Hi all,
well it's weekend and this is what I've come up with is my lost hours.... I've been puzzeling around with the hiraga style amps for some time now.. I think I've got rid off the buffer stage with a new biasing aragement...
please let me you what you think..
happy holiday greetings,
Thijs
PS sure it simulates great but let's not talk about that 😉
well it's weekend and this is what I've come up with is my lost hours.... I've been puzzeling around with the hiraga style amps for some time now.. I think I've got rid off the buffer stage with a new biasing aragement...
please let me you what you think..
happy holiday greetings,
Thijs
PS sure it simulates great but let's not talk about that 😉
Attachments
Come on... anyone ... doesn't have to be intelligent.. just make some comment on the design please 🙂
gr,
T
gr,
T
Thijs,
Good design, but needs some temperature compensation, depending on the mosfets technology... this is the major difficulty for this arrangement.
Regards, Pierre Lacombe.
Good design, but needs some temperature compensation, depending on the mosfets technology... this is the major difficulty for this arrangement.
Regards, Pierre Lacombe.
thanks guys....
About the temperature compensation.. good thinking! It can be a problem since it is desinged for about 1.2 amp Iq from 15 Volt rail, so everything will get hot.. I've tried to put 0.1 to 1 ohm resistors ion series with the output devices, that might help a little.. I guess I've got to do the real experiment to really find out..
greetings,
Thijs
About the temperature compensation.. good thinking! It can be a problem since it is desinged for about 1.2 amp Iq from 15 Volt rail, so everything will get hot.. I've tried to put 0.1 to 1 ohm resistors ion series with the output devices, that might help a little.. I guess I've got to do the real experiment to really find out..
greetings,
Thijs
I like this if you use Lateral Mosfets as you would need no temp compensation. I posted a similar schematic and got beaten up by using the inverted outputs. According to them the output impedance is too high with this arrangement. I still want to try it sometime.
Somehow my spiceprogram doesn't export component values.. here they come:
Input R-network : 1K series, 10K to ground
Load R of input BJT: 10K
Series R of output stage 0.1 to 1 Ohm
Feedback net work: 100 Ohm from output, 12 Ohm to ground
Biasing networks: 220 from voltage rails, 10 Ohm to feedback node, little unsure about this last one...
PS.
output impedance is simulates aroung 100mOhm..
gr,
Thijs
Input R-network : 1K series, 10K to ground
Load R of input BJT: 10K
Series R of output stage 0.1 to 1 Ohm
Feedback net work: 100 Ohm from output, 12 Ohm to ground
Biasing networks: 220 from voltage rails, 10 Ohm to feedback node, little unsure about this last one...
PS.
output impedance is simulates aroung 100mOhm..
gr,
Thijs
Thanks, I'm tempted to build one to try it out aftr I finish my Alephs. The other complaint I got was that the output devices supply gain and therefore vary with different loads. I'm not convinced. Here is my thread if you missed it. http://www.diyaudio.com/forums/showthread.php?threadid=4282&highlight=sziklai
Thanks! Great thread!
If I knew I would posted this design as a reply.
But I still don't agree on much of the output impedance discussion.
gr,
Thijs
If I knew I would posted this design as a reply.
But I still don't agree on much of the output impedance discussion.
gr,
Thijs
I did some simulations on an inverting two stage amplifier with this FET output topology.
I didn't bias them to class A but higher than usual. The (simulated) performance was quite good. The only disadvantage was indeed the output impedance that was a little high.
When the gates of the MOSFETs were driven through an emitter follower, the upper rolloff frequency was around 0.5 MHz (simulated !).
Regards
Charles
P.S.:http://www.gruensch.de/Produkte/GRUENSCH_MSE/gruensch_mse.html is a commercial high-biased two-stage amp although with a differential input stage.
I didn't bias them to class A but higher than usual. The (simulated) performance was quite good. The only disadvantage was indeed the output impedance that was a little high.
When the gates of the MOSFETs were driven through an emitter follower, the upper rolloff frequency was around 0.5 MHz (simulated !).
Regards
Charles
P.S.:http://www.gruensch.de/Produkte/GRUENSCH_MSE/gruensch_mse.html is a commercial high-biased two-stage amp although with a differential input stage.
Thanks for the values. We need 221 ohm gate stoppers for the FET's and possibly an RC network of 10 ohm and .01 for the output to prevent oscillation. Now let's bridge two of these and feed it with a balanced input for lower distortion, more power, and to eliminate the output offset. What do you think?
