Member
Joined 2009
Paid Member
I designed it for use with +/-50V rails which will give up towards 150W into 8R but I haven't verified this because in the end it depends how close the output will swing to the rails. To be honest, it's way more power than is necessary for domestic listening on reasonable speakers so I've never really bothered about this. I am in the process of putting my own boards into a nice box and I will use +/-42V rails. This will give me up towards 100W into 8R which is more than I will need for sure but I happen to have a nice transformer I want to use if it will fit into the box. What I did focus on was designing the output stage so that it would produce lovely sound at low levels from a single pair of BJT power transistors and yet have a pair of power MOSFETs as guardian angels to deal with peaks into lower impedances when called up on.
The LED indicates green when the solid state relay at the output is closed and the output is connected to the speaker, it indicates red when the relay is open and the output is disconnected from the speaker. I messed up first time around so that it only works on one of the colours. One fix is to mount the LED off-board with a series resistor back to the -ve power rail, i.e. mount the LED on the front panel where it can act as the power-on indicator and also alert you if there is a dc-fault condition (the other fix requires some work on the pcb but to be honest I think it's nicer to have the LED on the front panel) - you might want to read through the thread.
The LED indicates green when the solid state relay at the output is closed and the output is connected to the speaker, it indicates red when the relay is open and the output is disconnected from the speaker. I messed up first time around so that it only works on one of the colours. One fix is to mount the LED off-board with a series resistor back to the -ve power rail, i.e. mount the LED on the front panel where it can act as the power-on indicator and also alert you if there is a dc-fault condition (the other fix requires some work on the pcb but to be honest I think it's nicer to have the LED on the front panel) - you might want to read through the thread.
Last edited:
Member
Joined 2009
Paid Member
If you are new to this hobby you'll be rushing around reading stuff from many sources, mostly on the internet where people can say what they like, soaking it all into your brain like a sponge and trying to make sense of it all !
My experience would suggest to you that you keep it simple, build something that you can enjoy and then go from there, otherwise you may be forever second guessing yourself whilst you try to optimize the whole system from the start before you have gained very much experience. I don't think you can do it this way, you have to get your hands dirty as they say. It can take more effort than it looks to put a system together, or even build a DIY amplifier with all the wiring and parts that have to be installed in a box. And after all the trouble you may find you aren't spending much time listening to the music !
My experience would suggest to you that you keep it simple, build something that you can enjoy and then go from there, otherwise you may be forever second guessing yourself whilst you try to optimize the whole system from the start before you have gained very much experience. I don't think you can do it this way, you have to get your hands dirty as they say. It can take more effort than it looks to put a system together, or even build a DIY amplifier with all the wiring and parts that have to be installed in a box. And after all the trouble you may find you aren't spending much time listening to the music !
Hi Bigun
I am fascinated by your work on the single transistor input stage.
Recently finished working with JG and FdW on my little Nobrainer amp that has a very clever biasing scheme and a very stable CFP output stage... it sounds really good but lacks some power due to it's opamp input stage.... So I took the liberty to clone your input stage and got this scheme that seems to work.
As I am not a specialist in spice, maybe you can have a look and tell me if it can be done.
Even if you do not have the time for this, maybe I can have your permission to continue using your design in my case study
best
Ricardo
I am fascinated by your work on the single transistor input stage.
Recently finished working with JG and FdW on my little Nobrainer amp that has a very clever biasing scheme and a very stable CFP output stage... it sounds really good but lacks some power due to it's opamp input stage.... So I took the liberty to clone your input stage and got this scheme that seems to work.
As I am not a specialist in spice, maybe you can have a look and tell me if it can be done.
Even if you do not have the time for this, maybe I can have your permission to continue using your design in my case study
best
Ricardo
Rereading my last post I see there is room for some clarification.
The Nobrainer is a power amp designed by Joachim Gerhard that I built as is. After my initial less then good results I was able to get FdW interest that introduced some important mods in the floating mirror and than other mods in the bias scheme.
So this is not my design, I built it, asked for some mods that I implemented and finally had the work to finetune it to an outstanding sounding little amp.
The results are so good that I want to know if it is possible to use with a discrete input stage instead of the opamp.
The Nobrainer is a power amp designed by Joachim Gerhard that I built as is. After my initial less then good results I was able to get FdW interest that introduced some important mods in the floating mirror and than other mods in the bias scheme.
