Hi fab,
I did draw up a proposal circuit which should not suffer from CMRR at least.
U1 is a comparator, Q3 forms a CCS which bias the input of U1 so the circuit will start to detect when the output stage is apporaching Class B just before a little bit before it actually does, the current from Q3 CCS (depending on the value of R3) can be very very small and should not affect the amplifier.
The power supply to the comparator is formed around D1, D2, R6, R7, C1, C2, D3, D4 and is not any worser than a bootstrapping topology doing to the output, the current here is very small too.
The zeners dont have to be more then perhaps 5 volt each.
When the comparator sense the Class B it will go fully positive, which means around 5 volt, 5 V minus 0,65 v for the transistor Q1 and you can calculate the resistor R8 needed, as you see it's actually also a CCS which need to have a very small current to activate Q2 which is also a CCS giving the LED proper current.
Ok, bear with this circuit I present here, I did draw it up in only some 15 minutes and should be considered more as a source of inspiration. Theres no component values BTW but I think you figure them out youself!
Good luck!
Michael
I did draw up a proposal circuit which should not suffer from CMRR at least.
U1 is a comparator, Q3 forms a CCS which bias the input of U1 so the circuit will start to detect when the output stage is apporaching Class B just before a little bit before it actually does, the current from Q3 CCS (depending on the value of R3) can be very very small and should not affect the amplifier.
The power supply to the comparator is formed around D1, D2, R6, R7, C1, C2, D3, D4 and is not any worser than a bootstrapping topology doing to the output, the current here is very small too.
The zeners dont have to be more then perhaps 5 volt each.
When the comparator sense the Class B it will go fully positive, which means around 5 volt, 5 V minus 0,65 v for the transistor Q1 and you can calculate the resistor R8 needed, as you see it's actually also a CCS which need to have a very small current to activate Q2 which is also a CCS giving the LED proper current.
Ok, bear with this circuit I present here, I did draw it up in only some 15 minutes and should be considered more as a source of inspiration. Theres no component values BTW but I think you figure them out youself!
Good luck!
Michael
Attachments
CMRR issue solved
Hi Ultima
Thanks for your inputs about the CMMR issue lead. I remembered that a balance amplifier needed matched resistors to improve CMRR. The ratio of resistors at + and - input of balance amplifier must be the same and with 5% tolerance resistors it is not guaranteed...So I added a potentiometer in series with one of the input resistor and could finally null the output of the input balance amplifier when audio signal is applied with no amp load.
The ratio of ac output voltage for class B detecting level is now about closed to "2" for 8 ohms and 4 ohms loads.
Though the detecting level was not accurate since I have parallel mosfets (2), I had to get back to my first version of class B detection circuit where I could divide by 2 only the AC portion and not the DC portion of the signal (half of the current goes into each source resistor for each respective mosfet thus the DC voltage across the source resistor is divided by 2). See attached circuit. Since in my project of a Class A / B amp I intend to use 3 to 5 mosfets in parallel then I will have to divide the AC signal accordingly.
In some unfortunate wiring manipulation I burned out some output stage driver transistors of my power amp...so I will not have the chance to take time at this moment to analyse your circuit. Will your circuit take into account parallel output transistors while sensing only one resistor?
Thanks for all your comments so far.
Hi Ultima
Thanks for your inputs about the CMMR issue lead. I remembered that a balance amplifier needed matched resistors to improve CMRR. The ratio of resistors at + and - input of balance amplifier must be the same and with 5% tolerance resistors it is not guaranteed...So I added a potentiometer in series with one of the input resistor and could finally null the output of the input balance amplifier when audio signal is applied with no amp load.
The ratio of ac output voltage for class B detecting level is now about closed to "2" for 8 ohms and 4 ohms loads.
Though the detecting level was not accurate since I have parallel mosfets (2), I had to get back to my first version of class B detection circuit where I could divide by 2 only the AC portion and not the DC portion of the signal (half of the current goes into each source resistor for each respective mosfet thus the DC voltage across the source resistor is divided by 2). See attached circuit. Since in my project of a Class A / B amp I intend to use 3 to 5 mosfets in parallel then I will have to divide the AC signal accordingly.
In some unfortunate wiring manipulation I burned out some output stage driver transistors of my power amp...so I will not have the chance to take time at this moment to analyse your circuit. Will your circuit take into account parallel output transistors while sensing only one resistor?
Thanks for all your comments so far.
Attachments
dimitri said:
Dimitri my friend,
Congratulations, you are the first person to notify it!
I like your avatar btw, whats your favorite?
Fab,
well theres a small problem, but not an impossible one, if you use several FET's in parallele since theres som difference between them, I wonder if you should not consider having the sensing resistor between all the Drains connected together and the rail voltage, it would be the most easiest and less complex way!
Cheers Michael
Ultima Thule said:
...
Fab,
well theres a small problem, but not an impossible one, if you use several FET's in parallele since theres som difference between them, I wonder if you should not consider having the sensing resistor between all the Drains connected together and the rail voltage, it would be the most easiest and less complex way!
Cheers Michael
Steven said:
... another reason to sense in the collector/drain line...
Steven
Ok, both of you are convincing me but I see another problem with measuring the voltage closed to the power supply rail. I can not measure the real DC voltage corresponding to the DC bias. In my amp the voltage across a 0.27 ohms resistor between +60Vdc node and Drain + (common node to both postive parallel mosfets) is about -0.1 Vdc but my balance amplifier always measure about -0.5 Vdc whatever sensing resistor value I use. It is the same balance amplifier I used for measuring the voltage across the source resistor in my previous circuit and it was working fine with the CMRR improved with a pot.
See attached new circuit.
Is it because the high voltage node being measured is at the max supply rail? if so what could be the solution to measure that voltage?
Thanks
Attachments
I think you have exceeded the common mode input range of your LF351 diff amplfier.
I think it is better to use a few transistors instead of an opamp, e.g. a current mirror, with the emitters on both sides of the sense resistor. Small changes in voltage drop cause big changes in output current of the mirror.
Steven
I think it is better to use a few transistors instead of an opamp, e.g. a current mirror, with the emitters on both sides of the sense resistor. Small changes in voltage drop cause big changes in output current of the mirror.
Steven
Attachments
Hi Fab,
I see you still have an amplifier trying to detect the classB current.
I think you have more chance of measuring the ClassB point if you use a bridge and comparator to detect the almost zero current point.
e.g. connect a resistor string from +Vrail to ground. tap into the string at 0.1V below Vrail and connect this junction to the + input of the comparator. I have assumed for this example that your Vrail sits at 60Vdc.
Connect the - input of the comparator with a large resistor to the drain connection of the measuring resistor off the Vrail.
If my logic is correct then a low voltage on the drain (say = 59.8V) keeps the -input below the + input which should be sitting at 59.9V. Then the output of the comparator is high (with respect to the comparator's supply voltages) indicating that classA current is flowing. When the current to drain drops very close to zero the voltage drop across the measuring resistor is now 0.02v giving a - input voltage of 59.98v the comparator now sees a higher voltage on the - input than the + input and the comparator changes to a low on the output. That's the signal that you need.
By using a bridge to compare voltages, the Vrail can swing around and the comparator compares the voltage at it's two inputs irrespective of the absolute voltage on it's pins. Both pins are connected to the Vrail by resistors.
Q for comparator protection do we need a large resistor feeding each of the input pins?
The supply voltage to the comparator should be centred on the Vrail and have a high and a low greater than the maximum swing of the measuring points plus the spare volts needed because it is not a rail to rail opamp. Lets say +65v and +55v for the comparator rails and I think the supplies need to be referenced to the Vrail not independent. It might be worth putting a couple of diodes across the +- inputs to protect the chip at switch on and off and when the Vrails are doing odd things during protection etc.
I see you still have an amplifier trying to detect the classB current.
I think you have more chance of measuring the ClassB point if you use a bridge and comparator to detect the almost zero current point.
e.g. connect a resistor string from +Vrail to ground. tap into the string at 0.1V below Vrail and connect this junction to the + input of the comparator. I have assumed for this example that your Vrail sits at 60Vdc.
Connect the - input of the comparator with a large resistor to the drain connection of the measuring resistor off the Vrail.
If my logic is correct then a low voltage on the drain (say = 59.8V) keeps the -input below the + input which should be sitting at 59.9V. Then the output of the comparator is high (with respect to the comparator's supply voltages) indicating that classA current is flowing. When the current to drain drops very close to zero the voltage drop across the measuring resistor is now 0.02v giving a - input voltage of 59.98v the comparator now sees a higher voltage on the - input than the + input and the comparator changes to a low on the output. That's the signal that you need.
By using a bridge to compare voltages, the Vrail can swing around and the comparator compares the voltage at it's two inputs irrespective of the absolute voltage on it's pins. Both pins are connected to the Vrail by resistors.
Q for comparator protection do we need a large resistor feeding each of the input pins?
The supply voltage to the comparator should be centred on the Vrail and have a high and a low greater than the maximum swing of the measuring points plus the spare volts needed because it is not a rail to rail opamp. Lets say +65v and +55v for the comparator rails and I think the supplies need to be referenced to the Vrail not independent. It might be worth putting a couple of diodes across the +- inputs to protect the chip at switch on and off and when the Vrails are doing odd things during protection etc.
working circuit
Hi Steven,
Your suggestion was successful ! It is so simple ...
Here attached the final circuit and the simulation results.
I used 2 x 0.27 ohms (5W) in parallel on the positive mosfets rail as the sensing resistor.
V(1) is at "+" of opamp
v(2) is at "-"
V(7) is at base of Q1
R122 + R117 < R121 (before class B level)
Question: is there a bad side effect on audio to the modulation of the voltage across the sensing resistor upon large current drawn which modifies (reduces) the max power supply voltage on the mosfets?
It self asjusts to speaker load impedance and DC bias current. The resistor value chosen gives the class B threashold voltage at comparator to about 5V (with 60 Vdc supply) which is well within the operation of the comparator under +/- 15 Vdc supply. Also sine the voltage signal is clipped at class B the max output stage will stay close to the threashold voltage.
It seems perfect.
I am glad I used the forum to work out that idea.
Steven said:
Hi Steven,
Your suggestion was successful ! It is so simple ...
Here attached the final circuit and the simulation results.
I used 2 x 0.27 ohms (5W) in parallel on the positive mosfets rail as the sensing resistor.
V(1) is at "+" of opamp
v(2) is at "-"
V(7) is at base of Q1
R122 + R117 < R121 (before class B level)
Question: is there a bad side effect on audio to the modulation of the voltage across the sensing resistor upon large current drawn which modifies (reduces) the max power supply voltage on the mosfets?
It self asjusts to speaker load impedance and DC bias current. The resistor value chosen gives the class B threashold voltage at comparator to about 5V (with 60 Vdc supply) which is well within the operation of the comparator under +/- 15 Vdc supply. Also sine the voltage signal is clipped at class B the max output stage will stay close to the threashold voltage.
It seems perfect.
I am glad I used the forum to work out that idea.
Attachments
real circuit results
The real prototype did not reveal to be as magical as the simulator. Class B detection is well adapted to varying load impedance but not auto adapted to the DC bias current if it is changed. I had to modify the circuit and I am working to make the circuit auto adaptable to the varying of DC bias current.
However, the requirement of auto adaptation of detection of class B level requirement is imperative but less is the auto-adaptation to DC bias current change.
More to come....I hope.
The real prototype did not reveal to be as magical as the simulator. Class B detection is well adapted to varying load impedance but not auto adapted to the DC bias current if it is changed. I had to modify the circuit and I am working to make the circuit auto adaptable to the varying of DC bias current.
However, the requirement of auto adaptation of detection of class B level requirement is imperative but less is the auto-adaptation to DC bias current change.
More to come....I hope.
Final Circuit built
I finally got back to this attached circuit and got good results to detect class B operation since my amp is class A up to about 15Wrms/ 8ohms thus the bias voltage I measure across one mosfet source 0.5 ohm resistor is about 100 mV.
I will also post a picture of the circuit.
I finally got back to this attached circuit and got good results to detect class B operation since my amp is class A up to about 15Wrms/ 8ohms thus the bias voltage I measure across one mosfet source 0.5 ohm resistor is about 100 mV.
I will also post a picture of the circuit.
Attachments
Picture of pcb
The pcb contains :
- Class B detect with led indicator for one channel (right led);
- Clippin level detect with led indicator for one channel (left Led;
I use bi-color Led so left led is green for power indication of amp and becomes red when very close to clipping.
I use clear blue color led for class A operation indication and no lihthing when in class B. The revers could be done or bi-color led also to swith from green (class A) to red (class B).
You need 2 circuits for streeo operation.
The pcb contains :
- Class B detect with led indicator for one channel (right led);
- Clippin level detect with led indicator for one channel (left Led;
I use bi-color Led so left led is green for power indication of amp and becomes red when very close to clipping.
I use clear blue color led for class A operation indication and no lihthing when in class B. The revers could be done or bi-color led also to swith from green (class A) to red (class B).
You need 2 circuits for streeo operation.
Attachments
Why not use a simple comparator chip like LM311 to sample the output, as you know at which voltage into your load you leave class A, You can easily create a voltage divider and a similar reference voltage for the other pin, the output connected to an LED... I have done the inverse as a low battery detection module for RC aircraft, with a piezo that emits noise when servos are activated on low battery. You can get all you need from the datasheet.... If memory serves they can take about 50V
Re: Revisited
Hi Mikee55
Yes I still use it. It is accurate enough for me . Some improvement could be done for the low frequency filtering. About the write-up I am not sure if you are the only one interested. I could simply provide you the information with the clipping indicator.
For use on "any" amp, it is impossible to guaranty that since there is an almost infinite number of amp design possibility...
One thing is that you need a sufficient degeneration resistor on the output stage to have a minimum of accuracy which could be a drawback in some amp design.
mikee55 said:fab, do you still use the indicator, and would you say it was accurate? Also, is there a possibility of some sort of project write up, so the indicator could be used on any amp?
Cheers Mike
Hi Mikee55
Yes I still use it. It is accurate enough for me . Some improvement could be done for the low frequency filtering. About the write-up I am not sure if you are the only one interested. I could simply provide you the information with the clipping indicator.
For use on "any" amp, it is impossible to guaranty that since there is an almost infinite number of amp design possibility...
One thing is that you need a sufficient degeneration resistor on the output stage to have a minimum of accuracy which could be a drawback in some amp design.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.