By looking carefully, I noticed also the absence of tsunami & Godzilla protection in this amp 🤣🤣🤣Class A can be done without such mistakes.
Yes thats what driving an 8 ohm load with your amplifier setup calls for as long as the heatsinking can handle the accompanying heat dissipation .I think for safety I should set the voltage at 100-125mV. Am I correct?
Last edited:
Noted. Thanks...Non-polar (or two in series, connected back to back).
Yup, already read. 👍Please read post 22 again as I have edited / refined my reply .
Careful with those pots! If they are shorted out, the only limit to the current in the input bias and driver current is the current gain of the respective transistors.
I suggest it is wrong to have a potentiometer like that without a series limiting resistor.
Also, the arrow on the diagram may just be JLH's schematic symbol of a variable resistor, but may be interpreted as clockwise rotation moving the wiper. In the input stage case, a clockwise rotation gives a higher resistance, which reduces the current in the bias circuit, and so to maintain the bias the output voltage moves upwards, so that would be right, I suggest. In the case of the curernt setting, a clockwise rotation again increases the resistance, but that lowers the bias current, whereas I would normally prefer clockwise rotations to increase the current, or voltage.
As there is a wide variation in gains for the output transistors, and operating currents for various output powers and load resistances, choosing a limitng value for the driver current is not so easy. A typical set-up might be 2A and with a gain of 50 the o/p transistors may need 40mA (each) so for a single pair that needs 80mA in the driver, and a base resistance on the control transistor of about 8 ohms. I would recommend a 4.7 ohm as a series limiting resistor, but the potentiometer itself needs to be able to handle that current on just a part of its track. Therefore, the total power in the potentiometer if running at 80mA is 0.3W, and though this should be within the capability of a 0.5W pot, I would probably use at least a 1W component. Just so that the current in the track is within its rating.
I would also suggest measuring the output centre voltage to ground, rather than a supply rail. If you use a DMM it should indicate + or - but an anlogue meter may need to use its reverse connection button (or reverse leads if it does not have one). The reason I suggest this is that the PSU voltage may vary, despite no load, if the voltage acdjustment changes the current. And if you do use the +Vcc to output measurement, you should check the supply voltage in between readings (swap the negative lead between output and ground between settings).
Either way, it is probably an iterative set-up where you may need to set the current to an approximate value, then check the voltage for zero (with shorted input) then reset the current, and continue until you see no further changes. It would help if you have two meters to be able to monitor the current (that is to say, the voltage across an emitter resistor) at the same time as the output voltage.
I suggest it is wrong to have a potentiometer like that without a series limiting resistor.
Also, the arrow on the diagram may just be JLH's schematic symbol of a variable resistor, but may be interpreted as clockwise rotation moving the wiper. In the input stage case, a clockwise rotation gives a higher resistance, which reduces the current in the bias circuit, and so to maintain the bias the output voltage moves upwards, so that would be right, I suggest. In the case of the curernt setting, a clockwise rotation again increases the resistance, but that lowers the bias current, whereas I would normally prefer clockwise rotations to increase the current, or voltage.
As there is a wide variation in gains for the output transistors, and operating currents for various output powers and load resistances, choosing a limitng value for the driver current is not so easy. A typical set-up might be 2A and with a gain of 50 the o/p transistors may need 40mA (each) so for a single pair that needs 80mA in the driver, and a base resistance on the control transistor of about 8 ohms. I would recommend a 4.7 ohm as a series limiting resistor, but the potentiometer itself needs to be able to handle that current on just a part of its track. Therefore, the total power in the potentiometer if running at 80mA is 0.3W, and though this should be within the capability of a 0.5W pot, I would probably use at least a 1W component. Just so that the current in the track is within its rating.