So this is not my design, I built it, asked for some mods that I implemented and finally had the work to finetune it to an outstanding sounding little amp.
The results are so good that I want to know if it is possible to use with a discrete input stage instead of the opamp.
Member
Joined 2009
Paid Member
There is a floating mirror fed by a central very stiff ccs (thanks to Frans)...
The ccs current is then mirrored to the biasing transdiodes ( the bc with one leg shorted) in series with a trimpot that sets idle.
The signal is fed to the drivers via two high value resistors because we found it would greatly reduce distortion.
The output buffer is a very stable CFP.
All this I understand very well.... what I am having trouble with is your complex biasing schematic on the input stage.... maybe you could give me some guidelines so I can start developing it a bit.
The ccs current is then mirrored to the biasing transdiodes ( the bc with one leg shorted) in series with a trimpot that sets idle.
The signal is fed to the drivers via two high value resistors because we found it would greatly reduce distortion.
The output buffer is a very stable CFP.
All this I understand very well.... what I am having trouble with is your complex biasing schematic on the input stage.... maybe you could give me some guidelines so I can start developing it a bit.
Member
Joined 2009
Paid Member
There is the full implementation of the input stage as shown in post 245, and the simpler input stage as shown in post 266. The full implementation is more linear, provides for higher OLG. Both use the same biassing scheme.
With a Singleton input you always have to contend with the fact that the input to the base of the input transistor is dc-shifted from the feedback node at the emitter - the difference being one 'Vbe' of around 650mV. So you have to shift the dc-level of the input from ground by 650mV, or you have to shift the dc-level from the feedback from ground by 650mV - or you accept a dc-offset at the output and use a capacitor in the output (which isn't the case for TGM8).
Many people choose to dc-shift the input to the base. I chose to shift the dc of the feedback node and leave the input at the base close to ground. I did this by feeding a dc-voltage to top of the feedback shunt capacitor. In post 245 the schematic identifies C3 as the feedback shunt capacitor. The zener diode D1 creates a reference voltage of 12V. Q2 and R6 form a potential divider between this 12V reference voltage and ground. R12 picks off the mid point of this divider and feeds the voltage to C3. By adjusting the voltage drop across Q2 using VR1 we adjust the dc-voltage fed to the top of C3. Because of Q2 acting like a forward-biassed diode it will track temperature changes and ensure the dc-offset is kept stable with temperature.
Last edited:
Member
Joined 2009
Paid Member
This is phase lead compensation around the feedback series resistor. It is not unique to TGM8 and you may see it on other designs.
Hi Bigun
Just circling back on this project. I recall your original design brief called for an DC coupled input and that provisions for the input cap were only added after a little peer pressure from myself and others ;-)
We discovered that there is a DC offset at the input transistor base with or without coupling and so the input cap became a necessity.
1. Has anything about the design changed that could eliminate the cap?
2. If the cap is a necessity, then is there any benefit to shifting the DC of the feedback node as opposed to the conventional approach to shift it at the input base?
I guess stating the question in another way: can the input cap be eliminated and if not then what is the benefit of setting the offset at the feedback node?
Just circling back on this project. I recall your original design brief called for an DC coupled input and that provisions for the input cap were only added after a little peer pressure from myself and others ;-)
We discovered that there is a DC offset at the input transistor base with or without coupling and so the input cap became a necessity.
1. Has anything about the design changed that could eliminate the cap?
2. If the cap is a necessity, then is there any benefit to shifting the DC of the feedback node as opposed to the conventional approach to shift it at the input base?
I guess stating the question in another way: can the input cap be eliminated and if not then what is the benefit of setting the offset at the feedback node?
Member
Joined 2009
Paid Member
Excellent questions.
1. Well yes, peer pressure won out on including provision for an input cap on the pcb. Strictly speaking, it shouldn't be needed because the input should be happy being at d.c. ground potential (not a.c. of course or the signal is gone!). The issue is that the base of the input device does require current, bipolar transistors are voltage controlled but do not have infinite input impedance. This base current develops a voltage drop across any resistance at the input of the amplifier. You can null out this voltage drop with the dc-offset adjustment using VR1, but if you change the source impedance you potentially have to re-adjust VR1. So the cap could be said not to be a necessity but it's convenient to keep it. If we were to replace the input transistor with a nice JFET of suitable voltage rating(?) the gate current is so much smaller that it's negligible.
