Hi Guys
For the sim that had "several amperes" of idle current, what was the idle current value? Obviously that sim demonstrates full class-A behaviour.
The EF triple version of this buffer that I bread boarded uses 2SC/2SA where BDs are indicated. I used a different bias reg configured like a CFP. Self showed this in his book and even demonstrated that it functioned better than his own single-BJT reg with output compensating R added - but then he dismissed the CFP version as being "too complex" for a bias regulator.
I will try a couple of other bias regs that are current servos, since they monitor the output stage current directly and then adjust the drive to the triple or mosfet gate. Each uses two BJTs but one is complimentary and fits better with the "look" of Lineup's buffer.
One point about current sources: The two-BJT version using local feedback by monitoring the emitter current of the pass element was shown by Self to assist with slew rate, compared to the "passive" current source with just a voltage reference tied to the base. He tended to use a single control BJT to set multiple current sources to save pennies and a bit of board space. I prefer to make each CS independent - BJTs are cheap.
I love this circuit. I think it is brilliant!
At the risk of looking like back tracking: How well would the mosfet circuit perform with an EF driver stage?
Have fun
Kevin O'Connor
For the sim that had "several amperes" of idle current, what was the idle current value? Obviously that sim demonstrates full class-A behaviour.
The EF triple version of this buffer that I bread boarded uses 2SC/2SA where BDs are indicated. I used a different bias reg configured like a CFP. Self showed this in his book and even demonstrated that it functioned better than his own single-BJT reg with output compensating R added - but then he dismissed the CFP version as being "too complex" for a bias regulator.
I will try a couple of other bias regs that are current servos, since they monitor the output stage current directly and then adjust the drive to the triple or mosfet gate. Each uses two BJTs but one is complimentary and fits better with the "look" of Lineup's buffer.
One point about current sources: The two-BJT version using local feedback by monitoring the emitter current of the pass element was shown by Self to assist with slew rate, compared to the "passive" current source with just a voltage reference tied to the base. He tended to use a single control BJT to set multiple current sources to save pennies and a bit of board space. I prefer to make each CS independent - BJTs are cheap.
I love this circuit. I think it is brilliant!
At the risk of looking like back tracking: How well would the mosfet circuit perform with an EF driver stage?
Have fun
Kevin O'Connor
Last edited:
THe terrible mistake had 100mA through the BD139-BD140 drivers.
And the current in Output Was Like 5A idle!!!
But as you noticed eventhough it was a herrendous current the performance was GREAT.
It is like I predicted that this circuit benefit by driven in Pure Class A.
Good ideas.
I will add another transistor too the VBE multiplier. Makes the current settting more stable and high impedance.
In stead of LED as Voltage Reference I will try the two BJTs way.
As you say it can improve the circuit slewrate.
For me an input inpedance of 3 kOhm is no problem.
The voltage amplifier would easily cope with this.
Evenso 3 kOhm is no problem for a good Op Amp, liike OPA2134.
---
---
---
What I should do now is to concentrate to make this BJT triple output as good as possible.
Make the two changes mentioned by Struth.
More work, but I am happy with it
Kevin.
I have tried using BJT drivers to feed MOSFET gates in a SPICE.
It was in a power amplifier.
I thought, like you, it would be an improvement.
But I was disappointed.
It made no difference at all.
So, from my experience I would not do it.
But certainly I will go back to the MOSFET circuit.
Just have to make the Triple EF work the best possible first.
I am now ready to do the SPICE testing of the NEW triple EF version.
It certainly looks to be faster.
Kevin mentioned slewrate.
Now, this power follower circuit has an upper -3 dB at 8 MHz
as can be seen from the diagram I post here.
There is nothing bad about the behavior
This is done with two capacitors across the ouput.
They are chosen for stability: 1.2 nF each
The NEW circuit can be seen in diagram 2.
The currents all are at idle. 100 mA quiescence current.
I be back with the THD figures of the circuit.
It certainly looks to be faster.
Kevin mentioned slewrate.
Now, this power follower circuit has an upper -3 dB at 8 MHz
as can be seen from the diagram I post here.
There is nothing bad about the behavior
This is done with two capacitors across the ouput.
They are chosen for stability: 1.2 nF each
The NEW circuit can be seen in diagram 2.
The currents all are at idle. 100 mA quiescence current.
I be back with the THD figures of the circuit.
Attachments
Your idea is well worth to consider.
You mean we do not use the power transistors of last stage,
and make a less powerful output.
I certainly will look into it.
Andrew, this is no joke. It is a really good circuit for a follower.
But to be sure I understand you, please give more input what you have in mind
Hi Lineup,
I appreciate you work on this buffer, but have you checked it's behaviour on capacitive loads? Think that some unpleasant surprises are hiding there. As someone noted some more advanced compensation techniques should be used, or an output inductor.