I would also suggest measuring the output centre voltage to ground, rather than a supply rail. If you use a DMM it should indicate + or - but an anlogue meter may need to use its reverse connection button (or reverse leads if it does not have one). The reason I suggest this is that the PSU voltage may vary, despite no load, if the voltage acdjustment changes the current. And if you do use the +Vcc to output measurement, you should check the supply voltage in between readings (swap the negative lead between output and ground between settings).
Either way, it is probably an iterative set-up where you may need to set the current to an approximate value, then check the voltage for zero (with shorted input) then reset the current, and continue until you see no further changes. It would help if you have two meters to be able to monitor the current (that is to say, the voltage across an emitter resistor) at the same time as the output voltage.
The current level depends on the constant current source including Q8 and Q7 - particularly the temperature on Q7. If this is mounted near the output transistors on the heat sink the rise in temperature on the heat sink will increase conduction in Q7 - which will draw more current through R12 at 4k7 so the voltage drop between earth at one end and the base of Q8 at he opposite end will increase and move Q8 base closer to the voltage rail - and keep conduction in Q8 which provides drive to the output transistors in check.Sorry. The bias current depends on the HFE of Q1/Q1A. That will vary a lot with temperature. There is no protection from reverse Vbe on Q2/Q2A. Class A can be done without such mistakes.
Ed
In practice emitter resistors for Q1 and Q1a generate local feedback voltages which have the same effect but the trouble with this is the feedback is instantaneous and the variability flows through to Q3 emitter. In the upper circuit half the negative feedback from Q2 and Q2a is sucked out by the constant current source so there is less variability in the drive of these transistors. This emitter feedback problem in the lower circuit half does not arise if the resistors R14 and R16 are relocated to the collectors of Q1 and Q1a which improves the split phase drive symmetry of Q3. This is emitter placement is conventional practice with Class B Quasi-complementary circuits for good reasons.
Last edited:
mjona - you make an important point about putting Q7 on the same heatsink as (one of or both) of the output transistors, as I have also mentioned in other posts in the JLH thread.
But I don't see that putting emitter resistors into the collectors serves any useful purpose. (They might in a quasi complementary design in class AB with PNP drivers feeding through to them) . To ensure current sharing, the emitter resistors need to remain as emitter resistors!
The bias resistors can be reworked to reduce the effects though by splitting them into two. Use two 15k (or 16k if you insist on keeping close to the original value) instead of the 8.2k, and wire them from the base of the driver one to each emitter of Q1/Q1A. Similarly use two 4.7 (or 4.3)k resistors wired from the emitter of the driver, one to each emitter of Q1/Q1A. That removes the effective loading of the emitter resistors on those while maintaining the effectiveness of the feedback.
(However, it may be better to simply double up on the 8.2k and 2.2k resistors rather than using 16k/4.3k so that the high frequency response is maintained with parallel output transistors, but that may need further optimisation).
EdGr - OldDIY and I have both mentioned that the current is stable, but not necessarily constant, with one important caveat. For the MJ15003, it has been pointed out that the current gain changes by about 30% from 25 to 150C, so that limits the maximum current change provided that the heatsink is large enough to prevent the junction temperatures rising above say 100C. The heatsink temperature should be kept to "hand hot" or below, for safety, or it needs a stand-off cage. And as mjona mentions, the current source prevents the current becoming excessive. With Q7 in thermal contact, the current will be held more closely, but even in the original bootstrap design, a large enough heatsink was able to keep the current under control. People often become concerned about the thermal stability of a silicon transistor due to the large Vbe variation with temperature, but with constant current drive, the gain change with temperature, and the gain roll-off with current, are the parameters dominating the response.
The stability of the current was a concern at the time JLH published his original back in 1969. It isn't the greatest means of controlling the current - a complementary design was published shortly after, possibly by Nelson-Jones, with a more conventional bias stabiliser, but as many people have found, the JLH design works.
But I don't see that putting emitter resistors into the collectors serves any useful purpose. (They might in a quasi complementary design in class AB with PNP drivers feeding through to them) . To ensure current sharing, the emitter resistors need to remain as emitter resistors!