2. I still like setting the dc offset at the feedback node because it is low impedance point and so tends to be less sensitive to noise. Where I've injected the offset adjust the feedback cap, C3, shunts any noise to ground directly.
1. Well yes, peer pressure won out on including provision for an input cap on the pcb. Strictly speaking, it shouldn't be needed because the input should be happy being at d.c. ground potential (not a.c. of course or the signal is gone!). The issue is that the base of the input device does require current, bipolar transistors are voltage controlled but do not have infinite input impedance. This base current develops a voltage drop across any resistance at the input of the amplifier. You can null out this voltage drop with the dc-offset adjustment using VR1, but if you change the source impedance you potentially have to re-adjust VR1. So the cap could be said not to be a necessity but it's convenient to keep it. If we were to replace the input transistor with a nice JFET of suitable voltage rating(?) the gate current is so much smaller that it's negligible.
2. I still like setting the dc offset at the feedback node because it is low impedance point and so tends to be less sensitive to noise. Where I've injected the offset adjust the feedback cap, C3, shunts any noise to ground directly.
Last edited:
So there is a benefit
Are there any JFETs with nice specs that are still readily available? I've been trying to source some 2SK170 devices to complete a pair of Rudy's SYMASYM boards but these parts have reached "unobtainium" status with the high prices that ensue.
I might play with the idea of a singleton JFET fed input in my Quasi complimentary project thread.
Are there any JFETs with nice specs that are still readily available? I've been trying to source some 2SK170 devices to complete a pair of Rudy's SYMASYM boards but these parts have reached "unobtainium" status with the high prices that ensue.
I might play with the idea of a singleton JFET fed input in my Quasi complimentary project thread.
Member
Joined 2009
Paid Member
I would ask over at the PASS forum where JFETs are quite popular - I believe there are options available. There are plenty available from Digikey but maybe mostly in SMT packages (which is OK if using the larger SOT-23). The problem with JFETs is that they usually can't take the voltages we want to use in our powerful Class AB amplifiers without cascading. If you do go that direction let me know, I might like to get a few myself....
Member
Joined 2009
Paid Member
Bigun, I took the question to the Pass forum as you suggested and the 2sk209 appears to be the pick of the bunch. It is a TO-236 part so very small but I'm up to a soldering challenge...
http://docs-asia.electrocomponents.com/webdocs/10c3/0900766b810c3a4f.pdf
Am I correct in thinking that Vgds needs to be rail voltage plus half input voltage and some margin, say 30%. So this part would tolerate 25V rails or 35V at a pinch, sufficient for a low power design?
Do you see any real benefit in adapting your TGM8 circuit with a JFET input, and possibly MOSFET only outputs, either quasi or complimentary?
http://docs-asia.electrocomponents.com/webdocs/10c3/0900766b810c3a4f.pdf
Am I correct in thinking that Vgds needs to be rail voltage plus half input voltage and some margin, say 30%. So this part would tolerate 25V rails or 35V at a pinch, sufficient for a low power design?
Do you see any real benefit in adapting your TGM8 circuit with a JFET input, and possibly MOSFET only outputs, either quasi or complimentary?
Member
Joined 2009
Paid Member
Yes, I think you have a proper understanding of the voltage rating. The interesting thing that I had forgotten to consider about JFETs is that they are depletion mode devices, you have to bias them negatively to turn them off. This means that the dc-offset is the opposite polarity compared with a bipolar transistor - so it won't be a drop-in to the TGM8 without a couple of mods.
A MOSFET based amp would be an interesting project, rather different design constraints than a bipolar based amp and so an opportunity to learn some new things. I don't know if I'd start with TGM8, you might find it useful to look around at some existing MOSFET designs to see how they have been put together and what design considerations went into them. Carlos has an interesting design in the works here: http://www.diyaudio.com/forums/solid-state/255673-hiraga-amplifier-mosfet-insulation.html (post 12) which might be adapted to a JFET input.
A MOSFET based amp would be an interesting project, rather different design constraints than a bipolar based amp and so an opportunity to learn some new things. I don't know if I'd start with TGM8, you might find it useful to look around at some existing MOSFET designs to see how they have been put together and what design considerations went into them. Carlos has an interesting design in the works here: http://www.diyaudio.com/forums/solid-state/255673-hiraga-amplifier-mosfet-insulation.html (post 12) which might be adapted to a JFET input.