Fluorescence
Electrical
- Oct 19, 2008
- 42
Hello,
A circuit that doesn’t work is not nice, but a circuit that does work when it shouldn’t is even worse!……….
I am designing mains LED lighting and am using UC3842 PWM controller since it costs just 20 Cents. (Specific LED drivers are too expensive)
I have designed two similar current mode forward converters:-
1. Uses a comparator (no loop compensation) to set the output current.
*****SCHEMATIC OF LED DRIVER (NO LOOP COMPENSATION)
(switching frequency = 100KHz)
2. Uses a compensated external Op-Amp as an error amplifier to set the output current.
*****SCHEMATIC OF LED DRIVER (WITH LOOP COMPENSATION)
(error amp zero = 63Hz, Error amp pole = 63KHz)
*****LT1241 DATASHEET
(LT1241 is very similar to the UC3842)
The strange thing is, the version with absolutely no loop compensation works just as well on the LT Spice simulator as the one with loop compensation.
Both versions give a very satisfactory 5mA of ripple current (ripple at switching frequency)
Here is the LED current from the non-compensated LED driver
*****LED CURRENT (non-compensated version)
Here is the LED current from the compensated LED driver
*****LED CURRENT (compensated version)
(The difference in current level is due to the different COMP pin resistor values)
LT Spice yahoo forum assure me that LT Spice does not lie.
The non-compensated converter is cheaper as it uses less parts.
However, I am worried about using it because from an academic viewpoint, it is not stable.
Do you think I would be allright using the non-compensated version.?
I know that current mode forward converters have a double pole at half the switching frequency.
In the compensated version I have used a type II compensator to ensure that I roll the gain off well before this double pole, in order to give myself adequate gain and phase margin.
In the non-compensated version, the comparator does not give the 180 degree minimum phase shift that the op-amp compensated converter does.
-However, as the frequency gets extremely high, the comparator delay becomes significant and it will indeed begin to show a phase.
-Indeed there will be a frequency at which the loop phase comes to minus 360 degrees, and I am just presuming that my loop gain has gone through zero before that happens. But has it ? I am doubtful
I have actually built a hardware version of the non-compensated converter and amazingly it works like a dream and shows no instability.
………Even though from an academic standpoint it is severely bad practice.
-It cannot be induced to go unstable …even switching the load repeatedly full_on and full_off doesn’t make it oscillate.
I am afraid that I would have difficulty in deriving the modulator transfer function for these converters as I am using non-standard configurations.
The power stage transfer function is straight forward, with a double pole at half the switching frequency.
The error amp transfer function is less straightforward as the upper divider resistor does not go to the output voltage , like in a “normal” voltage regulating SMPS.
The error amp transfer function is Zfeedback/Zin .
-Usually, with a “normal” voltage regulating converter, Zin is simply the upper divider resistor.
……..However, I am not sure what it would be with my converter…..i would presume, possibly wrongly, that Zin for the error amplifier would be the lower divider resistor, since the upper divider resistor is connected to an AC ground.
-Also, I am unaware of the modulator transfer function due to the unusual configuration here.
Do you think I would be safe to use the cheaper, non- compensated converter ?
Or any thoughts at all appreciated.
Apologies for the length of this post
Thankyou for reading.
A circuit that doesn’t work is not nice, but a circuit that does work when it shouldn’t is even worse!……….
I am designing mains LED lighting and am using UC3842 PWM controller since it costs just 20 Cents. (Specific LED drivers are too expensive)
I have designed two similar current mode forward converters:-
1. Uses a comparator (no loop compensation) to set the output current.
*****SCHEMATIC OF LED DRIVER (NO LOOP COMPENSATION)
(switching frequency = 100KHz)
2. Uses a compensated external Op-Amp as an error amplifier to set the output current.
*****SCHEMATIC OF LED DRIVER (WITH LOOP COMPENSATION)
(error amp zero = 63Hz, Error amp pole = 63KHz)
*****LT1241 DATASHEET
(LT1241 is very similar to the UC3842)
The strange thing is, the version with absolutely no loop compensation works just as well on the LT Spice simulator as the one with loop compensation.
Both versions give a very satisfactory 5mA of ripple current (ripple at switching frequency)
Here is the LED current from the non-compensated LED driver
*****LED CURRENT (non-compensated version)
Here is the LED current from the compensated LED driver
*****LED CURRENT (compensated version)
(The difference in current level is due to the different COMP pin resistor values)
LT Spice yahoo forum assure me that LT Spice does not lie.
The non-compensated converter is cheaper as it uses less parts.
However, I am worried about using it because from an academic viewpoint, it is not stable.
Do you think I would be allright using the non-compensated version.?
I know that current mode forward converters have a double pole at half the switching frequency.
In the compensated version I have used a type II compensator to ensure that I roll the gain off well before this double pole, in order to give myself adequate gain and phase margin.
In the non-compensated version, the comparator does not give the 180 degree minimum phase shift that the op-amp compensated converter does.
-However, as the frequency gets extremely high, the comparator delay becomes significant and it will indeed begin to show a phase.
-Indeed there will be a frequency at which the loop phase comes to minus 360 degrees, and I am just presuming that my loop gain has gone through zero before that happens. But has it ? I am doubtful
I have actually built a hardware version of the non-compensated converter and amazingly it works like a dream and shows no instability.
………Even though from an academic standpoint it is severely bad practice.
-It cannot be induced to go unstable …even switching the load repeatedly full_on and full_off doesn’t make it oscillate.
I am afraid that I would have difficulty in deriving the modulator transfer function for these converters as I am using non-standard configurations.
The power stage transfer function is straight forward, with a double pole at half the switching frequency.
The error amp transfer function is less straightforward as the upper divider resistor does not go to the output voltage , like in a “normal” voltage regulating SMPS.
The error amp transfer function is Zfeedback/Zin .
-Usually, with a “normal” voltage regulating converter, Zin is simply the upper divider resistor.
……..However, I am not sure what it would be with my converter…..i would presume, possibly wrongly, that Zin for the error amplifier would be the lower divider resistor, since the upper divider resistor is connected to an AC ground.
-Also, I am unaware of the modulator transfer function due to the unusual configuration here.
Do you think I would be safe to use the cheaper, non- compensated converter ?
Or any thoughts at all appreciated.
Apologies for the length of this post
Thankyou for reading.