I'm looking for a (near enough) perfect follower stage for an Aikido, hope that something along your idea could lead to that direction.
I appreciate you work on this buffer, but have you checked it's behaviour on capacitive loads? Think that some unpleasant surprises are hiding there. As someone noted some more advanced compensation techniques should be used, or an output inductor.
I'm looking for a (near enough) perfect follower stage for an Aikido, hope that something along your idea could lead to that direction.
It was a casual remark in another Thread about how little gain was/is needed in the Power Amplifier feeding a horn loaded compression driver.
I suggested that a Follower/Buffer may be all that is needed, by my suggestion has not developed any discussion.
A 12W to 25W ClassA amplifier is probably sufficient for a 110dB to 120dB compression driver.
12W into 8ohms is only 14Vpk. +-20Vrails is plenty high enough. +-12Vrails might even do, if I settle for a bit less maximum power. 6W into 115dB/W @ 1m = 118dB from a pair @ 2.4m. That seems to indicate that a Buffer is probably good enough, maybe with a switchable +6dB signal level, gain stage. Even 2.2Vac from a CDP into 110dB is still 103dB @ 2.4m That is still loud.
I suggested that a Follower/Buffer may be all that is needed, by my suggestion has not developed any discussion.
A 12W to 25W ClassA amplifier is probably sufficient for a 110dB to 120dB compression driver.
12W into 8ohms is only 14Vpk. +-20Vrails is plenty high enough. +-12Vrails might even do, if I settle for a bit less maximum power. 6W into 115dB/W @ 1m = 118dB from a pair @ 2.4m. That seems to indicate that a Buffer is probably good enough, maybe with a switchable +6dB signal level, gain stage. Even 2.2Vac from a CDP into 110dB is still 103dB @ 2.4m That is still loud.
Last edited:
Well to get back on topic, there is always a possibility to adjust a circuit to 12 Watt output. This goes for this my buffer as well. As you say 20 Volt rails and to make it suitable for 8 Ohm output.
Now, the latest version I show can be made to acommodate for your needs. The difference is not that big.
I have shown figures for 81 Watt RMS output: 36Vp swing.
Anything less is only better.
To adjust for Class A output, is interesting. I have not yet done any Class A testing.
But my prediction is fantastic figures for Class A.
Now, the latest version I show can be made to acommodate for your needs. The difference is not that big.
I have shown figures for 81 Watt RMS output: 36Vp swing.
Anything less is only better.
To adjust for Class A output, is interesting. I have not yet done any Class A testing.
But my prediction is fantastic figures for Class A.
Andrew.
This is a version with +/- 20 Volts rails.
It runs in Class A with 1.245 A at idle.
The difference from my high power Class AB is minute.
I have not specially adjusted it for Class A operation.
Nevertheless it shows very good performance.
Here is the circuit and the THD figures at different levels - all in Class A.
This is a version with +/- 20 Volts rails.
It runs in Class A with 1.245 A at idle.
The difference from my high power Class AB is minute.
I have not specially adjusted it for Class A operation.
Nevertheless it shows very good performance.
Here is the circuit and the THD figures at different levels - all in Class A.
Attachments
No, I have not tested for capacitance in the load.
As I remember 2.2uF in parallelled to the load will make an ultimate test.
As far as I remember.
Would you help me to setup such a test?
I thank you for this, it is a very important test.
I know about the Akido. It is a tube circuit.
Tell me more, please.
I thank you for your post.
To be capacitive stable is an important thing.
I have done some changes to my circuit.
It is now stable into 8 Ohm//1uF.
This was done changing the compensation caps, increasing them to make the amplifier slower.
Of course we can make the amplifier stable into 2.2uF,
but I think 1 uF is enough.
Here is the modified circuit.
Those caps are in red circles.
Attachments
There has been a change.
It has to do with the remark by a guy here, that we should make it stable in capacitive loads.
This is essential, because it can avoid oscillation and burnt amplifier.
So, I have done a different compensation.
This makes the amplifier stable up to something like 2.2uF across 8 Ohm.
This is so far my testing went. And it behaves well.
The price for making stable using capacitors is somewhat more THD distortion.
It is still an extremely fast circuit .. -3dB well above 10 MHz
I show you here the new version. Stable in capacitive loads.
To change the power supply and increase bias to Class A is not too difficult for you, Andrew. It is purely math.
Attachments
Is it possible to have U3, U9, U6 and U7 handle your bias. May have come across a unity gain self biasing output stage ....Also the current needed to drive this stage is quite high.....you can make a dynamic class A output stage.........the question remains concerning the sound character of the entire amplifier from input to output....your on to something if you get it right, still got some ways to go
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.