Something weird is somewhere.
Do not blame the final pairs they are very good and suitable ones.
It is almost impossible to destroy an amplifier because of thermal runaway while experimenting: you usually check the temperature from time to time, very often, and you have the time to turn it off if hot. Perhaps you didn't check it that often?
The transistors I burnt are countless, but none burnt because of thermal runaway. The worst happened to me was a blister on a finger.
An oscillation is very unlikely. Unless the aluminum parts are not grounded with the board.
Do not blame the final pairs they are very good and suitable ones.
It is almost impossible to destroy an amplifier because of thermal runaway while experimenting: you usually check the temperature from time to time, very often, and you have the time to turn it off if hot. Perhaps you didn't check it that often?
The transistors I burnt are countless, but none burnt because of thermal runaway. The worst happened to me was a blister on a finger.
An oscillation is very unlikely. Unless the aluminum parts are not grounded with the board.
I checked it after powering up, touching the output devices, and I could not feel them heating up quickly, so I thought it would be ok. I sat down and started to listen, and after a few minutes one channel went 'pop' and there was smoke. They had also played for many hours with low heat before I changed the base resistors.
I don't have any common ground to the heat sinks etc. I never had this in any prototype setup, and it worked fine so far.. I also checked them with square waves with the standard base resistors. One channel had a slight overshoot at high output levels, but it looked fine IMO.
I don't have any common ground to the heat sinks etc. I never had this in any prototype setup, and it worked fine so far.. I also checked them with square waves with the standard base resistors. One channel had a slight overshoot at high output levels, but it looked fine IMO.
Last edited:
The heat sinks (and the cabinet) MUST be connected to the mass via a ground coupler. While experimenting you cannot leave them floating and unpolarized, connect them to the mass on the neg. speaker connector with a wire.
Right after the smoke were the sinks hot?
Are ALL transistors from Onsemi?
Right after the smoke were the sinks hot?
Are ALL transistors from Onsemi?
Hi marigno & others,
I enjoy reading your discussion, but let me add my 2C.
It is impossible to neglect the fact that even the original NHB-108 is high THD amplifier. The original DartZeel specification boasts with >1% THD. By all means THD is an important parameter that should be kept below 0.1% to attain the minimum values for Hi-Fi definition.
It is true, that some like high THD, some don't. It is important to state it clearly. Instead Deletraz presents his own theory of the insignificance of THD. He has his right to his opinion. The vast majority wouldn't agree with him.
After I have measured this clone, I have decided to abandon this project altogether. It is true, at low power this amplifier sounds rather pleasant, however as the power increases it sounds rather odd, for my taste. Measurements show significant increase of distortion.
Let me illustrate this dilemma with the illustration below. Some like the left, some like the right. I personally prefer lower THD and less calories.
I enjoy reading your discussion, but let me add my 2C.
It is impossible to neglect the fact that even the original NHB-108 is high THD amplifier. The original DartZeel specification boasts with >1% THD. By all means THD is an important parameter that should be kept below 0.1% to attain the minimum values for Hi-Fi definition.
It is true, that some like high THD, some don't. It is important to state it clearly. Instead Deletraz presents his own theory of the insignificance of THD. He has his right to his opinion. The vast majority wouldn't agree with him.
After I have measured this clone, I have decided to abandon this project altogether. It is true, at low power this amplifier sounds rather pleasant, however as the power increases it sounds rather odd, for my taste. Measurements show significant increase of distortion.
Let me illustrate this dilemma with the illustration below. Some like the left, some like the right. I personally prefer lower THD and less calories.
Last edited:
No, heat sinks were not hot, but they are heavy and take time to heat up. The rest of the transistors are from Mouser, I think onsemi is the manufacturer yes. I (almost) never buy transistors on ebay etc. Only for some old amps when they are discontinued etc have I done that (and it worked). CCS & drivers were fine, I only replaced the outputs and base resistors, and it was happy again.
