I was looking at geoff's class-a site and noticed this amp. It uses jfet upfront, and is fully complementory.
I would like to know:
1) how does it work? it has a fully complementory vbe generator,
. and it also has a very interesting feedback scheme. I would appreciate any insight into how it works.
2) I am interested in what has to be done to replace the jfets;
3) and I would also be interested to use some irfs as output devices,.
Any insight is apprciated.
I would like to know:
1) how does it work? it has a fully complementory vbe generator,
. and it also has a very interesting feedback scheme. I would appreciate any insight into how it works.2) I am interested in what has to be done to replace the jfets;
3) and I would also be interested to use some irfs as output devices,
.Any insight is apprciated.
john, here it is:
http://www.tcaas.btinternet.co.uk/jc-3.gif
your comments would be greatly appreciated.
http://www.tcaas.btinternet.co.uk/jc-3.gif
your comments would be greatly appreciated.
That design is correct in principle, but the design is over 25 years old, and the parts are obsolete. The basic design can be modified with FET's just about everywhere. I recommend Fairchild Mosfets, as they will be better than IR in this application. Jfets can be 2sk389 and j109.
This design is somewhat limited to inverting operation, but can be more useful with a servo added to control the offset.
This design is somewhat limited to inverting operation, but can be more useful with a servo added to control the offset.
This is a dual feedback arrangement. Especially in the early days, it seemed to be useful to reduce the amount of global negative feedback, because of potential TIM. This technique lowers the over global negative feedback to the ratio of the internal feedback resistors and the overall feedback resistor. This should be about 14-20dB. The reason for this is that the drive impedance to the output stage is lowered and this removes the dependence of very linear beta in the output stage. Another way of looking at this is that the pre-driver stage (6178-6180) generates a very high drive impedance. The darlington output stage reflects back from the speaker impedance another fairly high impedance, which is essentially 2*B(1)*B(2)*R(L). It is the relative levels of these two impedances that implies whether the output stage is voltage or current controlled. Lowering the drive impedance to the output stage makes the output stage into a voltage follower, rather than a beta multiplier (which is more nonlinear).
With fets in the circuit, this is unnecessary, as the gate is never current controlled.
With fets in the circuit, this is unnecessary, as the gate is never current controlled.
Upon re-reading my previous comments, I regret that I did not express things properly.
This is the situation: A transistor output stage can be thought to be controlled in two ways. It can either be a BETA MULTIPLIER or a VOLTAGE FOLLOWER. What matters is the source impedance that the output stage is driven with. If it is driven with a HIGH impedance, then the output stage will behave as a Beta Multiplier and will therefore be VERY dependent on BETA.
If, on the other hand the drive to the output stage is LOW impedance, then the transistor output stage will behave as a Voltage Follower to the drive voltage. Beta nonlinearity will then be less important.
The extra feedback loop reduces the drive impedance to the transistor output stage, so it should become more linear, and also be less sensitive to load variation on overall linearity.
I left out one important word in my previous statement:
It should have said "removes the dependence of HAVING a very linear beta in the output stage"
I hope you can understand what the extra feedback does better now.
This is the situation: A transistor output stage can be thought to be controlled in two ways. It can either be a BETA MULTIPLIER or a VOLTAGE FOLLOWER. What matters is the source impedance that the output stage is driven with. If it is driven with a HIGH impedance, then the output stage will behave as a Beta Multiplier and will therefore be VERY dependent on BETA.
If, on the other hand the drive to the output stage is LOW impedance, then the transistor output stage will behave as a Voltage Follower to the drive voltage. Beta nonlinearity will then be less important.
The extra feedback loop reduces the drive impedance to the transistor output stage, so it should become more linear, and also be less sensitive to load variation on overall linearity.
