphragms might also suffer from resonances, but to a much lower degree.
  • There is an increase in THD near 3,2 kHz.  Hopefully with more trial-and-error, that HD peak can also be tamed.  Time to use up more spare diaphragms. 
Other data:

Modified RT-5002 with Dayton H110 horn.

~2.83V/1m, no EQ FR.  Bone stock vs Horn loaded with modified Diaphragm.

Harmonic Distortion, modified diaphragm, no EQ, 2.83V  measured 10cm from horn. Also +5 dB.

Bone stock Frequency Response (black) vs modified diaphragm (red) with EQ, HPF ~1kHz.

Harmonic Distortion, bone stock no EQ vs modified diaphragm horn-loaded with EQ, ~90 dB reference sensitivity.



  • Nice. Something tells me the AMT industry may already know of this trick, looking closely at stock photos of some Mundorf AMTs. Next step to go "full Mundorf" is to partially obstruct the diaphragm in a few random spots, probably with some ferrous material since this is the motor gap after all.

    I'm not deaf, I'm just not listening.
  • edited April 2019
    You're right, I have the Mundorf catalog and had seen them, but honestly didn't know what they were.  I suppose it's only an issue on some of their AMT's because it seems most of them don't have those "lines", whatever they may be, in the pictures.  Mine is better because it's on the rear of the diaphragm, so it doesn't look ugly.   =)  While lounging around today after running all the data/charts, I figured I could add again more to the rear of the diaphragm to "break up" the diaphragm even further within the "triangles" created by the threads, into further smaller triangles, but not identically so as to not excite further resonances.  Luckily the front plate bars on these AMT's are wide enough that I could also add silicone'd thread to the front of the diaphragm right behind the bars, and no one would be the wiser, except my ears. ;)
  • Pretty incredible improvement! Like the low end with that H110 horn.
    Don, Donno, or "Hey you" all work for me, But never "Mr Johnson"
  • It's rather impressive to me that attaching a string with silicone doesn't appear to affect the frequency response or sensitivity of the AMT.
    I'm not deaf, I'm just not listening.
  • edited April 2019
    Thanks Don.  It's interesting [to me] how I actually came about this.  Before I started ordering from Hygeia, I bought a set of Airborne RT-5002s (identical of course) from a guy on the forum in 2017.  Well since I was later ordering mine (and yours) from Hygeia with the machined aluminum bevel, naturally I had to take apart the Airborne's to see what makes them tick.  Somewhere along the line, I ended up "damaging" the diaphragms.  Both times (shame on me!, LOL) when I removed the side "[" and "]" brackets that physically and magnetically connect the rear to the front plate, and sure enough the pull from the high powered Nd magnets pulled the bracket right into the freakin diaphragm!  Didn't tear it, but "ruined" the folds.   This happened twice, I might add, lol.  So fast forward a year, I was looking at the ruined diaphragms and I figured, what the heck, let me try to straighten them.  So I removed them, and using some business cards cut into strips, I gently coaxed all the twisted/folded folds back into position as straight as I could, both from the front and the rear of the diaphragm.  Even though they were spaced out as they should be, they were nowhere near straight (that was key!), as there were gentle curves to the folds I could not get rid of.

    Now go back to the future, lol, to March of this year when I was on Spring Break from my work.  I ended up testing the FR of the damaged diaphragms, and surprisingly the FR was almost back to normal.  Then the kicker was when I decided to see "how bad" the distortion increased, and surprisingly, the top end seemed almost as good as new, but the low-end had decreased!  When I say low I mean from about 1000-2,500 Hz.  So I begun my gedankenexperiment.  I always had a hunch that the HD spike in the 1,3kHz was due to a resonance.  Then I remembered looking at this video from the DIY AMT thread on  In that video series you can see how even though in theory of operation only the parallel pleats are supposed to move towards/away from each other during normal AMT operation, the tops of the folds also end up vibrating themselves as well, and start to move sideways (exaggerated in that video by purposely driving the diaphragm to the point of non-linear behavior).  Just like a rod suspended between two points;  short enough and it shouldn't wobble too much from external excitation, but if you make it long enough, it ends up being like a guitar string where it will have it's own resonances/harmonic frequencies.  In the case of the damaged diaphragms, the imperfect nature of the now curved pleats is what I felt was curbing the natural resonance that had caused the HD spike near 1,3 kHz.  I felt that the non-linear behavior in the video also applied to this AMT diaphragm since it was longer in one dimension than the other.  Not only that, the more homogeneous a system is, the easier it is to excite a resonance within it, in the absence of damping.  In my mind it was like a normal cone transducer having only one centering suspension (surround or spider) to center the whole moving assembly (in the AMT this would be the outside perimeter equivalent to one centering suspension).  But without a 2nd centering suspension, the voice coil in a cone transducer will have more degrees of freedom with regards to rocking around.  What was "missing" from the AMT diaphragm was it's own "spider" to "keep things centered" (ie: removing a degree of freedom, aka the tops/bottom of the folds immobilized, that is).  Keeping a voice coil in a cone transducer centered with a dual suspension system is the parallel of maintaining the AMT folds from having that extra degree of freedom and allowing that non-linear behavior.  That way, if the peaks/troughs of the folds could be "held" so that they don't move, the pleats themselves will be the only things moving, as they should be, which should lead to lower distortion.  I chose sewing thread because it had to be something lightweight and easy to apply.  Silicone was chosen because it is readily available and stays somewhat flexible after it has cured.  That was my answer to the question: How can I reduce the excitability of the hypothesized resonance without going as far as randomly bending the pleats around like the damaged diaphragms are?  

