What are your opinions on the best topology for a line driver follower? I'm designing the output of a CD player and want to do it with no feedback in the system.
Goals are for high linearity, low noise and relatively low output impedance. The power supply can be designed around the right follower.
There are options aplenty, NPN or PNP emitter followers, PMOS or NMOS, P/NJFETs, tubes etc. I'd like to learn about successfull output follower stages at line level.
Thanks for your contributions.
Børge
Goals are for high linearity, low noise and relatively low output impedance. The power supply can be designed around the right follower.
There are options aplenty, NPN or PNP emitter followers, PMOS or NMOS, P/NJFETs, tubes etc. I'd like to learn about successfull output follower stages at line level.
Thanks for your contributions.
Børge
This probably belongs in the tube forums but,
Cathode follower has low output impedance and is highly linear. You probably won't get the tube effect though, as it will add the least amount of coloration. This may or may not be a good thing for you.
If you want tube sound but low Zout I'd do one of the following.
-Use a common tube wired in the conventional grounded cathode stage using it's own dedicated High voltage line. Then use a small step down audio transformer. You could cap-couple it, but the tube wouldn't swing enough voltage at line-level to extract it's characteristics. The step down transformer will also lower it's output impedance and noise floor.
- Try building a low voltage one using these type of tubes.
http://www.duntemann.com/12vtubes/12vtubesindex.htm
Pick one with a low Rp. Do a grounded cathode or SRPP cap-coupled. I've been eager to mess with these myself but haven't got around to it.
Cathode follower has low output impedance and is highly linear. You probably won't get the tube effect though, as it will add the least amount of coloration. This may or may not be a good thing for you.
If you want tube sound but low Zout I'd do one of the following.
-Use a common tube wired in the conventional grounded cathode stage using it's own dedicated High voltage line. Then use a small step down audio transformer. You could cap-couple it, but the tube wouldn't swing enough voltage at line-level to extract it's characteristics. The step down transformer will also lower it's output impedance and noise floor.
- Try building a low voltage one using these type of tubes.
http://www.duntemann.com/12vtubes/12vtubesindex.htm
Pick one with a low Rp. Do a grounded cathode or SRPP cap-coupled. I've been eager to mess with these myself but haven't got around to it.
There are lot's of options but probably no "best" topology, so do some more search! John Broskie's tubecad site has plenty of information on cathode followers and their variations, e.g. Aikido CF, White CF, Super Linear CF, etc. ...
For a JFET source follower discussion see part 2 of the Borbely article JFETS: THE NEW FRONTIERS. You can download his articles from his WEB pages.
P.S. 12 V tubes are a very good idea, especially since you can more easily build versions with a bipolar power supply.
For a JFET source follower discussion see part 2 of the Borbely article JFETS: THE NEW FRONTIERS. You can download his articles from his WEB pages.
P.S. 12 V tubes are a very good idea, especially since you can more easily build versions with a bipolar power supply.
Thanks guys!
I'm not that familiar with tubes. As an alternative to the cathode follower I may work with anything that has a bit of gain and high input impedance. As I said in the first post I'm trying to get by without feedback.
I was also hoping for some transistor based suggestions to reduce the hi-voltage supplies.
If you have any suggestions for tubes that work well for line-level output stages, please let me know.
I'm trying to introduce as little distortion as possible. Hence the no-feedback approach.
Børge
I'm not that familiar with tubes. As an alternative to the cathode follower I may work with anything that has a bit of gain and high input impedance. As I said in the first post I'm trying to get by without feedback.
I was also hoping for some transistor based suggestions to reduce the hi-voltage supplies.
If you have any suggestions for tubes that work well for line-level output stages, please let me know.
I'm trying to introduce as little distortion as possible. Hence the no-feedback approach.
Børge
The supply has not been designed yet. But I should be able to crank it up to 100V if need be. However, I'm more comfortable below, say, +-30V.
The output of the follower (or output amp) is 2Vrms, +-2.82V. So in the follower case the input voltage will be about the same.
Alternatively, I have been considering making a class-A output stage with a gain in the neighborhood of 1-5. The important thing there, again, is very good linearity, low noise, high output impedance and low output impedance for line driving.
Børge
The output of the follower (or output amp) is 2Vrms, +-2.82V. So in the follower case the input voltage will be about the same.
Alternatively, I have been considering making a class-A output stage with a gain in the neighborhood of 1-5. The important thing there, again, is very good linearity, low noise, high output impedance and low output impedance for line driving.
Børge
I'm going to prove my ignorance now!
Does it? I've only real experience of emitter followers, but (unless its a cfp/sziklai arrangement, or maybe there is a cap in parallel with the emitter resistor) surely an emitter follower is just adding current gain, and the voltage output is Vbe less than the input voltage? Because of the current gain, you have the ability to 'match impendances'. Therefore, its open loop, and there is no feedback.
I may well have the wrong end of the stick there.
Does it? I've only real experience of emitter followers, but (unless its a cfp/sziklai arrangement, or maybe there is a cap in parallel with the emitter resistor) surely an emitter follower is just adding current gain, and the voltage output is Vbe less than the input voltage? Because of the current gain, you have the ability to 'match impendances'. Therefore, its open loop, and there is no feedback.
I may well have the wrong end of the stick there.
You're right about Vbe, but if you think a little harder, you'll realise that can't be exactly right, because if it was, there would be nothing to drive the transistor. The fuller way to look at an emitter follower is to move the emitter load resistor up to the collector, calculate the voltage gain under those conditions, then apply 100% feedback to see what the emitter follower gain would be:
As an example, suppose you had a 20V supply, a 10k load resistance and you biased to drop half the supply voltage across that load resistance.