The bias resistors can be reworked to reduce the effects though by splitting them into two. Use two 15k (or 16k if you insist on keeping close to the original value) instead of the 8.2k, and wire them from the base of the driver one to each emitter of Q1/Q1A. Similarly use two 4.7 (or 4.3)k resistors wired from the emitter of the driver, one to each emitter of Q1/Q1A. That removes the effective loading of the emitter resistors on those while maintaining the effectiveness of the feedback.
(However, it may be better to simply double up on the 8.2k and 2.2k resistors rather than using 16k/4.3k so that the high frequency response is maintained with parallel output transistors, but that may need further optimisation).
EdGr - OldDIY and I have both mentioned that the current is stable, but not necessarily constant, with one important caveat. For the MJ15003, it has been pointed out that the current gain changes by about 30% from 25 to 150C, so that limits the maximum current change provided that the heatsink is large enough to prevent the junction temperatures rising above say 100C. The heatsink temperature should be kept to "hand hot" or below, for safety, or it needs a stand-off cage. And as mjona mentions, the current source prevents the current becoming excessive. With Q7 in thermal contact, the current will be held more closely, but even in the original bootstrap design, a large enough heatsink was able to keep the current under control. People often become concerned about the thermal stability of a silicon transistor due to the large Vbe variation with temperature, but with constant current drive, the gain change with temperature, and the gain roll-off with current, are the parameters dominating the response.
The stability of the current was a concern at the time JLH published his original back in 1969. It isn't the greatest means of controlling the current - a complementary design was published shortly after, possibly by Nelson-Jones, with a more conventional bias stabiliser, but as many people have found, the JLH design works.
Last edited:
john_ellis - Okay, I think we agree.
I would feel okay if this were a 1W amplifier, but at 170W, I want more stable bias.
Ed
I would feel okay if this were a 1W amplifier, but at 170W, I want more stable bias.
Ed
mjona - you make an important point about putting Q7 on the same heatsink as (one of or both) of the output transistors, as I have also mentioned in other posts in the JLH thread.
But I don't see that putting emitter resistors into the collectors serves any useful purpose. (They might in a quasi complementary design in class AB with PNP drivers feeding through to them) . To ensure current sharing, the emitter resistors need to remain as emitter resistors!
The bias resistors can be reworked to reduce the effects though by splitting them into two. Use two 15k (or 16k if you insist on keeping close to the original value) instead of the 8.2k, and wire them from the base of the driver one to each emitter of Q1/Q1A. Similarly use two 4.7 (or 4.3)k resistors wired from the emitter of the driver, one to each emitter of Q1/Q1A. That removes the effective loading of the emitter resistors on those while maintaining the effectiveness of the feedback.
(However, it may be better to simply double up on the 8.2k and 2.2k resistors rather than using 16k/4.3k so that the high frequency response is maintained with parallel output transistors, but that may need further optimisation).
One could use small value base stopper resistors to help current sharing. I don't see this as an issue if the output transistors do not turn off which is the case in a Class A circuit. It could be different in Class B if parallel transistors switch off at different points within the crossover region due to current variations. If high power unit for 4 Ohm loads is required one could build two stereo units and bridge each one of these. The proportions could require housing in a rack case cabinet where all the connecting wires can be routed to and from inconspicuously.
One could use small value base stopper resistors to help current sharing. I don't see this as an issue if the output transistors do not turn off which is the case in a Class A circuit. It could be different in Class B if parallel transistors switch off at different points within the crossover region due to current variations. If high power unit for 4 Ohm loads is required one could build two stereo units and bridge each one of these. The proportions could require housing in a rack case cabinet where all the connecting wires can be routed to and from inconspicuously.
I'm not sure why you think base resistors would help current sharing, just because the transistors are working in Class A (or Class B for that matter). The point is that the emitter current, being far greater than the base current, will create a small voltage with a low value resistor which will have far greater effect on current sharing.
