Supertweeter Amplification and Protection

Yesterday I found that one channel of my supertweeters had died. Each channel has a forward firing unit and a backward firing unit. The forward firing unit is a Dynaudio D21AF, which has on-axis response to 40kHz. The backward firing unit is a Vifa NE19VTS-04, which also has on-axis response to 40kHz. Both drivers in the left channel had died and measure open. The drivers in the right channel seem to be ok.

I'm a follower of the idea that supertweeters make a difference, and in sighted testing experiments they always have. I have not felt that the effect was large enough to pass a blind test so I haven't bothered. I still believe the effect is real and wish to continue using supertweeters. But only after the recent failure did I decide to move my supertweeter assemblies (a tower made from a speaker stand and old speaker) out from in between my two main speakers (Acoustat 2+2 C-modded) to check that out, and I discovered that moving all that stuff out from in between the speakers was a bigger positive effect than I ever (imagined?) the powered supertweeters to be providing. So my next generation supertweeter system will be mounted more discretely behind and possibly also in front of the main speakers rather than to the side, where they interfere with the normal dipole and front wall reflections. So I'm not going to resume supertweeting until I have a whole new physical arrangement. Plus I need to obtain replacemet drivers. Both kinds are currently unobtainable new but I've ordered both on an auction site. One seller first time around cancelled the order due to unavailability so I'm not counting on anything until I have the replacement drivers in hand.

But I'm wondering if I should re-think the power amplifier part of the setup as well. I've been using an ATI 1502, rated at 150W/channel (probably has at least 200W before clipping). This seems like it might be excessive power, though both drivers seem to have ratings which suggest it is OK. I'd long planned to build a 25W/channel class A amplifier instead. But I'm not even exactly sure 25W would be sufficient. Most of the time the output is very tiny fraction of a watt, but perhaps for some brief instants it isn't. And if a 25W "class A" amp actually has more like 50W peak power, it could still burn the tweeters in a similar way. So then what about 15W, etc.?

Whatever happened did not destroy the amplifier, which is still being used because it's a 33 year old survivor. I could continue to use the same 150W amplifier and add a speaker fuse. I'm thinking a 1 amp fuse or less in slo blow. If I were sure that only tiny power were needed, I could add an attenuator, say with 1 ohm on the bottom and 11 ohms above for a fraction of 1/12. This would only add 1 ohm series impedance, which is probably fine. I have always used both digital electronic and physical crossover on the tweeters. The digital part is a 17kHz high order high pass linear phase FIR filter generated using RePhase and running in a miniDSP OpenDRC. The physical part is 0.47 uF capacitors in series with with each tweeter, which serves to protect it as well as eliminate hum from the amplifier. At 20kHz this capacitor has about 17 ohms impedance anyway. (That also increases the power required from the amplifier. If I made a 25W amplifier, it would have to be very hum free so I could avoid using this "protection" capacitor or use a larger one.)

In fact I wonder how the drivers got destroyed. Since I can't directly hear the supertweeters, I have no idea how long the right channel had actually been dead. I only decided to test the tweeters when one other thing in the series of components used--a Behringer DEQ 2496 which I used to monitor the supertweeter signal and fine tune the level in the digital domain, had died and was not passing signal. But after removing that, I found the supertweeters were still dead in one channel and I traced the problem to open windings in both drivers.

(Note: after the SPDIF digital signal passes through the DEQ for final monitoring and adjustment, it goes to an Emotive Stealth DC-1 dac for conversion to analog to produce the unbalanced analog signal that drives the ATI 1502 amplifier. So even when the DEQ died, it could not have directly produced a dangerous voltage at the input of the amplifier. The DC-1 was generally set to +7dB amplification and was turned even higher during testing, but it would have to receive an intact digital signal to be producing any output at all.)

I had been successfully using this arrangement without issues for over 4 years, including a large number of power failures and restarts, etc. Nothing damaged the tweeters until recently. They were measured and working fine a month ago or so.

My guess is that at some point an input cable came disconnected and produced a full power 60Hz from the amplifier. If this were hard clipping the amplifier, it could look like a 60 Hz square wave. That might deliver enough power to burn out the tweeters even through the 0.47uF series capacitor. That kind of issue could be a problem regardless of using a low power amplifier, and to protect against it either a fuse or an attenuator might be good (but might also be ineffective).

Although I was mostly very careful, this could even have happened while I was in the process of testing the tweeters after the digital equipment failed.