Thinking about it, I actually replaced one of the drivers (next to the outputs) at the same time, because the legs were a bit soft from being bent before, so maybe there could have been something related to that.. I'm a bit reluctant to try again though, I don't like smoking amps
Guys, it’s unclear why the output (and not only) transistors failed. I described everything in previous messages. With basic 22 ohm resistors, the heating of the radiators was within 38-40 degrees (thermal sensor and multimeter fluke 187), constant supply voltage 46 volts. I just reduced the base resistors to 11 ohms, the temperature of the radiators rose to 50 degrees at the same supply voltage. Again I ran the amplifier for 4 hours in a cabinet with the door closed, no problems. I’m not afraid, because I wrote that the temperature sensors were installed at 65-68 degrees almost close to transistors to turn off. They have never worked yet. It is unlikely that grounding is to blame for overheating.There are no complaints about the quality of the music. I don’t want to offend anyone, but I wrote all my reports.
No problem, I wanted to try it. Looking at the sim a little bit, I think the Vbe of the drivers and outputs is critical here, and that they track well with temperature. If the die on the outputs heat up quicker than the drivers, it might go into thermal runaway with low resistor values. Once it gets started, there is no stopping it. In my case it could be that starting the amp, or starting the music heated the outputs quicker than the drivers, and that increased the idle current, which heated the outputs more, increased the temperature etc..
I did a quick sim with variations in temperature between the drivers and outputs
Baseline (room temp I guess, no temp directive in ltspice):
10ohms 355mA
15ohms 290mA
27ohms 210mA
Outputs set @70degC
10ohms 753mA
15ohms 597mA
27ohms 410mA
Outputs and drivers set @70degC
10ohms 175mA
15ohms 155mA
27ohms 120mA
This is just a sim, but it seems very possible that there could be thermal runaway if the output dies heat up faster than the drivers. Also is seems the idle current decreases with temperature when the amp heats up (slowly and evenly).
I did a quick sim with variations in temperature between the drivers and outputs
Baseline (room temp I guess, no temp directive in ltspice):
10ohms 355mA
15ohms 290mA
27ohms 210mA
Outputs set @70degC
10ohms 753mA
15ohms 597mA
27ohms 410mA
Outputs and drivers set @70degC
10ohms 175mA
15ohms 155mA
27ohms 120mA
This is just a sim, but it seems very possible that there could be thermal runaway if the output dies heat up faster than the drivers. Also is seems the idle current decreases with temperature when the amp heats up (slowly and evenly).
@Berlusconi
I chose this amp because it was the one hitting my 3 conditions: no feedback from the speakers, availability of the PCB, and easy to modify to use my lot of Sanken final pairs. Just to try, I connected my Acoustat 1+1 to the "original clone" version, but it refused to drive them. The distortion is written in its schematic. Two things are odd in this schematic: the attenuators R7/R5, R8/R6, and the two diodes D3, D4. Once I removed the diodes and added one more pair as a buffer, things changed, it started sounding a lot better. So I was encouraged to go forward with this build. The attenuators (at first sight they seem to be attenuators) seem to be not a problem, but a point of interest. BTW the Acoustat 1+1 has around 1Ohm impedance, too low for a lot of amps.
@Rallyfinnen , @Paroxod4
Always ground any piece of metal part of your build. Otherwise, It can make components oscillate and you don't know what's going on. It happened to a friend of mine, and troubleshooting it on the phone was not simple at all. Now it is one of the first things I ask.
A thermal runaway is not a fast thing, usually you have the time to intercept it. But this requires time and patience because it can happen at any time when you raise the volume. At its edge, the thermal runaway is very fast. I would try a more conservative approach following Paroxod4's suggestions.
@Paroxod4
do you still have D3, D4 on board? Assuming 11Ohm as base resistors, and 70 as hFE of final pairs, the diodes switch on when the current is 3.8A peak on the collector, more or less 60Wrms/8Ohm, or 30W/4Ohm.