If the transistors have different gains, the use of base resistors will not change the currents to any significant degree, and if the transistors are mismatched in Vbe, at low currents they still won't have a significant effect. Base resistors may help if the transistors are matched - but then they may not need anything particular for current sharing! And of course it is generally very easy to obtain matched devices of the same type when purchasing several.
Usually the reason for using base resistors is as a high frequency response limitation, either as an RC effect on the base capacitance or gate capacitance in the case of a MOSFET, plus damping especially in MOSFET circuits for internal or external lead inductance.
Base resistors have also been used to reduce the effects of Vbe variations in circuits which depend on controlling current into the base, traditionally old saturated core switching converters, but not for parallelling.
I'm not sure why you think base resistors would help current sharing, just because the transistors are working in Class A (or Class B for that matter). The point is that the emitter current, being far greater than the base current, will create a small voltage with a low value resistor which will have far greater effect on current sharing.
If the transistors have different gains, the use of base resistors will not change the currents to any significant degree, and if the transistors are mismatched in Vbe, at low currents they still won't have a significant effect. Base resistors may help if the transistors are matched - but then they may not need anything particular for current sharing! And of course it is generally very easy to obtain matched devices of the same type when purchasing several.
Usually the reason for using base resistors is as a high frequency response limitation, either as an RC effect on the base capacitance or gate capacitance in the case of a MOSFET, plus damping especially in MOSFET circuits for internal or external lead inductance.
Base resistors have also been used to reduce the effects of Vbe variations in circuits which depend on controlling current into the base, traditionally old saturated core switching converters, but not for parallelling.
Last edited:
Have you looked at the Wolverine output stage lately this has low value base stopper resistors which limit base current. If these are low wattage resistors these will serve as fuses in the event of a base short to emitter or collector.One could use small value base stopper resistors to help current sharing. I don't see this as an issue if the output transistors do not turn off which is the case in a Class A circuit. It could be different in Class B if parallel transistors switch off at different points within the crossover region due to current variations. If high power unit for 4 Ohm loads is required one could build two stereo units and bridge each one of these. The proportions could require housing in a rack case cabinet where all the connecting wires can be routed to and from inconspicuously.
I'm not sure why you think base resistors would help current sharing, just because the transistors are working in Class A (or Class B for that matter). The point is that the emitter current, being far greater than the base current, will create a small voltage with a low value resistor which will have far greater effect on current sharing.
If the transistors have different gains, the use of base resistors will not change the currents to any significant degree, and if the transistors are mismatched in Vbe, at low currents they still won't have a significant effect. Base resistors may help if the transistors are matched - but then they may not need anything particular for current sharing! And of course it is generally very easy to obtain matched devices of the same type when purchasing several.
Usually the reason for using base resistors is as a high frequency response limitation, either as an RC effect on the base capacitance or gate capacitance in the case of a MOSFET, plus damping especially in MOSFET circuits for internal or external lead inductance.
Base resistors have also been used to reduce the effects of Vbe variations in circuits which depend on controlling current into the base, traditionally old saturated core switching converters, but not for parallelling.
These may not swamp the non-linear voltage of the power transistor emitter diode to the same extent as an emitter resistor but any resistor in series with a non linear emitter diode resistance including the transistor base I see as a good thing. The emitter diode resistance re' varies with a.c. current and temperature.
There is a current thread where zero emitter generation has been discussed and the results are surprising.
Nelson Pass raised this question and if you read his post https://www.diyaudio.com/community/threads/class-ab-biasing-article.393566/page-2 post 23 you will see that drain and collector feedback is not so far fetched as one might have imagined.
Would that be the Wolverine circuit which also uses emitter resistors?
The emitter resistors will act to distribute the current sharing properly. The base resistors are sprog stoppers but if they are also fuses, that's not the same design consideration as current sharing.
