If there is a limited selection range of NP0 ceramic capacitors in terms of capacitance (nothing over 100pF) and voltage rating of 50 volts as in hobby shops in Australia and New Zealand, the only available choice is buy the specified part from a warehouse source.
The claimed THD is 0.002% and S/N ratio is -125 dB. These results are ascribed to special efforts with circuit board layout and component choice.
The power stage and smoothing capacitors all reside on a single PCB from the original circuit. The unwanted parts, SMD circuitry, has been sawed off. Needless to state, this has been done after the circuit was inspected to decide whether this was possible. So, large smoothing capacitors will always be connected to the power stage. This means, there is a real danger, that in the event of failure, a very large current surge passes through the power stage. Although, for a small interval this can be damaging.
My strategy is to make sure the amplifier is very well behave with a low voltag power supply paying attention to stability. By the latter, I mean, bias and power stage's quiescent current stability. As far as self oscillations are concerned there are none. A frequency counter didn't find any. If I want to confirm this further, I can use an oscilloscope albeit a USB digital one from Hantek (200MHz, 1G samples per second).
My strategy is to make sure the amplifier is very well behave with a low voltag power supply paying attention to stability. By the latter, I mean, bias and power stage's quiescent current stability. As far as self oscillations are concerned there are none. A frequency counter didn't find any. If I want to confirm this further, I can use an oscilloscope albeit a USB digital one from Hantek (200MHz, 1G samples per second).
While rotating the VBE multiplier's preset to find the right bias, the current drawn from the power supply, at first remains almost constant at around 20mA, but then, it reaches an avalanche like state with the current drawn increasing to around 26A (~13A per rail).
When I cut the thick and wide tracks on the original PCB to insert the 0.1 Ohm resistors, I measured the resistance across the gaps so formed and confirmed the resistance increased. This should imply the tracks were successfully severed.
When I cut the thick and wide tracks on the original PCB to insert the 0.1 Ohm resistors, I measured the resistance across the gaps so formed and confirmed the resistance increased. This should imply the tracks were successfully severed.
A heavy current implies there must exist an easy path shorting the +ve and -ve supply rails. This path can only mean the power transistors are not working alternatively. An explanation for this can be the driver bases not being turned sufficiently off. Since, I used a diode in parallel with a capacitor in series with the bases, this may be the cause. Mooly has expressed his concern about this circuit element and advised me it may cause problems.
As a caution not to blow components, I am hesitant about testing while this monstrous current is flowing.
As LTSpice's simulations did not discover this heavy current condition, I am feeling in the dark. It seems, simulations are not that predictive about amplifier behaviour.
As a caution not to blow components, I am hesitant about testing while this monstrous current is flowing.
As LTSpice's simulations did not discover this heavy current condition, I am feeling in the dark. It seems, simulations are not that predictive about amplifier behaviour.
Yes, it could be that as the bias current increase it cause the amplifier to oscillate. If the frequency is high enough then this will cause cross-conduction in the output stage as each half can not turn off fast enough. The result is that both sides of the output stage conduct heavily.
I suggest you post the .asc file for the latest version of your circuit to see why you are getting this result. It seems clear you paid no heed my advice about using safety resistors in series with the supplies. If the is a current surge at turn on output transistors may have become suspect. On the other hand some of these may be in a state of mutual conduction - not able to turn off as suggested by Mooly.
I am attaching the circuit with the latest updates.
C1: marked as 222J100. Capacitor is a small grey rectangular box. This is used as a 220pF capacitor but doubts are creeping in with all these failures.
R38: is actually a 220 Ohm variable resistor used to change the bias voltage. The 110 Ohm value is half 220 Ohm. LTSpice does not have a variable resistor.
D13 up to D16: have been added to protect the large electrolytic capacitor. Value 1N4148
I wonder if there is a problem around D5 and D6. These diodes prevent the charge being pulled out of the base of the drivers.
Try adding something like a 220 ohm across each as a test. The bias preset range will alter and you may need to alter the resistors around the preset to be able to bring the current down to zero. R35 may need to be as high as 680 or 820 ohm.
Try adding something like a 220 ohm across each as a test. The bias preset range will alter and you may need to alter the resistors around the preset to be able to bring the current down to zero. R35 may need to be as high as 680 or 820 ohm.
R21 and R24 have to supply large amounts of current through one resistor - this would have to be a high power wire-wound component which would have a small amount of inductance. This is usually ignored in simulations.
With multiple output transistors sharing R21/R24 as their respective collector loads while having wire-wound resistor in their emitter connection to the supply rails could be a problem in combination with wiring and physical layout.
The driver transistor emitters where the carrier charges originate share the same connection whereas the output arrangement is a complementary feedback one between driver and multiple outputs.
On paper the latter will have identical characteristics that don't necessarily reflect reality - slight differences in current gain length of wiring, and traces etc.
Forgetting about stability issues, Silicon Chip power amplifier circuits run the supply wiring close to emitter resistors to have some mitigation of radiated fields to lower distortion.
Some of the problems mentioned might be avoided if the compensation for the circuit is changed such the Vas collector capacitor is taken to the inverting input so it is this signal that dominates at critical frequencies rather than spurious effects in the driver and output area.
I have deleted the files I drew up for this circuit earlier so I am commenting off the cuff. I don't know what good it might do to increase R21/R24 to 0.22R to see if the inductance values are closer on either connection to the outputs.
With multiple output transistors sharing R21/R24 as their respective collector loads while having wire-wound resistor in their emitter connection to the supply rails could be a problem in combination with wiring and physical layout.