I left out one important word in my previous statement:
It should have said "removes the dependence of HAVING a very linear beta in the output stage"
I hope you can understand what the extra feedback does better now.
http://www.ne.jp/asahi/evo/amp/J200K1529/report.htm
circuits 6,7 and 13,14 seem to show some relation to the JC-3 topology and could be interpreted as illustrating Mr Curl's point about output stage impedance loading of the VAS stage - of course i'm "interpreting" the schematics, not the Japanese commentary
circuits 6,7 and 13,14 seem to show some relation to the JC-3 topology and could be interpreted as illustrating Mr Curl's point about output stage impedance loading of the VAS stage - of course i'm "interpreting" the schematics, not the Japanese commentary
Mr.John,
Could you tell me why do you not use Non-inverted mode in JC-3?
The reason is easy to make amp stable?
I think that adding a resister (10K,f.e.) between the pre-driver stage and GND could obtain the same effect which lower the impedance of the pre-driver stage output,and you didn't do that.could you tell me why? Maybe the impedance of the pre-driver stage is still higher and the open loop gain of amp is lower ?
Sorry for my poor English,
thanks for your replies.
X.G.
Could you tell me why do you not use Non-inverted mode in JC-3?
The reason is easy to make amp stable?
I think that adding a resister (10K,f.e.) between the pre-driver stage and GND could obtain the same effect which lower the impedance of the pre-driver stage output,and you didn't do that.could you tell me why? Maybe the impedance of the pre-driver stage is still higher and the open loop gain of amp is lower ?
Sorry for my poor English,
thanks for your replies.
X.G.
john curl said:
thx ,Mr. Curl.
In my mind, the differental input stage needs/is designed to reject the common signal.In order to increase the CMRR , higher the output impedence of CC at the tail or change the resister of the tail to the CC.I had not heared 'the common mode distortion',but I knew the input stage is set in CommonSource-CommonGate stage expanding the freq-bandwide at the inverted mode .Maybe your meaning is the CMRR(but the CMRR of the inverted mode is same as the Non-inverted,IMO)? Could you explain or give me a link or info about 'the common mode distortion'?
BTW:I found JC-3 have the special circuit that used dual feedback loops.according to Mr. John Curl,the first feedback loop produced the effect of lowering the impedence of the second amplifier stage
(=pre-driver).The load of the second stage is a resister of several kilo-ohm not a CC in Some modern amp circuit.I think these amps have the effect like JC-3.How do you think,MR.Curl?
PS:I had studied some electronic knowlege in college for 4 years,my working is not relative to the electronic.I love DIY of audio.
Sorry for my poor English again.
THX a lot.
X.G.
Common mode distortion is caused by a common mode input signal appearing at the amp inputs. This is both + and - inputs. When you have a non-inverting amp with low voltage gain, the input signal and the feedback ride up and down in common mode, with the drive signal. If you do NOT cascode, then you get more distortion. This can be also seen with analog IC's, at least early devices that were not cascoded on the input.
The inverting input FORCES the - input to 0 volts at all times (for all practical purposes. The + input is already at 0 volts, so no common mode signal is produced and therefore no common mode distortion.
The JC-3 is an OLD design about 30 years old. It will still work, but it is not a good example of modern designs, except for the internal topology.
The inverting input FORCES the - input to 0 volts at all times (for all practical purposes. The + input is already at 0 volts, so no common mode signal is produced and therefore no common mode distortion.
The JC-3 is an OLD design about 30 years old. It will still work, but it is not a good example of modern designs, except for the internal topology.
I have looked at the JC-3 for a while and I think it is a fascinating amplifier, so simple yet so sophisticated! I think it would be a great platform for experimenting with newer devices. Since the creator of this amp is here on the forum, I would like to ask him a few questions:
1. Since there are no dominant pole capacitors in this circuit, I assume it has a large open-loop bandwidth. Do you remember how large? And is it stable with G=1?
2. When substituting transistors for newer, faster ones the bandwidth and gain will likely/hopefully increase and this might create stability problems. Do you recommend a small cap across the VAS collector-base or do you prefer another method?
