This is why I like TO3 transistors and flying wires ( Quad 303 , Naim NAP 250 mK1 ) . TO3 SOA is usually a tad better as a bonus . If the driver stage is designed as if the TO3 will fail the work to replace the output stage is very little . No broken PCB . No two hours doing a 20 minute repair then . Optional capacitor coupling to prevent speaker damage . If the customer breaks his amp and speakers then he did have a seat belt that should have been used . I would cover such a repair the first time as a guarantee job for the amp .
My feeling is whilst it is very proper to design a 99% bomb proof amp the 1 % will catch you out . With some people that would be 3.65 times a year . None of my PA friends talk about amplifiers that never go wrong . If that animal existed I am sure I would be the first to know of it . Some evenings we never get to talk of women or other such distractions , these people certainly take their work home or to the pub . Mostly DI boxes lately .
As regards secondary breakdown how is the better type of MOS FET thought of ? Exicon 10P/N 20 for example ? They seldom harm the driver stage when failing which is useful .
My feeling is whilst it is very proper to design a 99% bomb proof amp the 1 % will catch you out . With some people that would be 3.65 times a year . None of my PA friends talk about amplifiers that never go wrong . If that animal existed I am sure I would be the first to know of it . Some evenings we never get to talk of women or other such distractions , these people certainly take their work home or to the pub . Mostly DI boxes lately .
As regards secondary breakdown how is the better type of MOS FET thought of ? Exicon 10P/N 20 for example ? They seldom harm the driver stage when failing which is useful .
Second breakdown proceeds rapidly once the region is entered. I think the scenario Waly's is referring to is one where the output waveform is near or at a rail when a short occurs. Under this condition it is possible to have a large portion of the rail voltage impressed across the transistor while simultaneously having very high Ic. The rate of current rise will only be limited by any parasitic inductances present. Assuming the rails are directly connected to the filter caps (i.e. no series regulation or other supply current limiting) the power supply impedance will be near zero. The only real limits to current in this case is the transistor itself, the emitter resistor, and the resistance of the wiring. How much the transistor limits current depends on it's base drive at the time of the fault and how much it's beta drops at high current.
You are absolutely right .I think crowbar is to say a dramatic solution . It is possible if doing as you say to protect everything . Remember the speaker that might get damaged might have no spare parts available for it . That spells death if so and a quest for a replacement that might not be as good . Some Radford Tri Star 90 's for example .
I still say as I did before . Use TO3 on flying wires . Design as if the amp will fail . Repair time should be 20 minutes if sensible enough to beleive it will and factor it in . With class D you do not have the luxury of this . Celebrate the class AB advantages .
I still say as I did before . Use TO3 on flying wires . Design as if the amp will fail . Repair time should be 20 minutes if sensible enough to beleive it will and factor it in . With class D you do not have the luxury of this . Celebrate the class AB advantages .
Totally agreed!
I've seen far more such failures than I would ever want to. Drivers are very often to blame, they seem to be installed with much less care than they should have been.
However, I have a feeling that such amps could benefit from carefully designed electronic protection circuits along classic current robbing lines. In my experience, drivers tend to overheat and burn out, so preventing this from happening is not a bad idea.
Obviously, the first step would be to choose more powerful drivers, which will take more to much more before overheating. Next, one really needs to make sure those drivers are cooled as best as possible, which is not always the case.
Actually, this is why I am so hell bent on Motoroloa/ON Semi's MJE family; after using Motorola devices for over 30 years, I have come to rely on their sturdy construction. I also like to hark back to the times when we used Motorola's TO-3 metal cased power devices - one had to have a hammer at hand, or do something incredibly stupid to harm those babes. And let's face it, the sound quality of MJ 21195/21196, relatively slow devices and all, is hard to come by in the modern, plastic TOP-3 world.
John who works with PA equipment was very pleasantly surprised to find a catastrophic amplifier failure resulted in one blown transistor . Previous experience suggested far worse . John was the cause of it and had done something stupid . John supposed the switch-mode PSU had saved the day . On contacting the manufacturer they said it was true . The PSU protection is fast enough to prevent much damage . Up to a point that is OK . One must protect the speakers as a priority . I can see even the crudest of crowbars with one of these PSU's a solution . John uses them increasingly as the company selling them has certificated them ( saves having a separate certificate ) . John prefers class H over D but says the weight difference with D is appreciated . AB double H double D , even so a D weighing 30 Kg .