Hi Dshortt9,
gate stoppers: that's probably a good idea.
RC network at output: that's might not be necesary. The amp simulated very stable but with a real speaker in a real world this is probably also a good idea.
Brigdging: this will only lower the asymmetric distortion (even harmonics?), but won't lower the symmetric distortion (odd harmonics ?). The Iq bias need to be double for class A operation in brigde mode. Class A operation is manditory due to the relative low amount of feedback avialeble to remove crossover distortion.
More power: off +/- 15 Volts this little amp will give about 8 Watts in 8 ohm, that maybe 25- to 32 Watts brigde mode.. I know I only need about 10 Watts. I once measured my total system gain, most of the time it is about one (that's pre and power amp cascaded!). That means about 2 volt rms into 6 ohms in my case.
I think it would be wise to first make a low power version to test the concept. I think that the adjustment for Iq and Vq out need to be done by varying the 220 ohm biasing resistors of the input BTJ pair. Let's first find out if that will work.
Thanks for your enthousiasm! Keep up the good thinking.
greetings,
Thijs
PS
Allthough the circuit topology is now where near innovative, I've never seen this kind of combined biasing/feedback topology. This allows for a very simple, low component count, DC coupled, high bandwidth, low distortion, high power, capacitor-less, amplifier with gain. Why isn't this going to be a classic like JLH's 10watts, hiraga's monstre etc ? Hmmmmmm just dreaming...
gate stoppers: that's probably a good idea.
RC network at output: that's might not be necesary. The amp simulated very stable but with a real speaker in a real world this is probably also a good idea.
Brigdging: this will only lower the asymmetric distortion (even harmonics?), but won't lower the symmetric distortion (odd harmonics ?). The Iq bias need to be double for class A operation in brigde mode. Class A operation is manditory due to the relative low amount of feedback avialeble to remove crossover distortion.
More power: off +/- 15 Volts this little amp will give about 8 Watts in 8 ohm, that maybe 25- to 32 Watts brigde mode.. I know I only need about 10 Watts. I once measured my total system gain, most of the time it is about one (that's pre and power amp cascaded!). That means about 2 volt rms into 6 ohms in my case.
I think it would be wise to first make a low power version to test the concept. I think that the adjustment for Iq and Vq out need to be done by varying the 220 ohm biasing resistors of the input BTJ pair. Let's first find out if that will work.
Thanks for your enthousiasm! Keep up the good thinking.
greetings,
Thijs
PS
Allthough the circuit topology is now where near innovative, I've never seen this kind of combined biasing/feedback topology. This allows for a very simple, low component count, DC coupled, high bandwidth, low distortion, high power, capacitor-less, amplifier with gain. Why isn't this going to be a classic like JLH's 10watts, hiraga's monstre etc ? Hmmmmmm just dreaming...
How about a cascode for the input BJT pair?
How about ..... ?
Or even ........... ?
And why not ............ ?
In plain text: To me, the most tempting feature of your circuit is it's KISS character, so why spoil it ?
Regards
Charles
P.S. I tried a variant of an input cascode as well for simulations with a two-stage amp. Mine was inverting though. If you like I can dig out the file.
How about ..... ?
Or even ........... ?
And why not ............ ?
In plain text: To me, the most tempting feature of your circuit is it's KISS character, so why spoil it ?
Regards
Charles
P.S. I tried a variant of an input cascode as well for simulations with a two-stage amp. Mine was inverting though. If you like I can dig out the file.
I always look for input pair with higher voltage capability.tschrama said:
I never use any cascode.
This slows down the amplifier,
as current has to pass another pair of transistors.
I find seldome any advantages
by using cascode - I can manage without.
KISS= Keep It Simple (if possible)
that is my philosophy - copied by many other designers 😀
/halo - totally uncascoded
Hi Halojoy,
Thanks for your reply. I still do like the simple circuits the best. Mainly because they are more easy to understand and to build...but...
Cascodes and CCS are quite simple ingredients and can improve a design significantly I think. In this design the voltage swing on the collectors (or actually Vce)of the input BJT is low so a cascode wont help much in terms of lineairity or speed.
There still might be a problem with slewrate (MOSFET capacitance driven by 10K ?) allthough the simulations show >500KHz bandwidth. Still a CCS as collectorload for the input BJT would not only improve slewrate, but also openloop gain (and thus more feedback availeble which will propably improve all the spec's). Will it be impossible to adjust Iq and Vq and stabilize than?