Thanks!
1. Since there are no dominant pole capacitors in this circuit, I assume it has a large open-loop bandwidth. Do you remember how large? And is it stable with G=1?
2. When substituting transistors for newer, faster ones the bandwidth and gain will likely/hopefully increase and this might create stability problems. Do you recommend a small cap across the VAS collector-base or do you prefer another method?
Thanks!
OJG, this JC-3 design is about 30 years old. It is essentially a JC-2 line driver, coupled with a darlington output stage. You are correct that the dominant pole is created by the miller capacitance of the driver stage transistors. This design, as it stands, can do about 100V/us and should have an open loop bandwidth of about 20KHz (as I remember). This was to fit the critereon set by Matti Otala at the time.
Of course, newer, faster output transistors would make this amp faster, the initial parts had an F(t) of 4 MHz. We can get 10 times this, today.
The RCA devices initially used were actually better at the time than most equivalently rated transistors, because they had less distortion due to their internal construction, but I doubt that they are available today.
This design topology is what is used in all the Parasound amp designs that I am associated with.
Because of faster output transistors, we have eliminated the output coil and still get 100+ V/us slew rate.
This design, with improved output transistors, would work with unity gain, but it seems to be overdesign for just a buffer.
Of course, newer, faster output transistors would make this amp faster, the initial parts had an F(t) of 4 MHz. We can get 10 times this, today.
The RCA devices initially used were actually better at the time than most equivalently rated transistors, because they had less distortion due to their internal construction, but I doubt that they are available today.
This design topology is what is used in all the Parasound amp designs that I am associated with.
Because of faster output transistors, we have eliminated the output coil and still get 100+ V/us slew rate.
This design, with improved output transistors, would work with unity gain, but it seems to be overdesign for just a buffer.
john curl said:
Thanks, John!
That is what I thought. I like it better already!
Is this a hint as to what is inside the "critically acclaimed" Halo JC-1?
If I substituted the VAS and driver with mosfets...?My DAC outputs 10Vp so if I keep the 20V rails I would configure it for a gain of 2. I like low-gain power-amps to maximise SNR.
I think it is time for me to play with Spice a little and I am sure I will have more questions then!
Yes, the JC-1 is just bigger, and more complex. It has to swing +/- 90V rails. I cascode the input stage with fets, the driver stage is composed of 200V fets, and the 'darlington' driver is now fets as well. All that is left are the bipolar output transistors. By the way, be sure to add 10 ohms to the base of any output transistor used. We found that with ring emitter devices (high speed) that this is important. The 10 ohms adds a real resistance that cancels any negative resistance generated by the combined phase shift of the base emitter junction and any capacitive load.
Again, the reason for the inverting input was to reduce COMMON MODE distortion to a minimum. It is not necessary IF you CASCODE the input stage. All modern designs are non-inverting or balanced input.
Again, the reason for the inverting input was to reduce COMMON MODE distortion to a minimum. It is not necessary IF you CASCODE the input stage. All modern designs are non-inverting or balanced input.
thanks, John.
Has anyone tried to adapt the design for bipolar or MOSFETs? I simply don't feel comfortable with jfets and wanted to know if bipolar would have worked.
In the one version that I tried (in simulation), it started out well but then began to oscillate. haven't had time to figure out why.
John: on the MOSFETs, you had mentioned somewhere that you tried some Fairfield MOSFETs with good results. do you have a preferred list of mosfets?
Thanks in advance.
Has anyone tried to adapt the design for bipolar or MOSFETs? I simply don't feel comfortable with jfets and wanted to know if bipolar would have worked.
In the one version that I tried (in simulation), it started out well but then began to oscillate. haven't had time to figure out why.
John: on the MOSFETs, you had mentioned somewhere that you tried some Fairfield MOSFETs with good results. do you have a preferred list of mosfets?
Thanks in advance.