    I actually expected to loose some sensitivity, and there was a little on the last octave, but I reckoned that since the tops of the folds weren't supposed to move anyways, the silicone'd sewing thread which is extremely thin is not going to interfere with the conductive pleats that are supposed to be [slightly] moving and creating sound.  The pleats themselves shouldn't be moving all that much anyways, under normal operation, so something "grabbing" to the tops of the folds shouldn't phase them either.  I'm only holding immobile what is already supposed to be immobile.  :) Perhaps with more sophisticated equipment/procedures one would see a loss of sensitivity, as I'm only using a Dayton OmnimicV2 system, in an untreated room.  
  • Excellent detail.  I love this stuff.
     John H, thanks to JP I did get that email
  • Excellent detail.  I love this stuff.
    Thanks.  Now the question is:  do I have the balls to experiment with ridiculously expensive diaphragms from my Beyma TPL-150H's and Aurum Cantus AST 30130's ?!?!    Granted, their behavior is much better as they more "gracefully overload" as frequency goes lower (the Beyma motor, even though is encapsulated, is still similar to the Aurum Cantus in that they are both open behind the rear plate since the Nd magnets are to the sides of the diaphragm instead of directly behind like the Hygeia RT-5002).  But I can't help but wonder if their distortion profiles could be improved upon on the low-end as well, since they are both much longer top-to-bottom and side-to-side.
  • Excellent analysis and experiments!
    How are you fastening the siliconed thread to the tops of the diaphragm? 
    You might be able use a weak glue for the test(s) and then take off the thread and return the TPL-150 to (hopefully) stock.
    Don, Donno, or "Hey you" all work for me, But never "Mr Johnson"
  • IMO the value is creating something from nothing, what can you do with the HiVi and Bestons?

    I'm now wondering what the distortion mode is for a bunch of other drivers...

     John H, thanks to JP I did get that email
  • The silicone'd thread is silicone'd to the diaphragm, lol.  I literally coated it with silicone, then immediately laid it across the diaphragm.  So it cured while on the diaphragm.

    On a different note, going back on some previous measurements of mine, I need to stop using this class-D amp for distortion measurements, as they sure did come out lower when I was using a class A/B amp with an even higher drive level!
  • Yes, the amp can be a factor on drivers with ultra-low distortion. The whole signal chain in fact, from personal experience.
    = Howard Stark: "This is the key to the future. I'm limited by the technology of my time, but one day you'll figure this out."
  • edited April 2019

    IMO the value is creating something from nothing, what can you do with the HiVi and Bestons?

    I'm now wondering what the distortion mode is for a bunch of other drivers...

    For planar ribbons, it would be much trickier, IMO.  The disadvantage lies in the fact that if there are spurious, unwanted resonances traveling up & down the diaphragms, to tame them would be to dampen & constrain the planar diaphragm in the direction of travel (front to back), which would then adversely affect FR and/or sensitivity.  In the AMT scenario, the direction of back-and-forth motion of the pleats/traces is perpendicular to the edges of the folds, so it causes little to no restraint of motion.

    Perhaps this is a job for Mundorf Angel Hair (treating planar ribbons).  :)  Extremely lightweight for the space it takes up!  Perhaps it can be used as a dampening material by every so slightly touching the diaphragm, but without adhering to it and adding mass to what is supposed to be a near zero-mass diaphragm.   3.9g of Mundorf Angel Hair takes up more space than 20g of polyfill!