The problem with resistors being used to linearise an exponential response is that they cannot work over a large range. Either they are linearising, or they aren't. As you said, the emitter resistance is dependent on the operating conditions. At low currents, the exponential term of Vbe will dominate, and re will depend on the Vth/Ie term unless high values of resistance are used. In which case they will limit the current drive at high currents.
And that raises another point: in many amplifier circuits, the drive to the output transistors (and drivers) is generally speaking a current, rather than a voltage. Therefore, the overall response is actually less dependent on the exponential Vbe of the transistor than the more linear current gain -but, of course, that situation changes at higher audio frequencies when the voltage amplifier stage becomes more voltage-driven due to compensation capacitors giving local feedback. So at higher frequencies, the exponential characteristic becomes more significant, and generally increases distortion. One reason TMC or TPC is used, which have the effect of maintaining current drive for a greater part of the audio spectrum.
I take it you mean zero emitter degeneration. It is no surprise that zero degeneration gives the best response simply from the point of view of having the highest possible gm. Even though I suggested that current drive is effective in the voltage amplifier stage, some voltage increase is obviously nevessary to increase Vbe to increase the output current, and if gm is high, the overall gain is high. The use of emitter resistors is, once again, a necessity rather than desirable simply to control currents - either quiesecent current in the case of Class AB, or for current balance in multi-device output stages in class A or AB. Or, indeed, in Class A circuits other than a single-pair JLH.
I have not seen Nelson Pass's XA25 circuit, and cannot comment on that. I could accept that collector or drain resistors can be used to control current in circuits where, say, an N-channel or NPN device is used with a PNP driver attached (in the same way as complementary feedback pair circuits work), so that there is a feedback signal derived from it. Otherwise, the high output impedance from a collector or drain will mean a low value will have negligible effect on current balance.
The emitter resistors will act to distribute the current sharing properly. The base resistors are sprog stoppers but if they are also fuses, that's not the same design consideration as current sharing.
The problem with resistors being used to linearise an exponential response is that they cannot work over a large range. Either they are linearising, or they aren't. As you said, the emitter resistance is dependent on the operating conditions. At low currents, the exponential term of Vbe will dominate, and re will depend on the Vth/Ie term unless high values of resistance are used. In which case they will limit the current drive at high currents.
And that raises another point: in many amplifier circuits, the drive to the output transistors (and drivers) is generally speaking a current, rather than a voltage. Therefore, the overall response is actually less dependent on the exponential Vbe of the transistor than the more linear current gain -but, of course, that situation changes at higher audio frequencies when the voltage amplifier stage becomes more voltage-driven due to compensation capacitors giving local feedback. So at higher frequencies, the exponential characteristic becomes more significant, and generally increases distortion. One reason TMC or TPC is used, which have the effect of maintaining current drive for a greater part of the audio spectrum.
I take it you mean zero emitter degeneration. It is no surprise that zero degeneration gives the best response simply from the point of view of having the highest possible gm. Even though I suggested that current drive is effective in the voltage amplifier stage, some voltage increase is obviously nevessary to increase Vbe to increase the output current, and if gm is high, the overall gain is high. The use of emitter resistors is, once again, a necessity rather than desirable simply to control currents - either quiesecent current in the case of Class AB, or for current balance in multi-device output stages in class A or AB. Or, indeed, in Class A circuits other than a single-pair JLH.
I have not seen Nelson Pass's XA25 circuit, and cannot comment on that. I could accept that collector or drain resistors can be used to control current in circuits where, say, an N-channel or NPN device is used with a PNP driver attached (in the same way as complementary feedback pair circuits work), so that there is a feedback signal derived from it. Otherwise, the high output impedance from a collector or drain will mean a low value will have negligible effect on current balance.
An optocoupler with an executive circuit is used (on collector current sensors).Nelson Pass's XA25 circuit
- Home
- Amplifiers
- Solid State
- Class A bias help