I chose this amp because it was the one hitting my 3 conditions: no feedback from the speakers, availability of the PCB, and easy to modify to use my lot of Sanken final pairs. Just to try, I connected my Acoustat 1+1 to the "original clone" version, but it refused to drive them. The distortion is written in its schematic. Two things are odd in this schematic: the attenuators R7/R5, R8/R6, and the two diodes D3, D4. Once I removed the diodes and added one more pair as a buffer, things changed, it started sounding a lot better. So I was encouraged to go forward with this build. The attenuators (at first sight they seem to be attenuators) seem to be not a problem, but a point of interest. BTW the Acoustat 1+1 has around 1Ohm impedance, too low for a lot of amps.
@Rallyfinnen , @Paroxod4
Always ground any piece of metal part of your build. Otherwise, It can make components oscillate and you don't know what's going on. It happened to a friend of mine, and troubleshooting it on the phone was not simple at all. Now it is one of the first things I ask.
A thermal runaway is not a fast thing, usually you have the time to intercept it. But this requires time and patience because it can happen at any time when you raise the volume. At its edge, the thermal runaway is very fast. I would try a more conservative approach following Paroxod4's suggestions.
@Paroxod4
do you still have D3, D4 on board? Assuming 11Ohm as base resistors, and 70 as hFE of final pairs, the diodes switch on when the current is 3.8A peak on the collector, more or less 60Wrms/8Ohm, or 30W/4Ohm.
My boards and equipment correspond to the original ones, only the base resistors are set to 11 ohms. The diodes are in place. The power supply is 46 VDC. All transistors are selected as best as possible, now I don’t remember about everything, but the output transistors have a gain of 110-115.
So the diodes seldom switch on, they would at: 144W/8Ohm (theoretically, you don't reach this power with 46V rails), 72W/4Ohm in case of an hFE=110 (worst case) and a diode Vf=0.6V. Here's why you like your amp! You did a good job, the hFE of final pairs is very important. If you have a spare diode check its Vf then apply the formula (((Vf/R) x hFE) / 1.41) power 2) x Rload and see at which power RMS the diodes switch on.
I'm waiting for a case from Modushop, and I had an idea and had to make some sims:
Increasing the base resistors for the drivers increases the idle current too. Not sure what adverse effects there might be. Tradeoff is that decreasing base resistors on the outputs lowers output impedance, but this increases it slightly, but should be 'safer'. Maybe some 150 Ohm trimmers instead of the driver base resistors could be an interesting experiment..
Comments welcome!
Below are some sim plots of idle current vs chip temperature (drivers/CCS/outputs) with varying base resistors on the drivers.
Standard 20 Ohm resistors:
50 Ohm resistors
100 Ohm
Increasing the base resistors for the drivers increases the idle current too. Not sure what adverse effects there might be. Tradeoff is that decreasing base resistors on the outputs lowers output impedance, but this increases it slightly, but should be 'safer'. Maybe some 150 Ohm trimmers instead of the driver base resistors could be an interesting experiment..
Comments welcome!
Below are some sim plots of idle current vs chip temperature (drivers/CCS/outputs) with varying base resistors on the drivers.
Standard 20 Ohm resistors:
50 Ohm resistors
100 Ohm
I tried trimmers, and ended up with 100 Ohms fo the drivers base resistors. Then it has about 0,6A idle current total (around 0,5A through the outputs). I also did a re-check on the base resistors (with trimmers) for the NPN outputs, and minimum distortion was with 16 Ohms on one board, and 18 Ohms on the other. With these values H2 was basically gone while the amp was in class A. PNP outputs have the standard base resistors.
I ended up using 20 Ohms on the NPN outputs, 'recycling' the base resistors from the drivers. Then H2 & H3 are basically equal in level while in class a, on a low level of abt -80dB with 8 Ohms load, and abt -70dB with 4 Ohm load.
I had to put in alu-oxide (?) washers to keep the outputs from overheating, but after that the heat transfer to heat sink was more than enough.
