Hi Everyone,
I have been trying to design a symmetrical bridge amp and have come up with an (I think) unique design. I would like to discuss this design if you have anything interesting to say I would like to hear it.
I tried to make the input stage fully symmetrical but I am not sure if there are better ways to do it. Any advice? Is my input ground wrong?
I know the output transistor bias is non-optimal, I left that out for clarity. But the way the compensation cap is set (like I remember from NDFL amps) seems to help remove crossover distortion at high freq, so maybe I might not even change the bias.
What input voltage should satuare the output? I figure a bit less than 1v, right?
You can see a GIF of the circuit made with LTspice at:
http://www.deleveld.dds.nl/bridgebjt.gif
Thanks for any comments,
Doug Eleveld
I have been trying to design a symmetrical bridge amp and have come up with an (I think) unique design. I would like to discuss this design if you have anything interesting to say I would like to hear it.
I tried to make the input stage fully symmetrical but I am not sure if there are better ways to do it. Any advice? Is my input ground wrong?
I know the output transistor bias is non-optimal, I left that out for clarity. But the way the compensation cap is set (like I remember from NDFL amps) seems to help remove crossover distortion at high freq, so maybe I might not even change the bias.
What input voltage should satuare the output? I figure a bit less than 1v, right?
You can see a GIF of the circuit made with LTspice at:
http://www.deleveld.dds.nl/bridgebjt.gif
Thanks for any comments,
Doug Eleveld
Hi Doug,
I would suggest that if symmetry is the goal, you should scrap the quasi-complementary output circuit. This only complicates things because the signal for driving the positive half is a follower, and the negative half is a voltage amp. Use complementary pairs. There are many devices to choose from better than 2N3055. Your VAS stages could use current sources, as does the output(& driver) bias. Optional of course, but will help with stability. Does this circuit clip at equal voltages, positive and negative? If not, bridged output will be limited by the lesser.
You could use two seperate differentials for the input, one for each channel, driven by a phase splitter circuit. LTP is not required, but a single transistor input stage has it's drawbacks. I don't see how this circuit would make two equal but opposite signals.
I would suggest that if symmetry is the goal, you should scrap the quasi-complementary output circuit. This only complicates things because the signal for driving the positive half is a follower, and the negative half is a voltage amp. Use complementary pairs. There are many devices to choose from better than 2N3055. Your VAS stages could use current sources, as does the output(& driver) bias. Optional of course, but will help with stability. Does this circuit clip at equal voltages, positive and negative? If not, bridged output will be limited by the lesser.
You could use two seperate differentials for the input, one for each channel, driven by a phase splitter circuit. LTP is not required, but a single transistor input stage has it's drawbacks. I don't see how this circuit would make two equal but opposite signals.
I guess its a fine point but I want looking for symmetry in the NPN, PNP sense, I was looking for it in the bridge sense. NPN/PNP symmetry looks nice on paper but in reality, even for 'matched-paírs' they are different enough not to be truly symmetrical. But a bridge is truly symmetrical even for quasi-complementary pairs.
Another advantage of bridge amps (I think) is that they can have low suseptibility to supply line variations. This means you can use lower power supply capacitance without causing any hum. With the shematic now even 1 volt p-p at the supply causes almost nothing at the output.
You make a good point about the clipping, I am worried a bit about the clipping. The action of Q5 is to hold the average of the two outputs at some steady value. This occurs even if one side of the bridge has clipped and this means a 'fake' clip for the other side. The 'fake' clip means unwanted dissipation and lost overhead.
For some bridge amps you can make the gain of one side of the bridge to be higher than the other. So when the high gain side clips the otherside still has some headroom. The output sees gain-halfing instead of hard clipping. This might sound better than hard clipping but I dont know. I would have liked to be able to do that with this amp but the use of Q5 makes this impossible. I'm reluctant to replace Q5 with a more conventional tail beacuse 1) I have to have extra circuitry to make the outputs sit at 50% of the rail. 2) its the only think that makes my amp a bit original :-/
I wanst really planning on using 2n3055 its just the only big power transistor in LTspice. I was thinking of using the TIP series that I used to use in canada, but I'm not sure if there are readily available in europe where I live now. Any suggestions for a good high-current darlington power transistor?
Thanks very much for your comments!
Doug Eleveld
Another advantage of bridge amps (I think) is that they can have low suseptibility to supply line variations. This means you can use lower power supply capacitance without causing any hum. With the shematic now even 1 volt p-p at the supply causes almost nothing at the output.