I don't understand your comments on the Cascode (slowong a circuit down). I always understood that a Cascode would speed up a circuit since:
There is non Vce swing, no BJT capacitance to charge which otherwise slows it down. The upper cascode transistor is 'collector driven' which is a high speed configuration.. so all in all a cascode speeds things up, shouldn't it?
gr,
Thijs
Thanks for your reply. I still do like the simple circuits the best. Mainly because they are more easy to understand and to build...but...
Cascodes and CCS are quite simple ingredients and can improve a design significantly I think. In this design the voltage swing on the collectors (or actually Vce)of the input BJT is low so a cascode wont help much in terms of lineairity or speed.
There still might be a problem with slewrate (MOSFET capacitance driven by 10K ?) allthough the simulations show >500KHz bandwidth. Still a CCS as collectorload for the input BJT would not only improve slewrate, but also openloop gain (and thus more feedback availeble which will propably improve all the spec's). Will it be impossible to adjust Iq and Vq and stabilize than?
I don't understand your comments on the Cascode (slowong a circuit down). I always understood that a Cascode would speed up a circuit since:
There is non Vce swing, no BJT capacitance to charge which otherwise slows it down. The upper cascode transistor is 'collector driven' which is a high speed configuration.. so all in all a cascode speeds things up, shouldn't it?
gr,
Thijs
Just an idea:
1] how about using a CCS for the collectorloads of the input BJT and
2] replacing the output MOSFET with CFP's
This would essentially be a BTJ input version of a hiriga monstre amp, with CCS and no cascode...
would it wotk?
1] how about using a CCS for the collectorloads of the input BJT and
2] replacing the output MOSFET with CFP's
This would essentially be a BTJ input version of a hiriga monstre amp, with CCS and no cascode...
would it wotk?
Fact is that the cascoding the input pair
do make somethings worse.
say input voltage is +-1 Volt
and Vc-e over input transistor is 20 volts cascoded
40 volts uncascoded
the Vce will fluctuate +-1 Volt/20 and +-1volt/40
And you need somethings to set working voltage for
the cascode transistor, and some capacitor to filter
WEll, I can do without those extras,
does not give me just about anything.
Note, that this is about input pair.
Cascoding can sometimes give a more linear behavior
when used in other places of an amplifier.
But as said, I do not use it.
I think cascoding is heavily overrated,
and the negative aspects of it are seldome pronounced.
/halojoy - still happily uncascoded
do make somethings worse.
say input voltage is +-1 Volt
and Vc-e over input transistor is 20 volts cascoded
40 volts uncascoded
the Vce will fluctuate +-1 Volt/20 and +-1volt/40
And you need somethings to set working voltage for
the cascode transistor, and some capacitor to filter
WEll, I can do without those extras,
does not give me just about anything.
Note, that this is about input pair.
Cascoding can sometimes give a more linear behavior
when used in other places of an amplifier.
But as said, I do not use it.
I think cascoding is heavily overrated,
and the negative aspects of it are seldome pronounced.
/halojoy - still happily uncascoded
A significant property of a cascode circuit is the tremendous reduction of the miller effect and therefore extended bandwidth compared to a common emitter circuit. In this respect I have to disagree with Halojoy (though there are some risks involved for disagreeing with him 😉 )
The input circuit of the above amp is not a simple common emitter circuit however.
For the forward path it is a common emitter circuit and for the NFB loop it is a common base circuit with a poorly grounded base (maybe there is a more elegant term for that ).
Using a current source as load for the input devices could indeed be tricky (in fact two current sources in series feeding a voltage into an infinite load resistance). With BJT output devices it would be easier.
Maybe there would be a solution by making this CCS dependant on the output device's source current (this would result in an additional inner feedback loop for each half of the circuit). But then you sacrifice KISS again.
Regards
Charles
The input circuit of the above amp is not a simple common emitter circuit however.
For the forward path it is a common emitter circuit and for the NFB loop it is a common base circuit with a poorly grounded base (maybe there is a more elegant term for that ).
Using a current source as load for the input devices could indeed be tricky (in fact two current sources in series feeding a voltage into an infinite load resistance). With BJT output devices it would be easier.
Maybe there would be a solution by making this CCS dependant on the output device's source current (this would result in an additional inner feedback loop for each half of the circuit). But then you sacrifice KISS again.
Regards
Charles
- Status
- Not open for further replies.