Magic smoke appeared from my DH-220 power amplifier today when I applied a 1KHz square wave with no load. It was C20, a 0.1µF 100V film capacitor in the output snubber that burnt. The Hafler will be subject of a future repair video.
I built this as a kit in 1979 or 1980.
That's surprising, it sounds as if it's unstable without a load. Unusual for a solid state design.
That's surprising, it sounds as if it's unstable without a load. Unusual for a solid state design.
It was not unstable. I was watching the output on an oscilloscope, the amplifier was still working. I just think the ESR of the snubber capacitor had increased with age and it was dissipating to much power. I will investigate further.
When I made the video of the magic smoke with my cell phone I had already turned it off. I shut it off as soon as I had a wiff of the magic smoke smell.
That's a relief!
Make sure you look in the MHz region too though, output network frying oscillations don't necessarily show up when the timebase is set to view 1kHz (although I would probably expect the resistor to fry before the capacitor, or at least, both together).
P.S. Looking at you're sig, I am sure you are well aware.
Here is a cell phone photo
That brewed up nicely, not quite Rifa grade though!
That's a relief!
Make sure you look in the MHz region too though, output network frying oscillations don't necessarily show up when the timebase is set to view 1kHz (although I would probably expect the resistor to fry before the capacitor, or at least, both together).
P.S. Looking at you're sig, I am sure you are well aware.
The O-scope I was using was a 100 MHz unit at full bandwidth. The square wave looked clean and continued to look good as I smelled the magic smoke. Oscillation is possible, but I think a well used 43 year old capacitor that lived 200 yards from the Pacific Ocean for 20 years just went bad.
Are the bad capacitors "wrap and fill" Mylar?
Are the bad capacitors "wrap and fill" Mylar?
I don't know. I have set the Hafler aside for the moment. I currently have other projects that are higher priority. I will do a more in-depth diagnosis when I have more time to work on this.
That amp was never meant to pass squarewaves at 1Khz with harmonics to at least 50Khz on a continuous basis. That snubber never expected to see so much energy at that high of a frequency on a 100% duty cycle. Your 1Khz squarewave would have had more energy above 20Khz than a 20Khz sinewave. No wonder that thing decided to smoke. Next test, try 20Khz squarewave, both channels driven to full RMS rated output power and see how many milliseconds before the amp totally fries?
That amp was never meant to pass squarewaves at 1Khz with harmonics to at least 50Khz on a continuous basis. That snubber never expected to see so much energy at that high of a frequency on a 100% duty cycle. Your 1Khz squarewave would have had more energy above 20Khz than a 20Khz sinewave. No wonder that thing decided to smoke. Next test, try 20Khz squarewave, both channels driven to full RMS rated output power and see how many milliseconds before the amp totally fries?
I disagree. I have done 1 KHz square wave testing in the past with no issue. The first time was 43 years ago when I assembled the kit. The second time was about approximately 25 years ago when I replaced the power supply filter capacitors.
The snubber consists of a 10Ω 5W resistor in series with the late 0.1µF 100V film capacitor. If the ESR of the capacitor << 10Ω (it should be < 0.1Ω) then the 3dB frequency of the snubber network should be 159 KHz. Since the amplifier is slew rate limited when driven by a 1 KHz square wave the amount of energy that makes it into a properly working snubber should not hurt it.
I have owned this amplifier since 1979, I know it’s full history. This amplifier lived 200 yards from the Pacific Ocean for almost 20 years. It was regularly exposed to “salt air” during that time. That is why there is rust on the case that has bubbled up from under the paint, and signs of oxidation inside and out.
My guess as to what happened is the long-term exposure of the capacitor to the marine environment caused it’s ESR to increase orders of magnitude. The ESR increase simultaneously increased the proportion of power dissipation in the capacitor and lowered the 3dB frequency of the snubber network which also increased the amount of power dissipated by the network.
I have several fix-it projects ahead of this one, but when I work on it I will do a full diagnosis of failure.
What was the amplitude of the square wave at the output when the capacitor fried? That would determine whether this was an obvious overload.
The
rated slew rate for the DH-220 is 30V/µS. If one assumes a maximum 100V step from -50V to 50V, that would translate to a 3.3333µs rise time in SPICE.
Let's assume for argument that the rise time is a faster 3µS for a maximum 100V step at 1KHz. Then how much power is dissipated in the snubber?
To find out I ran an LTSPICE simulation. It turns out the average power dissipation in the 10Ω snubber resistor is 470mW for a 100VPP 1KHz slew rate limited square wave. Therefore the average power dissipation in the series 0.1µF capacitor with an ESR=0.1Ω would only be 4.7mW. Not nearly enough to make it flame.
The peak power in the 10Ω resistor reaches 100W. Therefor the peak power in the capacitor would only be 1W. However, the entire power pulse V(PR1) only lasts 4µs, just a bit longer than the rise time. That works out to be less than a 0.8% duty cycle for the power waveform since it is not rectangular.
The first order thermal time-constant of something the size of the capacitor is at least on the order of tens of ms if not longer. Thus the transient heating is not going to be that great. Again not enough to hurt a properly working capacitor
If the capacitor had degraded with time in the damp salty marine environment as I suspect, then the power dissipation could be much greater as the ESR increases.
Thus the only way that capacitor fried so spectacularly was that is was defective, and even perhaps partially shorted. If the capacitor was working correctly even a full output 1 KHz square wave should not damage it.