You make a good point about the clipping, I am worried a bit about the clipping. The action of Q5 is to hold the average of the two outputs at some steady value. This occurs even if one side of the bridge has clipped and this means a 'fake' clip for the other side. The 'fake' clip means unwanted dissipation and lost overhead.
For some bridge amps you can make the gain of one side of the bridge to be higher than the other. So when the high gain side clips the otherside still has some headroom. The output sees gain-halfing instead of hard clipping. This might sound better than hard clipping but I dont know. I would have liked to be able to do that with this amp but the use of Q5 makes this impossible. I'm reluctant to replace Q5 with a more conventional tail beacuse 1) I have to have extra circuitry to make the outputs sit at 50% of the rail. 2) its the only think that makes my amp a bit original :-/
I wanst really planning on using 2n3055 its just the only big power transistor in LTspice. I was thinking of using the TIP series that I used to use in canada, but I'm not sure if there are readily available in europe where I live now. Any suggestions for a good high-current darlington power transistor?
Thanks very much for your comments!
Doug Eleveld
Yes it would if I thought I could get balanced inputs to work with my CD player. It's just standard unbalanced outputs. If I had balanced output from the CD I would have put the ground at a more usual place (-ve supply).
Bu I havent built it yet so there could well be problems with how the input grounding is.
Thanks,
Doug
Bu I havent built it yet so there could well be problems with how the input grounding is.
Thanks,
Doug
well interesting amp
i have few questions though
1. i donot understand how you want to avoid crossover distortion- at first look it seems pure class b, it us tricky somehow i guess
...confusion...
2. at suppy you intend to use one huge electrlitic cap from + to - rails, right ??? classic use of two caps and ground would fail
here- you would amplify 0 potential.
3. aside from symetry, how you intend to keep low dc through speaker?? as the result of using 1% components-worst case may produce huge dc
sorry if my questions are stupid- i may not understand some ideas
cheers
i have few questions though
1. i donot understand how you want to avoid crossover distortion- at first look it seems pure class b, it us tricky somehow i guess
...confusion...
2. at suppy you intend to use one huge electrlitic cap from + to - rails, right ??? classic use of two caps and ground would fail
here- you would amplify 0 potential.
3. aside from symetry, how you intend to keep low dc through speaker?? as the result of using 1% components-worst case may produce huge dc
sorry if my questions are stupid- i may not understand some ideas
cheers
Well there certianly might be crossover distortion. But in simulations at least C2 and C5 really lower crossover distortion, especially at HF. Notice how the compensation caps C2 and C5 go to the base and the output. Normally they go between the base and collector. At high frequency they are local feedback loops linearizing base-output transfer function. So it tends to lower crossover distortion. I got this idea from an old series of articles called NDFL (nested differential feedback loop) by a name I dont remember (Cherry?). Some people claim that it can cause oscillations but if I remeber correctly the zobel networks at the output are really necessary to reduce this.
Yes the supply is a single cap. Rememebr that caps store energy compared to the square of the voltage. So if you have +40 volts and say 1000uf you have more energy stored than 2 1000uf caps at +-20 volts. And the single cap is going to be cheaper too.
DC at the output depends pretty much only on R4 and R10 and the Q6,Q8 matching. I'm not really sure the best polace for a DC offset trimmer, maybe ontop of R23.
Doug
Yes the supply is a single cap. Rememebr that caps store energy compared to the square of the voltage. So if you have +40 volts and say 1000uf you have more energy stored than 2 1000uf caps at +-20 volts. And the single cap is going to be cheaper too.
DC at the output depends pretty much only on R4 and R10 and the Q6,Q8 matching. I'm not really sure the best polace for a DC offset trimmer, maybe ontop of R23.
Doug
Commentable Experienced Thoughts
Hi Doug,
Nice to hear about bridge amps in professional manner from u.
Your Point regarding QUASI-COMPLIMENTARY is absolutely complainces with professional standards.
Their is far better symmetry obtained using NPN/NPN devices rather than using NPN/PNP devices.
Complementary doesnt stands tall in pro environment very well.
We also use bridge topology exclusively in our amps with only N-channel Mosfets at output.
ur comments regarding advantages of Bridging are quite right.
I have studied ur circuit in detail,
Acc to me some issues must be dealt with,
Biasing Network
Balanced input
Use higher bandwidth devices at output such as MJL21196 or use Mosfets
Feedback network
Replace Bootstrapping with Active Current source.