So the question is, should I use fusing, attenuation, or some other kind of protection on the tweeters next time around? Should I be using a lower power amplifier (what kind and how much power) instead of or in addition to protecting with a fuse or whatever?
Last edited:
  • Like
Reactions: 1 user
I suggest you to disassembly one tweeter to check "how it is blown", I mean if it got low freq. as you suspect the coil should show mechanical damage (over excursion); but if it blew because of high power/oscillation then you should see a "cooked" coil.
Once you know what happened you can decide how to protect it.
Externally, the NE19VTS dome is smashed into the plastic protector, so it looks like overexcursion occurred, or maybe it was just persistent over temperature that caused everything to melt or buckle. The D21AF dome has a dimple. I had figured the Dynaudios (rated at 600W fwiw) would be indestructible. Since the amplifier is fine, and fully protected against DC etc with a relay, and since the 0.47uF capacitor, low frequencies not likely a problem I believe. The capacitors measure fine, not shorted.
In testing the amplifier, it's remarkably well behaved. If I pull an RCA input plug out half-way, with no other plug connected, there's only tiny hum just above the noise floor, even if amp plugged into different AC line than the generator (nominally they are plugged into same outlet, however the AES/EBU source for the dac...the plugged into a different line). I was using very solid Blue Jeans LC-1 audio cable which has Canare plugs--they do not come out very easily or by accident. So there is no evidence so far supporting the "loose" or "accidentally pulled" plug theory. If I only pulled one plug, the ground would still be made by the other plug. That was the most I did during testing.

I could have accidentally turned the miniDSP doing the High Pass crossover to the "flat" position. That would have been catastrophic and I should have immediately noticed. With all the downstream digital amplification it would be dumping a highly overamplified and clipping full range signal to the power amp. I'd expect to at least hear a pop or some kind of garbage, which I don't recall happening. But I may have turned the knob by accident within the last few months when no music was playing, so the burnout occurred when system was turning on sometime later. This now seems like better explanation than loose cable (which simply could not have happened by accident anyway, nor was a loose cable found when I started investigating problem, though I did pull one cable by accident IIRC thinking it was the speaker cable.

I'd like to know more about the "lamp" protection idea, but it looks like a DC lamp is simply wired across the lines so that it shorts under especially high voltage. That would probably work ok, especially if the lamp is placed AFTER the capacitor, so the amplifier would not see a short as such, just a very small capacitor.

I'm also thinking of using two fuses, one fast and one slow, in series of course. Perhaps 1a fast and 0.5a slow.

My biggest question is really "do I need a smaller amplifier." But it seems like I should not use a lower power amp unless I change the capacitor to something larger or remove it altogether, because of the voltage loss at the capacitor around 20kHz. A fuse or lamp protection looks like a better idea and probably a good idea in any case.

Also in the past I always left the ATI amplifier turned on. It has very low idle consumption. But from now on I plan to be automatically switching that on and off with the main power amp, so bad stuff doesn't happen when I'm not around to notice.
Last edited:
The burst of noise could have happened on an automatic startup when I wasn't there to hear it. I need to protect against the possibility of selecting a non-crossover program on the miniDSP. Problem is, I have 3 identical miniDSP's, and by default the software always loads only one set of options #1-#4. I'd always been saving the super tweeter program in #1, the panel option in #2, the subs in #3, and a flat option in #4. Going forwards I'm going to have have only the correct crossover for tweeters and subs, so I can't select anything wrong. I'll save the default file that miniDSP loads in multiple versions, and copy over the default file for each one I'm going to load.
I already had a "high pass filter" in the form of a capacitor in series with the tweeter. Instead of being lower than the crossover frequency, I chose to make it even higher, effectively, a 40kHz capacitor cutoff (0.47uF capacitor in series with 8 ohm tweeter). Cranking up the level to compensate, that effectively gives me a further 6dB boost above 20kHz, which is useful in extending the useful output to 40kHz from a 4 ft distance.

The smaller capacitor also more easily lets me use a high grade film, and suppresses hum and all other undesired LF.

So I thought I was good already, but apparently I wasn't.
I was using Dayton Audio 0.47uF polycarbonate caps rated at 400V. (I assume that's a 400V DC rating.) I'm wondering if those could have failed in some extreme amplifier overload situation. They still measure fine (zero conductance) at low voltages. In any case, they failed to prevent damage it seems they should have prevented, so I'm planning to get something different next time, either polypropylene or polystyrene. Metallized polypropylene are self-healing and available in ratings up to 1200V with "automobile grade" meaning they are very robust to vibration and heat. I'm thinking of getting 1000VDC and 600VAC rated cap by Kemet (100,000 hours) for about $10 each (C4ASNBU3470A3EJ). I could get mere 630DC (310AC) for about $1, it seems each extra uptick in specs costs a lot.

In addition to fusing and other new measures to prevent catastrophe. This is not the first set of super tweeters I have fried. After the previous catastrophe, I decided to make the protection caps much smaller and have a dedicated DAC (with relay that switches outputs off until warmup) to prevent the ubiquitous turn on surges in the analog output of the Behringer DEQ 2496. That kept me from having any failures for 5 years.

I had figured nevertheless that the Vifa NE19VTS-04 tweeters or something similar would still be readily available, and at $20 they were too cheap to worry too much about fusing. But now they've become very hard to get (I'm paying over $100 for a no-brand-name replacement which looks identical) just like the Dynaudio D21AF's and there appear to be fewer options than in 2018. Supersonic response is not commonly produced, there's little market for it.