The driver transistor emitters where the carrier charges originate share the same connection whereas the output arrangement is a complementary feedback one between driver and multiple outputs.
On paper the latter will have identical characteristics that don't necessarily reflect reality - slight differences in current gain length of wiring, and traces etc.
Forgetting about stability issues, Silicon Chip power amplifier circuits run the supply wiring close to emitter resistors to have some mitigation of radiated fields to lower distortion.
Some of the problems mentioned might be avoided if the compensation for the circuit is changed such the Vas collector capacitor is taken to the inverting input so it is this signal that dominates at critical frequencies rather than spurious effects in the driver and output area.
I have deleted the files I drew up for this circuit earlier so I am commenting off the cuff. I don't know what good it might do to increase R21/R24 to 0.22R to see if the inductance values are closer on either connection to the outputs.
Once again, I tried the amplifier off-bias to avoid the current overload condition. To my surprise the amplifier played music brilliantly clear. In fact, at this moment, I am listening to music amplified by this very amplifier.
I was using a multimeter to measure the current drawn from the power supply. However, the point of measurement was from between the rectifier bridge and the smoothing capacitors. This means, although the current is unidirectional, it is pulsating at twice the mains frequency. The multimeter should have been giving me false readings far lower than the average current. Therefore, to find the correct bias, I will need to use a far more sophisticated instrument to allow me to correctly read the currents drawn.
I am using the same power supply that delivered the heavy overload current without issues. It uses a modified large UPS transformer and a 35A rectifier bridge.
I am using the circuit as originally posted.
Thanks to all.
I was using a multimeter to measure the current drawn from the power supply. However, the point of measurement was from between the rectifier bridge and the smoothing capacitors. This means, although the current is unidirectional, it is pulsating at twice the mains frequency. The multimeter should have been giving me false readings far lower than the average current. Therefore, to find the correct bias, I will need to use a far more sophisticated instrument to allow me to correctly read the currents drawn.
I am using the same power supply that delivered the heavy overload current without issues. It uses a modified large UPS transformer and a 35A rectifier bridge.
I am using the circuit as originally posted.
Thanks to all.
You should be looking at dc current drawn when you set up your amplifier, in particular you want to see how much the output stages draw with no input signal and no speakers connected.
All the current from each output half including the driver transistors has to pass through R21 or R24 whose values you know so you can make use of Ohms law by measuring the voltage drop across these. The safest way to do this is to put one probe on a point which is connected to earth - the black speaker binding post for instance since you can tighten this a little so the connection is secure.
Then with your spare hand in your pocket then probe separately the voltages at the junction between R21 and the emitter of Q13 and note the reading and do likewise at the junction between R24 and Q14. Ignoring polarity differences of the readings add the numbers together and divide the result by R21+R24 resistance (0.2R).
All the current from each output half including the driver transistors has to pass through R21 or R24 whose values you know so you can make use of Ohms law by measuring the voltage drop across these. The safest way to do this is to put one probe on a point which is connected to earth - the black speaker binding post for instance since you can tighten this a little so the connection is secure.
Then with your spare hand in your pocket then probe separately the voltages at the junction between R21 and the emitter of Q13 and note the reading and do likewise at the junction between R24 and Q14. Ignoring polarity differences of the readings add the numbers together and divide the result by R21+R24 resistance (0.2R).
R21 and R24 are made of an array of ten 1 Ohm metal film resistors in parallel. This has the effect of reducing the inductance by a factor of 10. I am ignoring the added inductances of the PCB tracks to have this arrangement. The ten resistors are rated at 3W each. So, that resistance is capable of 30W of power dissipation.
To set the correct bias I will measure the voltage across these resistors and use Ohm's Law.
To set the bias I had to add another two resistors to the circuit. I added two Emitter 3.3 Ohm resistors to the driver transistors. As a result the distortion figure rose slightly. I am attaching the schematic with the latest changes.
The next thing is to decide from which point in the circuit the earth terminal will be connected. This is the 0V terminal. The power feeding wires are located between the rectifier bridge and the smoothing capacitors which mean a very large pulsating current will flow in them. This has the effect of creating a virtual signal source that has 100Hz as the fundamental frequency. Needless to state I want to avoid having this virtual source injecting its harmanics in the signal path. This means, I must make sure, in the event of an earth loop, the loop current does not flow through the part where there is the smoothing power pulses. I am assuming, according to my logic, given an amplifier and a signal source with separate power supplies, there is always an earth loop. My aim is not to allow such an earth loop to inject its signals.
The fact that I had to solder and desolder parts for several times to get the a bias that is acceptable is discouraging me.
The next thing is to decide from which point in the circuit the earth terminal will be connected. This is the 0V terminal. The power feeding wires are located between the rectifier bridge and the smoothing capacitors which mean a very large pulsating current will flow in them. This has the effect of creating a virtual signal source that has 100Hz as the fundamental frequency. Needless to state I want to avoid having this virtual source injecting its harmanics in the signal path. This means, I must make sure, in the event of an earth loop, the loop current does not flow through the part where there is the smoothing power pulses. I am assuming, according to my logic, given an amplifier and a signal source with separate power supplies, there is always an earth loop. My aim is not to allow such an earth loop to inject its signals.
The fact that I had to solder and desolder parts for several times to get the a bias that is acceptable is discouraging me.
If that is cause of all this trouble, I will remove D5 and D6, the parallel capacitors, and replace them with 22 Ohm resistors. Without those diodes the bias voltage requirement will be lower and I will have far more control over it. The reason is for the same preset change, the change in bias voltage will be lower as the multiplying fact of the VBE multiplier would be lower. As it is, the factor is about 4, without the diodes, it will be about 2.