Hope that it helps
Hi Doug,
Nice to hear about bridge amps in professional manner from u.
Your Point regarding QUASI-COMPLIMENTARY is absolutely complainces with professional standards.
Their is far better symmetry obtained using NPN/NPN devices rather than using NPN/PNP devices.
Complementary doesnt stands tall in pro environment very well.
We also use bridge topology exclusively in our amps with only N-channel Mosfets at output.
ur comments regarding advantages of Bridging are quite right.
I have studied ur circuit in detail,
Acc to me some issues must be dealt with,
Biasing Network
Balanced input
Use higher bandwidth devices at output such as MJL21196 or use Mosfets
Feedback network
Replace Bootstrapping with Active Current source.
Hope that it helps
Hi amp_man_1,
I'm glad to hear that I'm not the only one who sees that bridge amps have some unique advantages.
I dont understand what you mean about the feedback network. The amp already has a feedback network. The junction of R4/R10 is kept at some steady volts, R3 and R29 are the feedback resistors. R43 and R30 are just there for the LTP (well not very long tail) base current. The base of Q6 is a virtual ground (I think) and R1 is the input resistor.
The amp is very easy to make for balanced inputs, even easier than the single ended one I showed. You just change the ground from where it is to the supply negative where it usually is for single supply amps. But my CD player is normal unbalanced outputs so I cant use a fully balanced input. You know what I mean?
As far at the voltage amp biasing with a active current source, all the simulations that I have done so far show no improvement in distortion. And you lose just a little efficifiency because of the lower clipping with active current source. Thats why I prefer the bootstrap. I might check if PSRR is better with active current source but it should still be pretty good even without it becasue of the bridge.
As far as class B is concerned you can easily add a Vbe multiplier there if you want. As I said before I left it out for simplicity. Its not a 'finished' amp. If someone thinks its a joke, well, I dont know what to say.
I'd love to simulate some faster power output transistors but I only have spice models than came with LTspice and the 2n3055 is the only high power transistor there. Do you have any other models for power transisitors that can work with LTspice?
Thanks,
Doug
I'm glad to hear that I'm not the only one who sees that bridge amps have some unique advantages.
I dont understand what you mean about the feedback network. The amp already has a feedback network. The junction of R4/R10 is kept at some steady volts, R3 and R29 are the feedback resistors. R43 and R30 are just there for the LTP (well not very long tail) base current. The base of Q6 is a virtual ground (I think) and R1 is the input resistor.
The amp is very easy to make for balanced inputs, even easier than the single ended one I showed. You just change the ground from where it is to the supply negative where it usually is for single supply amps. But my CD player is normal unbalanced outputs so I cant use a fully balanced input. You know what I mean?
As far at the voltage amp biasing with a active current source, all the simulations that I have done so far show no improvement in distortion. And you lose just a little efficifiency because of the lower clipping with active current source. Thats why I prefer the bootstrap. I might check if PSRR is better with active current source but it should still be pretty good even without it becasue of the bridge.
As far as class B is concerned you can easily add a Vbe multiplier there if you want. As I said before I left it out for simplicity. Its not a 'finished' amp. If someone thinks its a joke, well, I dont know what to say.
I'd love to simulate some faster power output transistors but I only have spice models than came with LTspice and the 2n3055 is the only high power transistor there. Do you have any other models for power transisitors that can work with LTspice?
Thanks,
Doug
deleveld said:
Hi DOUG,
u and me are the only members in this forum i think who supports bridging in power amps thats why its also a unique point too
Bridging is far reliable when u go beyond 500WRMS , highpower is obtained best with bridging only thats why CROWN Audio amps also uses bridging in their amps.
In my opinion the input Should be totally Dual Differential type which can easily be converted into any type balanced or unbalanced and output stage must be quasi complementary.
Secondly In this type of stage the feedack must be symmetric with respect to its signal reference.
A common mode node must be implemented to reference the bridge as in differential in/out opamps.
Ur feedback network is good only in simulationbut it wont be good in reality because of asymmetric swing resulting from unbalanced input, a balanced type has some symmetry preservation of both output tunnels.
The current source increase PSRR and also increases the slew rate of amp, thats why its important and it also has a size reduction factor rather than using a big cap for bootstrapping.
The spice models are available form the manufactures of power devices such as WWW.ONSEMI.COM
Doug u r the rite guy in this thread with whome i can share my views, never pay attention to useless remarks made by others.
