Interesting to know, why? 🙂
Here is a model of the modified Leach Poweramp with a THD of 0,000.05% in the amplification of the signal with frequency of 20 kHz to 40 V / 8 Ohms.
Look at the R10, R11, Q22, Q21 and a current mirror with unequal resistance of resistors R5, R13. If there are problems with a positive temperature drift quiescent current Q9, Q10, it is necessary to put a diode in series with R12.
As some have suggested, I have tried a new VAS again, the Hawksford type. Now voltage across 10 ohm VAS resistors is 100mV (10ma). It is rock steady at that voltage. The measured DC bias measured between 0V and negative feedback point (2nd and 3rd stage power transistors not in circuit) is 0.2-0.4mV.
I will put the output stage back in circuit now.
I will put the output stage back in circuit now.
Last edited:
It's all good. What made you change circuit to control the quiescent current of Q3, Q4?
Hi Guys
I would lose the current mirrors and keep the EF-VAS, as the latter provides an order of magnitude THD reduction.
The feedback loop is nothing that Leach used and I would change it to something more conventional. Sometimes a high-frequency oscillation is triggered by a low-frequency instability.
I would also remove the 22ks in the hum filters, if you want them just to be hum filters. If you want them to regulate, change the 22ks to zeners.
LC wrote: " I'm really getting sick viewing the same old crap in almost every new thread started"
So don't look and don't comment.
Have fun
I would lose the current mirrors and keep the EF-VAS, as the latter provides an order of magnitude THD reduction.
The feedback loop is nothing that Leach used and I would change it to something more conventional. Sometimes a high-frequency oscillation is triggered by a low-frequency instability.
I would also remove the 22ks in the hum filters, if you want them just to be hum filters. If you want them to regulate, change the 22ks to zeners.
LC wrote: " I'm really getting sick viewing the same old crap in almost every new thread started"
So don't look and don't comment.
Have fun
Last edited:
Hi. I haven't read everything, but I would try a few things to experiment (I'm referring to your original schematic):
- Remove T4, T7, T9, R7, R12, R13, R18 and R20, and the same on the other side
- Decrease R16 a bit, maybe 3.9k?
- Increase R6 to about 1.5k (or where you will end up with 1.5 V over it)
- Determine your amplifier stage current by setting R29 (should have about 0.3 V over it)
My reasons are that you can start simple, and work it more complex, but then you have a better understanding of the design points you wish to follow, and can calculate them easier rather than guessing. By decreasing R16, you will have 2 mA in the tail, giving 1 mA in each pair. It also gives more current to play with (you can keep your resistors a little smaller). I have found it easier working with 1 mA than 0.7 mA, and I find the extra bias helps. Then eliminating R7 and R18 will eliminate that forced local feedback (you don't really need to stabilize the emitters here - you can even use small resistors, 56 ohm or less if you really want to), giving you more open-loop gain, and thus justifying getting rid of the current mirrors. Doing all of this will make your base voltage on Q2 very predictable, which makes your voltage on R29 more predictable.
To stabilize this part of the circuit, I would try shunting C9 to the collector of Q2, rather than Q4, since Q2 and Q4 do not make a Darlington. You may even try changing Q4 for a PNP, but you will need to change the design a little here, but Q4 being a PNP will give you much better thermal stability. As an NPN, you already have at least 350 mW on it at your original design point, and I'm afraid this thing will run like the wind towards thermal breakdown, cooking Q4 and R29 in the process. You can replace it directly with the circuit shown below (NPN part), and this should give much better thermal stability, while keeping your design points and topology more or less the same. Oh, and a capacitor from base to collector on Q4 might be worth a try too, say 100 - 150 pF (circuit below already includes it).
(Replace MJE's with whatever you prefer, and heatsink them)
Also, working with small current and small resistors is quite difficult; you need to give quite a lot of tolerance. The reason is that Vbe voltages can vary tremendously from simulations to real life specs. The compound pairs shown give this tolerance, and will be happy with (I'm guessing) up to 50 mA, which is a massive tolerance.
I may be way off, but if you would like, give this a try.
- Remove T4, T7, T9, R7, R12, R13, R18 and R20, and the same on the other side
- Decrease R16 a bit, maybe 3.9k?
- Increase R6 to about 1.5k (or where you will end up with 1.5 V over it)
- Determine your amplifier stage current by setting R29 (should have about 0.3 V over it)
My reasons are that you can start simple, and work it more complex, but then you have a better understanding of the design points you wish to follow, and can calculate them easier rather than guessing. By decreasing R16, you will have 2 mA in the tail, giving 1 mA in each pair. It also gives more current to play with (you can keep your resistors a little smaller). I have found it easier working with 1 mA than 0.7 mA, and I find the extra bias helps. Then eliminating R7 and R18 will eliminate that forced local feedback (you don't really need to stabilize the emitters here - you can even use small resistors, 56 ohm or less if you really want to), giving you more open-loop gain, and thus justifying getting rid of the current mirrors. Doing all of this will make your base voltage on Q2 very predictable, which makes your voltage on R29 more predictable.
To stabilize this part of the circuit, I would try shunting C9 to the collector of Q2, rather than Q4, since Q2 and Q4 do not make a Darlington. You may even try changing Q4 for a PNP, but you will need to change the design a little here, but Q4 being a PNP will give you much better thermal stability. As an NPN, you already have at least 350 mW on it at your original design point, and I'm afraid this thing will run like the wind towards thermal breakdown, cooking Q4 and R29 in the process. You can replace it directly with the circuit shown below (NPN part), and this should give much better thermal stability, while keeping your design points and topology more or less the same. Oh, and a capacitor from base to collector on Q4 might be worth a try too, say 100 - 150 pF (circuit below already includes it).
(Replace MJE's with whatever you prefer, and heatsink them)
An externally hosted image should be here but it was not working when we last tested it.
Also, working with small current and small resistors is quite difficult; you need to give quite a lot of tolerance. The reason is that Vbe voltages can vary tremendously from simulations to real life specs. The compound pairs shown give this tolerance, and will be happy with (I'm guessing) up to 50 mA, which is a massive tolerance.
I may be way off, but if you would like, give this a try.
- Status
- Not open for further replies.