  • edited April 2019
    Getting a quote from Hygeia for a lot of extra diaphragms (14 of each of 5002's and 4001's, amongst other things) to test different variations of my silicone-string modification.  Who said String Theory was only for PhD theoretical physicists?  :)
  • Once again, they are pulling their stalling tactics !  I was hoping they would custom make some faceplates for me, and they said yes, but their machine shop that they outsource this to is claiming too little quantity.  Lets see what I can convince them of, since I do want those extra diaphragms!
  • I just had an epiphany!  If my gedankenexperiment plays out as I anticipate, I will all but eliminate all distortions in an AMT!   :o   =)    

  • Sooner or later some of these companies will realize that they need to pay you as a consultant to make their products better. Thanks for sharing at 0 $$. Stay safe down there  :#
    Don, Donno, or "Hey you" all work for me, But never "Mr Johnson"
  • Of course I exaggerated a little bit on completely eliminating all distortions, but if my scheme works, it should provide even better results than what I previously achieved.  Right now, my worked picked up quite a bit since it's the beginning of the school year, but the hamster wheel is always in warp-speed!
  • Oh, it's on!  Word is getting around of my brilliance!  =)   Someone is going to make HEADPHONES using the RT-5002's as the headphone drivers!  That's a lot of weight!  LOL

    Looks like they developed a new AMT as well, looks to be larger than the 4001, but smaller than the 5002:

    I think I might order a few of these to play with as well.

    I also asked if there was any possibility of them producing a mega AMT like the big Beyma's or the Aurum Cantus AST's, or the Dayton AMTPRO-4.  Let's see what they say.

  • rjj45
    I'm not deaf, I'm just not listening.
  • dcibel said:
    Nope, hadn't seen those.  Look pretty cool!
  • Wolf said:
    That new one looks intriguing.
    Want to order a few?
  • Need more details 'fore I can commit to that.
  • Haven't forgotten about this.  Work comes first, and I been swamped this semester!  I'm on vacations for two weeks starting on Saturday, and I'm excited about continuing my work on this!  =)
  • edited January 2020
    Modified a 2002 and two 5002 diaphragms last night.  Hoping to get to test them today or tomorrow, to see if my cojecture holds water.  No I'm not gonna say what I did.  Not without a signed NDA. :)   If it works, I will try to get the idea sold to some company.  Maybe Legacy Audio!

    You'd be able to see it in-person, but not here.  Crossing my fingers it works.

  • Well, I can say without violating NDA, that your preliminary results looked outstanding. Maybe Purifi would also
    be interested.
    Don, Donno, or "Hey you" all work for me, But never "Mr Johnson"
  • I'm hoping I didn't go overboard with this new modification.   The only thing I worry about is the accumulation of distortion in the upper frequency range.  I noticed in the previous iteration, there was a tiny peak that showed up as a positive gain in distortion, which is likely the result of the shift of energy into higher modes.    I have faith though.

    Oh, and I'm also gonna see if I can twin-stack AMT's to make Super-AMT's, lol.  Sort of making a cheap "Beyma-killer".   I noticed that in the RT-5002, there is some "dead space" (non-magnetic) above (&below) the magnet array that is seemingly there because of the size of the cartridge that is the diaphragm.  I'm going to eventually put two in my mill to shave off the tops to merge them vertically.   I'll have to modify the rest of the steel plates, but that's eazy peazy on a milling machine with some high speed steel endmills.  B)

    Something like this:

    That's about as close as you can get them without messing with the folds themselves.  The separation distance would be about ½" between the actual moving folds.  Imagine that sucker horn-loaded with some good compression!  Betcha it can get low. 
  • Darn it!  I twitched and moved the volume up too much at a time and smoked one diaphragm on the 5002!  lol.  The other surviving diaphragm shows promise, but I think the other one was the ticket!  I better be careful as I only have 8 diaphragms left!
  • edited January 2020
    Here is a stock 5002 vs the 2nd heavily modified one.  I haven't ran the charts/numbers but aside from that rise in distortion near 1kHz, I see an overall reduction in average THD from 1.5kHz+ and a huge decrease in average 3rd order HD from that same frequency and up.  This was just a quick test with the tweeters on the couch, with the mic 3" from them, tested stock one, then immediately swapped in the heavily modified one. Back-to-back. Pretty sure the graphs would be cleaner with testing them in the middle of the room away from all surfaces.  However the raw output of the fundamental is lower starting around 3kHz and below.  But it's a relatively high sensitivity tweeter, so careful notching of the response will still work wonders.

    I'm trying to fix my original modified one with the thread, since the tiny lead wires have all but broken off.  If I can get it working, I will test the three back-to-back.  Well four when I re-do the diaphragm that I smoked.
  • Here is a stock RT-2005 vs a modified one (similar to the 5002 diaphragm  that I just smoked).  Interestingly, the huge peak in the distortion that I had previously measured is not there in the stock one.  Which leads me to think perhaps it was an anomaly that showed up in initial testing.  Overall, from about 3kHz+ there is lower average THD, and lower average 3rd order HD.  The little distortion there is, is mostly all just 2nd order.

Sign In or Register to comment.