Bob, I think Baxandall showed this in Wireless World in the early 70's for power amps.
It doesn't seem to be in his WW Amp series but it might have been in a Letter to the editor. I was certainly playing with TPC in the 70's inspired by him and might have had some correspondence with him on the subject.
It was well before Cherry's 1982 JAES article.
Alas, dis beach bum dun hav stuff from True Gurus anymore.
If anyone would like a bit of reading would they like to say if subharmonics are relevant to class D and even conventional amplifiers . I saw someone state without hesitation that they don't exist in the context of class D amplifiers , I know Fourier was troubled by this . My spectrum analyzer thinks they do when a Hypex 180 module . My Ferrograph analogue bridge may have to come out of retirement , hope the book of words is still with it . I am about to construct a 8 th order Bessel or Butterworth filter to help my analyzer see through the fog . Any thoughts ?
http://www.gim.ntu.edu.tw/gia/dizi/diss/ch09.pdf
http://www.gim.ntu.edu.tw/gia/dizi/diss/ch09.pdf
To generate subharmonics, a system has to do something different each time when presented with the same input twice. This implies the need for some sort of "memory" in it. A very simple example is the flip-flop used to generate a sub oscillator in an analog synth. It has 1 bit of memory and divides the frequency by 2.
Class-D amps are full of nonlinear components that could easily have a memory effect. They are close cousins to switchmode power supplies and these often exhibit subharmonic oscillations and chaotic behaviour if not designed quite right.
On the original topic of the thread, Douglas Self's "beta enhanced VAS" doesn't take kindly to being clipped. I'm pretty sure I've seen it generate subharmonics when overdriven at high frequencies. I might have been getting confused with intermodulation from ripple on the rails though.
Class-D amps are full of nonlinear components that could easily have a memory effect. They are close cousins to switchmode power supplies and these often exhibit subharmonic oscillations and chaotic behaviour if not designed quite right.
On the original topic of the thread, Douglas Self's "beta enhanced VAS" doesn't take kindly to being clipped. I'm pretty sure I've seen it generate subharmonics when overdriven at high frequencies. I might have been getting confused with intermodulation from ripple on the rails though.
Last edited:
If I ever get to know for sure I will post on this . The Hypex is very hard to measure . I am sure what it clams will be true . Just want to know the route to that truth . I don't think my analyzer is lying . If not it is the Kings New Suit of Clothes and I refer to what went before in that also . My conventional oscilloscope sees it the same mostly . Output of the switching frequency is circa 400 kHz . It ( due to filtering ) is a sine-wave of about 5% THD with the first harmonic being the 3 rd . The harmonic decay seems exponential and of square-wave ( triangle ) origin as we would expect . The output level 266 mV into 8 R . That's about 20 dB below 1 watt and 40 dB below full power voltage wise . I would guess 5 mW typical due to the nature of the tweeter as a load .
As you say slightly off track . Not really because it might say class AB is still the better choice for many .
As you say slightly off track . Not really because it might say class AB is still the better choice for many .
Measuring distortion on Class-D amps is hard. The designers will invest in one of these to go with their AP test sets. AP High Performance Audio Analyzer & Audio Test Instruments : Accessories
This was my first stab at it . I will fit an attenuator for when above 5 V rms . The op amp will corrupt the results a bit . Better that than working blind . Even hum is hard to get accurate measurements of . MC33079 as a minimum that might work if using a quad package . NE5534 should be OK if using a single ( 10 MHz GBP ) . 1.6 dB loss at 20 kHz is no problem . 30 kHz was what Stereophile quoted in a review . Seems about right .
you can put a passive RC pole in front of the 1st multiple feedback filter and recalc for a odd order filter
Bessel is poor for meaurement accuracy - I'd make it a Butterworth for passband flatness and steeper attenuation slope
the extra passive RC should help rejecting switching frequency ripple
fet input op amps are reputed to be better in general for rejecting emi at the input - there are also AD "highly linear" front end bjt with multi-tanh related flattened gm inputs
Bessel is poor for meaurement accuracy - I'd make it a Butterworth for passband flatness and steeper attenuation slope
the extra passive RC should help rejecting switching frequency ripple
fet input op amps are reputed to be better in general for rejecting emi at the input - there are also AD "highly linear" front end bjt with multi-tanh related flattened gm inputs
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.