Yes, only bottom compound pair have 47pF and amp oscillates around 10MHz frequence without it. Upper pair don't need capacitor. This is weird. Who knows the answer?
This is a question I forget to ask...simulation should also oscillate without C2.
This is a question I forget to ask...simulation should also oscillate without C2.
Some updates.
If someone thinks to build this use the following schema, not schema from message #52, this have better thermal stability and can drive real loads. 100mA through output transistors is not enough (my mistake). This one can give 3.3 rms volts to 8ohm resistor which is around 1.3Watts if calculated correct, more than promised one watt.
Also measured distortion to one watt and 8 ohm resistor is 0.1%, quite close to simulated resultat from previous messages. And it looks like this, lower trace in the middle is from HP distortion analyzers output. Bigger trace from amps output.
Maybe some words more. Original design from message #34 have a risk for thermal runaway, I have build and measured also this, so it's more clever to use two diodes in series to bias output stage.
And "if it works don't fix it" the better one "if it works , fix it, until it's broken". 😉
If someone thinks to build this use the following schema, not schema from message #52, this have better thermal stability and can drive real loads. 100mA through output transistors is not enough (my mistake). This one can give 3.3 rms volts to 8ohm resistor which is around 1.3Watts if calculated correct, more than promised one watt.
Also measured distortion to one watt and 8 ohm resistor is 0.1%, quite close to simulated resultat from previous messages. And it looks like this, lower trace in the middle is from HP distortion analyzers output. Bigger trace from amps output.
Maybe some words more. Original design from message #34 have a risk for thermal runaway, I have build and measured also this, so it's more clever to use two diodes in series to bias output stage.
And "if it works don't fix it" the better one "if it works , fix it, until it's broken". 😉
Use of two diodes doubles the temperature coefficient of the bias voltage. If you want to lower the quiescent current and still maintain this, put a variable resistor in parallel with one or both of the diodes. Don’t let it’s value go low enough to completely starve them, though. If that ends up happening, use a TYPE with a lower forward voltage. 1N4001’s are fine here - speed does not matter if the diodes never switch off. Another trick is using one 1N4148 and a small signal schottky, such as BAT43.
Have to try two 1N4001 in series for next version. And bias can go higher around 280mA through output transistors. This one with higher bias sounds and measures better. Not a big surprise.
Yes, why not 24volts.
Needs only some other power supply for this and bigger heat sinks for BD135/136.
Needs only some other power supply for this and bigger heat sinks for BD135/136.
BD135 etc are marginal for a speaker amp. For a few more pennies I would go with TIP41/42 or even 2n6488G/91G, BD909/910 if you can get them. These will survive a lot more abuse. Using the smallest part possible is a recipe for failure. BGW used to make amps with so much silicon that they could withstand the worst abuse possible. Given the price of failure, a wise plan.
For a 24V single rail D44H/D45H can be used. They are far better at low voltage than TIP types or 2N6488 derivatives, and they are higher speed which improves crossover distortion. Don’t run them much higher - if you go to 36V or more use TIP41 or 2N6488 series, or make the jump to TO-3P audio types like NJW0281. But that’s a way bigger amplifier.
I have at home two pairs MJL21193/21194, but they have fT min = 4MHz.
These bd135/136 or mje340/350 could be used as drivers. For sure they will go outside SOA if used at 24volts class A speaker amp output transistors. Just noticed this, good points.
After this mod all changes and it isn't "a simple one watt amp".
These bd135/136 or mje340/350 could be used as drivers. For sure they will go outside SOA if used at 24volts class A speaker amp output transistors. Just noticed this, good points.
After this mod all changes and it isn't "a simple one watt amp".
If you are biasing into class A, a 4 MHz output tranny is more than fast enough. Even biased at 3A at these voltages 2119x’s will stay inside SOA.
Some updates:
-diodes D1&D3 are now 1N4001
-R2 is 220ohm to get more current
-no need for Q6 comp cap (dropped away)
-added zener 6.2V and C3, 10uF
Measures better and maybe also sounds better 🤔
20kHz square is quite clean now. Timediv 10uS/div, voltsdiv 1V/div.
-diodes D1&D3 are now 1N4001
-R2 is 220ohm to get more current
-no need for Q6 comp cap (dropped away)
-added zener 6.2V and C3, 10uF
Measures better and maybe also sounds better 🤔
20kHz square is quite clean now. Timediv 10uS/div, voltsdiv 1V/div.
Define “better”. Better is the one with the right forward voltage drop, and this is better than what was in there before. One does NOT need a high speed diode there. There is ALWAYS forward current flowing so switch off characteristic is unimportant. There is never going to be a reverse recovery spike. Even if they were a couple hundred pF who would care with C1 there?
They are. So what? You think that 1500 uF capacitor speeds things up any?
Actually, on second thought it does, but it’s effect is on the charge storage in the driver transistors, not the current through the diodes. When looked at from this standpoint, one can see that 1500 uF is overkill (smaller works fine).
Actually, on second thought it does, but it’s effect is on the charge storage in the driver transistors, not the current through the diodes. When looked at from this standpoint, one can see that 1500 uF is overkill (smaller works fine).
D4, the Zener is not a good idea, unless you are trying to do some kind of low voltage shut-down. It needs to be high-Z, ~CCS, not a constant voltage. A capacitor on the base of Q1 would be quieter but it's not likely a problem. RCR is probably the best PSRR.
Two diodes (D1, D3) plus the pot is a lot of bias for two driver BEs unless those 1n4007 are so large they never reach their knee voltage with ~2mA.
C1 only needs to be about 100nF, but 10uF would be fine.
Two diodes (D1, D3) plus the pot is a lot of bias for two driver BEs unless those 1n4007 are so large they never reach their knee voltage with ~2mA.
C1 only needs to be about 100nF, but 10uF would be fine.
It isn’t really a “knee voltage”. Vf changes by 100 mV per decade of current, until effects of series resistance take over. 4001’s are being run 2 decades below their max, 4148’s only one. At a given current, below a couple hundred mA, the 4001’s Vf will be about a tenth lower. It’s just a better match for a medium power transistor.
C1 only needs 10-100X the charge storage of the base-emitter plus collector-base junctions of the driver transistors. More is fine, but big electrolytics often make poor high frequency capacitors. The diodes try to maintain a constant voltage between the bases, but they do have dynamic resistance which is non zero, plus the bias adjustment pot. That little bit of R (couple hundred ohms) slows down the charge suckout as the upper cuts off and lower cuts on. Faster diodes won’t help that a bit either, as their dynamic resistance will probably be higher, and then there is the fixed resistor in series. The cap provides a reservoir to dump the charge, and if very large compared to the transistor the voltage between the bases will be held constant as the charge comes out of the base. Low ESR in the cap ensures the time constant is as short as it is going to be (Rbb in the transistor fights you too).
C1 only needs 10-100X the charge storage of the base-emitter plus collector-base junctions of the driver transistors. More is fine, but big electrolytics often make poor high frequency capacitors. The diodes try to maintain a constant voltage between the bases, but they do have dynamic resistance which is non zero, plus the bias adjustment pot. That little bit of R (couple hundred ohms) slows down the charge suckout as the upper cuts off and lower cuts on. Faster diodes won’t help that a bit either, as their dynamic resistance will probably be higher, and then there is the fixed resistor in series. The cap provides a reservoir to dump the charge, and if very large compared to the transistor the voltage between the bases will be held constant as the charge comes out of the base. Low ESR in the cap ensures the time constant is as short as it is going to be (Rbb in the transistor fights you too).
Last edited: