Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Audio amplifier ever in class B

Status
Not open for further replies.

northumber82

Member level 3
Joined
Jan 6, 2018
Messages
66
Helped
2
Reputation
4
Reaction score
2
Trophy points
8
Activity points
564
Hello, straight to the problem: my class AB amp remains ever in class B.

Here is the schematic:
Immagine.png

What is the oscilloscope out: (simulated because I don't have a photo)
Immagine2.png

In the simulation it is perfect.

What I've tried:
- Lowing resistor of bias (so much that it is even finished in self-oscillation)
- Put a single transistor instead of the push-pull bias and change resistances

Why it ever remains in class B with the crossover distortion? Thanks for answers
 

Hi,

R12, R22 .. one of these should be variable to adjust bias current of the power stage.

This current should be somewhere between 10mA and 100mA per transistor.
This means a voltage drop of 1mV ... 10mV across the source_resistors.

Be careful. Too high current will cause a lot of heat. Don´t overheat the transistors.

Klaus
 

The best designs add bias current through the output fets as you get near the zero crossing.
 

Hi,

R12, R22 .. one of these should be variable to adjust bias current of the power stage.

This current should be somewhere between 10mA and 100mA per transistor.
This means a voltage drop of 1mV ... 10mV across the source_resistors.

Be careful. Too high current will cause a lot of heat. Don´t overheat the transistors.

Klaus

Yes, are the value of R12 and R22 that I've changed, lowing seriously down, but nothing changed :(

Down from 180 to 47 ohm R12 and R22 to 220
 

It is completely symetrical so both output transistors are identically biased. To allow class A operation at low levels you need to ensure both have a small residual current flowing through them. The components around Q13 & Q20 are to overcome the Vgs threshold of the output transistors, if more voltage is dropped across them, it increases the voltage producing the threshold and takes them into a conductive state. So what you have to do is decrease the value of R12 or R22 until a quiescent current flows through the output transistors. It might be worth using a pre-set variable resistor to find the optimum value. Be careful not to let it go open circuit or to a much higher value as it could damage the output transistors.

Brian.
 
Hi,

Yes, are the value of R12 and R22 that I've changed, lowing seriously down, but nothing changed :(

Down from 180 to 47 ohm R12 and R22 to 220
You need to make them higher ohmic, not lower.


Klaus
 
Hint;
in your simulation, perform a .DC analysis.

Check the voltage values across the gates of Mi and M2, and the current thru R16 and R18.

It will become clear to you what is going on.
 

Done some tests in the real-life. Lowing down R13 and R23, and increase R12 and R22. The crossover distortion is disappearing. Thank you
 

Measure or simulate the DC bias points! Don't just change values without understanding what is going on.

You could go all the way to full class A mode. Which is not a bad thing from the audio perspective, but could cause excessive output transistor heating.
 

Measure or simulate the DC bias points! Don't just change values without understanding what is going on.

You could go all the way to full class A mode. Which is not a bad thing from the audio perspective, but could cause excessive output transistor heating.

Voltage into M1: 4V
Voltage into M2: 0.9V

Current R15: 130uA
Current R16: 10uA

I need to increase the bias?
 

According to R12,R13,R22,R23 and theoretical bias voltage for M1,M2,M3,M4 you should read approx. 8Vdc between opposed MOSFET gates. That is theoretical. In practice, although I have not reversed engineered that schematic completely, it appear at first glance that this circuit would be quite unstable when overall temperature is raised. Because Q10 and Q44 DC bias is controlled by temperature dependent 5 transistors KSA992 (assumed to be 2SA992 and complementary 2SC1845).
A minute change in Base bias voltage for BD140(and BD139) will dramatically change the bias voltage for M1-M2-M3-M4. Hence, dramatically changing bias current on R16-R18-R39-R42. So, my first observation is: BD139 and BD140 Emitter must be in series with an appropriate value resistor to lower its base bias voltage variation effect.
According to manufacturer specs, a change of 0.1V on BD140 VBE will change its collector current by 100 folds. This could be catastrophic to both the output MOSFET and BD140 itself, (Caused by excessive heat). Notice that those BD140 and BD139 MUST be heat sink.
Approximation attempt of calculating BD140 functional collector current leads to a value in the vicinity of 50mA, which is bare minimum to be able to decently drive those 4 MOSFET gates. So, loose a little voltage from an Emitter resistor, something like 0.1V/.05A= 20ohms, and it should substantially help bias current regulation. R1 and R2 should then be raised to compensate this 0.1V change. Values that I cannot calculate because the original values are not marked on your schematic.

Over and above those modifications, R12 - 180ohms resistor could be changed for a 800 or 1k ohms precision 10 turns TrimPot and a 250 ohms parallel resistor to prevent catastrophic current if the pot ever become intermittent. This way you will be able to accurately set the bias current.
Further improvement could be implemented but I would need all parts values.

Where did you get this schematic? I'd love to reverse engineer this thing and add it to my collection of power amp designs. It's a decent setup.
Hope this help
Cheers
 

According to R12,R13,R22,R23 and theoretical bias voltage for M1,M2,M3,M4 you should read approx. 8Vdc between opposed MOSFET gates. That is theoretical. In practice, although I have not reversed engineered that schematic completely, it appear at first glance that this circuit would be quite unstable when overall temperature is raised. Because Q10 and Q44 DC bias is controlled by temperature dependent 5 transistors KSA992 (assumed to be 2SA992 and complementary 2SC1845).
A minute change in Base bias voltage for BD140(and BD139) will dramatically change the bias voltage for M1-M2-M3-M4. Hence, dramatically changing bias current on R16-R18-R39-R42. So, my first observation is: BD139 and BD140 Emitter must be in series with an appropriate value resistor to lower its base bias voltage variation effect.
According to manufacturer specs, a change of 0.1V on BD140 VBE will change its collector current by 100 folds. This could be catastrophic to both the output MOSFET and BD140 itself, (Caused by excessive heat). Notice that those BD140 and BD139 MUST be heat sink.
Approximation attempt of calculating BD140 functional collector current leads to a value in the vicinity of 50mA, which is bare minimum to be able to decently drive those 4 MOSFET gates. So, loose a little voltage from an Emitter resistor, something like 0.1V/.05A= 20ohms, and it should substantially help bias current regulation. R1 and R2 should then be raised to compensate this 0.1V change. Values that I cannot calculate because the original values are not marked on your schematic.

Over and above those modifications, R12 - 180ohms resistor could be changed for a 800 or 1k ohms precision 10 turns TrimPot and a 250 ohms parallel resistor to prevent catastrophic current if the pot ever become intermittent. This way you will be able to accurately set the bias current.
Further improvement could be implemented but I would need all parts values.

Where did you get this schematic? I'd love to reverse engineer this thing and add it to my collection of power amp designs. It's a decent setup.
Hope this help
Cheers

I did not get him anywhere, I designed this myself using Cordell and Self essays and done a lot and a lot of attempts (simulation and real life).

About the heat developed in BD139-140 I don't care, because I already have an amplifier like this (a prototype) and those do not exceed normal room temperature without the emitter series resistor. In addition, at least according to simulation, put a series resistor the distortion would increase considerably, and that's exactly what I want to avoid.

About instead the voltage bias of the transistor, you said 8Vdc? R13 and R23 where changed with 220ohm resistor, and R12-22 with an unique 10k res. In the simulation I read 4Vdc for the mosfet gate, should I increase this?
 

Hi,

From post#2:
This current should be somewhere between 10mA and 100mA per transistor.
With 100mA you surely are in class A mode for quiet listening.

Klaus
 

In the simulation I read 4Vdc for the mosfet gate, should I increase this?

Datasheet tells you about Vth range of 2 to 4V, simulation models use a typical value. Respectively the gate voltage for your intended bias current must be expected to vary between transistor exemplars by worst case +/- 1V around the simulated value. This means, it's impossible to operate this circuit without individual bias adjustment. Selecting paralleled transistors for equal Vth is also suggested.

I'm not sure if the bias circuit is safe against thermal runaway.
 

Northumber82, good practice would be to thermally bond the bias circuit to the output transistor heatsink in such a way that the bias compensates for the rise in temperature.

I agree with FvM though, unless you selected individual MOSFETs for near identical characteristics, you will have trouble finding a good class A to class B changeover point. You may be forced to over bias one transistor to bring the other the the optimum quiescent current. To some degree the negative feedback might try to do that for you but at the expense of the center rail shifting toward one or the other supplies.

Brian.
 
Can I in this case use a LT1166 to replace completely the bias? For exampe
 

About the heat developed in BD139-140 I don't care, because I already have an amplifier like this (a prototype) and those do not exceed normal room temperature without the emitter series resistor. In addition, at least according to simulation, put a series resistor the distortion would increase considerably, and that's exactly what I want to avoid.

Sorry for the delay to get back. I have taken a look at this document from Cordell:
https://www.cordellaudio.com/papers/MOSFET_Power_Amp.pdf
Look at figures 7 and 12. In both examples the drivers (ie: your Q10 Q44) are connected with an Emitter resistor. Furthermore instead of using a PNP to drive the Positive Mosfet, he is using an NPN as a voltage follower. Contrary, in your circuit you are using the PNP as a current amplifier and also a voltage amplifier. In my opinion, it is not necessarily a bad design but you need an emitter resistor to lower the voltage gain of these transistors otherwise you may endup with too much gain and see internal instability probably enough to generate oscilation. Also, in an ideal world your design may look fine to the simulator but in real life no two transistors are alike. There is always a minute difference between each unit and this is why engineers are using Emitter resistor to compensate for this difference. I realise that you will lower the internal gain of your driver stage, hence raise the distortion a little, this is the price to pay for stability. Also, in real pratical situation, I am not convinced that this would really raise distortion. This is one test to be done with real instruments.

I am also commenting on your other argument in another post below

Good luck and welcome to comment again. It keeps my brain alive! Cheers

- - - Updated - - -

About instead the voltage bias of the transistor, you said 8Vdc? R13 and R23 where changed with 220ohm resistor, and R12-22 with an unique 10k res. In the simulation I read 4Vdc for the mosfet gate, should I increase this?

I said 8Vdc, I meant 8V between the two gates. This would be like 4Vdc for each gate (M1 and M2).
Over and above those modifications, R12 - 180ohms resistor could be changed for a 800 or 1k ohms precision 10 turns TrimPot and a 250 ohms parallel resistor to prevent catastrophic current if the pot ever become intermittent. This way you will be able to accurately set the bias current.
Further improvement could be implemented but I would need all parts values.
Your modification with 220ohms and 10k dont seem to make sense to me. It would raise the R13 current to 15mA. Considering the 10K would only see 1.3V between the two bases (Q13,Q20) a 130microAmp. that means your two Q13 Q20 are taking some 14.9mA into base. Enough current to saturate the two Qs and bring bias to a minimum, like 0.1V , which will shot your output mosfet bias to almost zero. Cant work. Unless you mistakenly said 220ohms and instead you plugged in 22k resistors for R13 and R23. In that case it may work fine. But since your Mosfet Gate bias current is in the range of mAmps, I see no problem with the original values. My suggestion with the 1k trimpot and 250 ohms resistor was to give you an opportunity to perform a very precise bias current adjustment. But your original resistor choice were correct to start with.

- - - Updated - - -

Can I in this case use a LT1166 to replace completely the bias? For exampe
Absolutely, the LT1166 is an excellent choice for this job.
 

Attached is a PDF. It show my modified version of your amplifier with the LT1166 implementation.
DIDnt try it but pretty confident that it should work as expected since your version was said to be validated with the simulator.
Hope you will try that and let us know how it goes.
 

Attachments

  • EDA2.png
    EDA2.png
    47.9 KB · Views: 223

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top