Ic would be 10V/10k = 1mA.
gm = 35Ic = 35mA/V
Av = gmRL = 35 x 10 = 350.
Your common emitter stage would have an open-loop gain of 350. But you slide the 10k load resistance through the transistor back into the emitter circuit, causing 100% feedback (B=1). You calculate the new gain:
A = A0/1+BA0 = 350/(1+1x350) = 350/351 = 0.997
Which is very nearly 1, but not quite.
As an example, suppose you had a 20V supply, a 10k load resistance and you biased to drop half the supply voltage across that load resistance.
Ic would be 10V/10k = 1mA.
gm = 35Ic = 35mA/V
Av = gmRL = 35 x 10 = 350.
Your common emitter stage would have an open-loop gain of 350. But you slide the 10k load resistance through the transistor back into the emitter circuit, causing 100% feedback (B=1). You calculate the new gain:
A = A0/1+BA0 = 350/(1+1x350) = 350/351 = 0.997
Which is very nearly 1, but not quite.
I follow you until you get to ...
" But you slide the 10k load resistance through the transistor back into the emitter circuit, causing 100% feedback (B=1). "
I need to understand that. I'll go away and investigate. Otherwise, it looks like a different way to get to the same result.
Thanks for the time to describe it. I need to read up.
Anyway, back on thread, I've just finished an IV stage for my DAC, and I also need to add some form of buffer. I was probably going to use a complimentary feedback pair acting as an emitter follower, with a two transistor current source instead of a load resistor. I've used a similar circuit in a headphone amp and it seemed to work well. Are there any obvious ways to improve it? I suspect cascoding would improve it, but I've never looked into it.
I guess thats an answer but also more questions.
Cheers,
Phil
" But you slide the 10k load resistance through the transistor back into the emitter circuit, causing 100% feedback (B=1). "
I need to understand that. I'll go away and investigate. Otherwise, it looks like a different way to get to the same result.
Thanks for the time to describe it. I need to read up.
Anyway, back on thread, I've just finished an IV stage for my DAC, and I also need to add some form of buffer. I was probably going to use a complimentary feedback pair acting as an emitter follower, with a two transistor current source instead of a load resistor. I've used a similar circuit in a headphone amp and it seemed to work well. Are there any obvious ways to improve it? I suspect cascoding would improve it, but I've never looked into it.
I guess thats an answer but also more questions.
Cheers,
Phil
How about Allen Wright's Super Linear Cathode Follower?
FVP5 circuit
If you look at the output stage (the right-hand three valves) the middle one is a cathode follower, the lower one is a current sink, and the top one bootstraps the output to the cathode of the CF so that it has a constant voltage between cathode and anode.
A CF is substantially improved by the current sink on the cathode, as it makes it a more of a bidirectional device, and the bootstrapping improves the linearity.
I built the differential RTP3, and it sounds great to my ears.
Alex
FVP5 circuit
If you look at the output stage (the right-hand three valves) the middle one is a cathode follower, the lower one is a current sink, and the top one bootstraps the output to the cathode of the CF so that it has a constant voltage between cathode and anode.
A CF is substantially improved by the current sink on the cathode, as it makes it a more of a bidirectional device, and the bootstrapping improves the linearity.
I built the differential RTP3, and it sounds great to my ears.
Alex
If you want tubes ECC86 / 6GM8 is about perfect for that voltage range. Maximum anode voltage is 30 volts and accepts much lower voltages, e.g. 12 Volts. If you build an Aikido or a SLCF stage with bipolar supplies you may not need an outpout cap eiter. All necessary information can be found at the Tubecad pages.
Fair point. I think the NPNs were from Farnell and the PNPs off of Ebay - neither were free of charge 
To be honest, I was really pleased with the circuit first time round and will probably just use it again. It might be foolish, but I'd rather have each circuit perform a single task, so the IV stage should do just that, and an additional follower circuit should be used to drive the interconnect IMO.
To be honest, I was really pleased with the circuit first time round and will probably just use it again. It might be foolish, but I'd rather have each circuit perform a single task, so the IV stage should do just that, and an additional follower circuit should be used to drive the interconnect IMO.
EC8010 said:
It was absolutely obvious from borges initial post that the desire was for no global feedback.
EC8010 if you are going to wade in with the "operates with 100% negative feedback" type of statement relating to local feedback then you could have at least had the common courtesy to explain the difference. (So go on then...)
(Perhaps people might question why you just waded in to demonstrate your 'superior knowledge' without any attempt to actually help at that point).
Borges, philpoole, etc., the emitter resistor in an emitter-follower acts in effect as local feedback. It's not a problem and is just part of normal circuit design.
(There are several threads on this forum where people get into childish pi55ing contests about what is and what is not feedback, and silly arguments based on semantic miss-interpretation, all of which does no more than confuse those who simply wish to learn).
Alex M said:
Thanks Alex,
I must ge getting a bit rusty because I had forgotten about the bootstrap.
So I guess what I'll end up doing is use a cathode follower, feed it a constant current and give it a nice bootstrap to keep it happy. That should be as linear as anything that can be made with as few components.
The output buffer is for a differential signal that comes from a passive IVC. The topology allows me to bias the grid with a DC voltage through a resistor in order to keep the output DC at 0V. I'm actually contemplating to feed the output DC back through an integrator to have the whole differential output DC coupled. I have too little experience with output caps and transformers. I guess this can be done if the active devices are well matched.
Borge
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.