Part two:
After some listening, I found the sound not as open and detailed in the treble as I remembered it, so I first added some parallel resistors on the driver base resistors of one channel, and I could hear the diff between channels, so I did the same on the other, and the sound was back to the open and spacious sound as I remembered it. (I'm focusing mostly on treble as mentioned before)
Square wave looked fine with the higher driver base resistor values, and bias current was up, and distortion down, but subjective sound was worse IMO This is a confusing journey for me, almost like some black magic! Maybe I should name it the woodoo amp! I always liked taking measurements and improving measured performance in amps, and mostly my ears agreed, in some cases I could not hear an improvement, but at least they did not sound worse!
Anyway, here I am enjoying the sound of this amp again after making it objectively worse.. :S
I did think that bass was a bit better with the higher bias, but mids and treble sounded a bit dull, and similar to what I heard when increasing bias with the base resistors for the outputs (before one channel was smoked).
I guess I have to accept it for what it is, and enjoy the distorted sound!
I ended up using 20 Ohms on the NPN outputs, 'recycling' the base resistors from the drivers. Then H2 & H3 are basically equal in level while in class a, on a low level of abt -80dB with 8 Ohms load, and abt -70dB with 4 Ohm load.
I had to put in alu-oxide (?) washers to keep the outputs from overheating, but after that the heat transfer to heat sink was more than enough.
Part two:
After some listening, I found the sound not as open and detailed in the treble as I remembered it, so I first added some parallel resistors on the driver base resistors of one channel, and I could hear the diff between channels, so I did the same on the other, and the sound was back to the open and spacious sound as I remembered it. (I'm focusing mostly on treble as mentioned before)
Square wave looked fine with the higher driver base resistor values, and bias current was up, and distortion down, but subjective sound was worse IMO This is a confusing journey for me, almost like some black magic! Maybe I should name it the woodoo amp!
I always liked taking measurements and improving measured performance in amps, and mostly my ears agreed, in some cases I could not hear an improvement, but at least they did not sound worse!Anyway, here I am enjoying the sound of this amp again after making it objectively worse.. :S
I did think that bass was a bit better with the higher bias, but mids and treble sounded a bit dull, and similar to what I heard when increasing bias with the base resistors for the outputs (before one channel was smoked).
I guess I have to accept it for what it is, and enjoy the distorted sound!
No, I didn't make the Wolverine. I saw its schematic, it should be a good amp after moving the feedback loop from the speakers' output to the VAS's output. However, after assaying the Circlotron, I will never step back to a N/P push-pull. You have the best possible N/P push-pull because it is a "diamond buffer". IMHO the Wolverine is a step forward in the VAS, and a step back in the back-end. The VAS would be a lot simpler and better by using tubes.
The A60+ is a beautiful project, cascodes in the VAS and the driver, with no feedback at all. You need to accurately match the transistors. I do not understand why they didn't use a triplet in the back-end, the VAS could be overloaded. So the intention is a good one, but the application doesn't follow. Difference from the NHB-108? It is different! Not a single common point! I don't know if I saw the same schematic you intended, please send it to me in a PM, we are already off-topic.
Now it's boxed and decal-tuned (have a decal cutter at work). There are still some minor things to finish up, and maybe play around with the resistors/idle current some more.
Another way to do that could be adding a 'spreader resistor' between the VAS transistors, just a few ohms would increase the current significantly.
After that it's time to put it in the living room and try it for tweeters in the active system. The transformer is from a Onkyo surround receiver, and I was able to split the secondary windings to get 4x22VAC and have about 29V rails after the ripple eaters.
Some pics:
Another way to do that could be adding a 'spreader resistor' between the VAS transistors, just a few ohms would increase the current significantly.
After that it's time to put it in the living room and try it for tweeters in the active system. The transformer is from a Onkyo surround receiver, and I was able to split the secondary windings to get 4x22VAC and have about 29V rails after the ripple eaters.
Some pics: