Fired up one of the channels today. Biasing went smoothly. After dialing it in i connected my scope and function generator. The response seemed perfect... Until i had a slip of the finger and set the input frequency to 100kHz and R125A/B started smoking quickly after 🙁
This was with a full swing output. Is it an expected failure with that input frequency, or might it be a symptom of something else?
EDIT:
Upon closer inspection it was actually R127. 🙁
This was with a full swing output. Is it an expected failure with that input frequency, or might it be a symptom of something else?
EDIT:
Upon closer inspection it was actually R127. 🙁
Last edited:
You should be fine once you replace those smoke zobel parts...
Good to hear that it should (hopefully) be an easy repair 🙂
I noticed also (before burning the zobel network), that the front heatsink was a bit warmer than the back one. This is the one that cools the four transistors closest to the output. Is that any cause for concern? I initially thought it was R125A/B that went smokey due to them being warmer to the touch than the rest of the degeneration resistors also.
I noticed also (before burning the zobel network), that the front heatsink was a bit warmer than the back one. This is the one that cools the four transistors closest to the output. Is that any cause for concern? I initially thought it was R125A/B that went smokey due to them being warmer to the touch than the rest of the degeneration resistors also.
Perhaps at those exceptionally high frequencies the lower trace impedance of the BJTs closer to the load causes current sharing to shift regardless of emitter resistors?. I wouldn’t worry as long as they don’t measure shorted in circuit with the load and power supply disconnected (just remember to hold the measurement for a while as sometimes you will get a short at first contact due to stray voltages). The zobel should have taken one for the team here. If you still have problems, the next place to look might be the feedback path.
nice progress with the GB4 build. Both channels perform as expected. Next is to optimize S/N
Hi there
I think that the quiescent current for the output BJT with 93mA per pair seems very high. It result in a power dissipation of 100W (EF4) with no output load. I tried to find the design decision in the Wolverine thread for such hi class A operating level, with no success.
Usually quiescent current on outputs on class A/B designs is arround 10 to 40mA in my experience. Can anybody point me to a discussion to decide for 93mA? Or any good reason to keep it so high?
I think that the quiescent current for the output BJT with 93mA per pair seems very high. It result in a power dissipation of 100W (EF4) with no output load. I tried to find the design decision in the Wolverine thread for such hi class A operating level, with no success.
Usually quiescent current on outputs on class A/B designs is arround 10 to 40mA in my experience. Can anybody point me to a discussion to decide for 93mA? Or any good reason to keep it so high?
Keeping it simple without taking the base current into account.
Idle current 22mV / 0.22R = 100mA
power dissipation for the EF3-4 with the following rail voltage:
P (56v) = 100mA x 56V x 8 Transistors = 44.8W
P (64v) = 100mA x 64V x 8 Transistors = 52W
P (70v) = 100mA x 70V x 8 Transistors = 56W
Output Transistors are biased at this level to minimise crossover distortion. If you google the Oliver condition you'll see where the 22mV comes from.
Last edited:
Ok 100W for a stereo amp. Still quit a lot in one case, not taken into account drivers and PSU losses.
Not a problem with good ventilation. But as said before almost all class A/B amps to my knowledge have much less quiescent current, with therefore less power dissipation at all.
The only pointer i found was the so called Oliver criteria of about 26mV over emitter resistor. Even hi end commercial amps do not follow this criteria.
This is not meant as negative feedback! Just trying to understand the issue!
Not a problem with good ventilation. But as said before almost all class A/B amps to my knowledge have much less quiescent current, with therefore less power dissipation at all.
The only pointer i found was the so called Oliver criteria of about 26mV over emitter resistor. Even hi end commercial amps do not follow this criteria.
This is not meant as negative feedback! Just trying to understand the issue!
See my article on Class AB biasing.
If you want lower bias current, the emitter resistors should be increased. Note that distortion will increase.
Ed
It is customary among designers of HQ power amps, to design with a quiescent current of approximately 100 mA per output BJT, for a push pull power amplifier with emitter follower configuration. I recommend reading D. Self's book. In the Wolverine case, this also has its origins in the original HK design and its Slewmaster derivatives.
It is not high but optimized. Even so low that Douglas Self argues that it is better to be referred to as class B instead of the often used class AB.
The total quiescent power consumed in the output stage of a PP-EF output stage can be calculated as follows:
Pq = Iq (single) * n output pairs * V+ → V- * n channels.
So with a rail voltage of 63V this is for a stereo EF3-4
0.1 * 4 * 126 * 2 = 100.8 Watts
If you determine the quiescent current by means of a measurement across the emitter resistors, the result of the calculation includes the base current of the outputs.
If you object to the energy consumption of 877 kWh per year at 24/7 operation, for reasons of heat development, costs or climate, it is imo best to start with conversion to EF3-3, which reduces consumption by 25%. I do not recommend lowering the quiescent current, it causes an increase in crossover distortion in the open loop amplifier, which then has to be corrected by the negative feedback.
It is not high but optimized. Even so low that Douglas Self argues that it is better to be referred to as class B instead of the often used class AB.
The total quiescent power consumed in the output stage of a PP-EF output stage can be calculated as follows:
Pq = Iq (single) * n output pairs * V+ → V- * n channels.
So with a rail voltage of 63V this is for a stereo EF3-4
0.1 * 4 * 126 * 2 = 100.8 Watts
If you determine the quiescent current by means of a measurement across the emitter resistors, the result of the calculation includes the base current of the outputs.
If you object to the energy consumption of 877 kWh per year at 24/7 operation, for reasons of heat development, costs or climate, it is imo best to start with conversion to EF3-3, which reduces consumption by 25%. I do not recommend lowering the quiescent current, it causes an increase in crossover distortion in the open loop amplifier, which then has to be corrected by the negative feedback.
Last edited:
There is no issue. If you have specified 4U heatsink, it just gets pleasantly warm, not hot. I can have wolverine on all day long.
Not like running AN39, 40 watts pure classA amp. Its hot even with forced air.
Hi Guy's,
I've finally found some time to work on my chassis. Though you guys might like to see the progress and some images of the completed project.
Wolverine project
I've finally found some time to work on my chassis. Though you guys might like to see the progress and some images of the completed project.
Wolverine project
Very nice Stuart. Unsurprisingly, even better than the professional "gold plated" process control gear of yours that I have seen. At the chocolate factory from memory.
Pointer appreciated. thanks
Thanks for your replies, Just read Bob Cordells "Designing Audio Amplifiers" page 113ff about EF biasing. His measurement with different bias current shows that the minimal crossover distortion could be less than 26mV on RE 0.22Ohm. Could be as less as 15mV/0.22R. (68mA per BJT) The Wolverine settings with 44mV/0.44R fits Cordell's recommendation perfectly.
Nevertheless most commercial Amps have much lower bias current to avoid high power usage and high temperatures.
Can not reduce to EF-3 cause have to drive a pair of vintage Infinity Kappa 9 speakers with terrible low impedance at some frequencies. (High current)
May be i'm trying to reduce to 15mV/0.22R and checking the effect to THD+N on my usual output requirement.