I myself wanted to be part of this project to design a full dual differential balanced/unbalanced input with differential outputs using cascodes,current mirrors and every thing possible by simple intellect manner. just want to quest deep in amp world.
remember me as a friend
ampman
Hi ampman
I will think about your comments about symmetrical feedback and common mode. I think you are right that an good amp should have differential inputs, especially a bridge amp. Its a good solution. I'll try to think how I can get my amp design to have this without adding too many pieces.
I am not too much of a fan of cascodes although I do see that they might be useful in specific places. Its just too many pieces to connect. I prefer "elegant" designs, especially unusual ones, and I prefer to have a low parts count. Or at least a fairly simple schematic.
I am not an audiophile, I just like to design and build amps (although I havent built one in a long time). I see it more of an artistic process, like painting a picture.
The site you mentioned is a GREAT site for spice models! Thanks very much!
Doug
I will think about your comments about symmetrical feedback and common mode. I think you are right that an good amp should have differential inputs, especially a bridge amp. Its a good solution. I'll try to think how I can get my amp design to have this without adding too many pieces.
I am not too much of a fan of cascodes although I do see that they might be useful in specific places. Its just too many pieces to connect. I prefer "elegant" designs, especially unusual ones, and I prefer to have a low parts count. Or at least a fairly simple schematic.
I am not an audiophile, I just like to design and build amps (although I havent built one in a long time). I see it more of an artistic process, like painting a picture.
The site you mentioned is a GREAT site for spice models! Thanks very much!
Doug
Commentable Experienced Thoughts
Hi Doug,
Some Helppul Tips though.
Input _ Dual Differential For symmetric feedback
Current sources must be implemented along with cascode loaded stage
Transconductance Stage
Again Casode it for better high Freuency response and phase response
Driver stage
use deep emitter stage to increase current amplification thereby inc. slew rate.
output stage
Use parallel devices with emitter resistor of value around 0.25 ohms 5W if BJTs are used.
Finally implement a common mode node to reject unwanted noise
and use thermal tracking for biasing.
hope it helps
Hi Doug,
Some Helppul Tips though.
Input _ Dual Differential For symmetric feedback
Current sources must be implemented along with cascode loaded stage
Transconductance Stage
Again Casode it for better high Freuency response and phase response
Driver stage
use deep emitter stage to increase current amplification thereby inc. slew rate.
output stage
Use parallel devices with emitter resistor of value around 0.25 ohms 5W if BJTs are used.
Finally implement a common mode node to reject unwanted noise
and use thermal tracking for biasing.
hope it helps
Hi amp_man,
I changed the amp input stage to make it single ended. I couldnt get it to work the way I want with a high input impedance. I only seem to be able to get low inpedeance inputs. Do you have any suggestions as to how to make a real differential input? I would really like to keep the amp simple and not add opamps....
I dont really like this amp as much as the last one, the schematic is less elegant but it is still pretty unusual.
I wont put in any cascodes I never use them and dont really thing i know enough about to use them optimally.
http://www.deleveld.dds.nl/bridge2.gif
I also checked that I get no big bumps at startup.
http://www.deleveld.dds.nl/startup.gif
I just realised that this amp has LOTS of feedback paths.
1) The normal feedback from the output
2) The average output path to the third diff head
3) The boostrap feedback via the tail current
I am not certian that the diff triple tail current from the bootstrap might be inviting problems because of the big electrolytic cap in the feedback path. But the gain for that path is low so it might not really be a problem.
Doug
I changed the amp input stage to make it single ended. I couldnt get it to work the way I want with a high input impedance. I only seem to be able to get low inpedeance inputs. Do you have any suggestions as to how to make a real differential input? I would really like to keep the amp simple and not add opamps....
I dont really like this amp as much as the last one, the schematic is less elegant but it is still pretty unusual.
I wont put in any cascodes I never use them and dont really thing i know enough about to use them optimally.
http://www.deleveld.dds.nl/bridge2.gif
I also checked that I get no big bumps at startup.
http://www.deleveld.dds.nl/startup.gif
I just realised that this amp has LOTS of feedback paths.
1) The normal feedback from the output
2) The average output path to the third diff head
3) The boostrap feedback via the tail current
I am not certian that the diff triple tail current from the bootstrap might be inviting problems because of the big electrolytic cap in the feedback path. But the gain for that path is low so it might not really be a problem.
Doug
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.