# 50A,15V power supply ; help needed on error amplifier circuit

#### nycokello

Joined Aug 10, 2009
35
I intend to use this circuit from Texas instruments tl494 datasheet to build 50a,15v power supply. My input voltage is 24dc from a step down ordinary transformer.
From this circuit, that av screenshot with my phone ,the output current is 10a and is controlled with the current sensing resistor R13 which is 0.1 ohms ; meaning it's dissipating 10w, at 1V. This one implies that my 50a will dissipate 50w, at 1v.
Now my question is;
Can I reduce this error amplifier's sensing voltage to 0.5v so that I can save 25w of dissipating power ; or will it affects the normal operation of tl494?

#### Attachments

• 65.2 KB Views: 30

#### Dodgydave

Joined Jun 22, 2012
9,940
You can set any ref voltage on pins 15,16 of the error amps, so you can lower the sense resistor also if you wanted.

NOTE, the circuit shown is fixed at 5V output.

#### ebp

Joined Feb 8, 2018
2,332
The data sheet isn't really helpful, but it does say that the recommended operating range for the amplifier inputs is -0.3 V to the chip supply voltage minus 2 volts. You should be able to simply reduce the voltage at the inverting input of the current limit amp and reduce the current sense resistor in proportion. You might run into noise sensitivity if you try to push it too far, but I think it is likely you could go down to something around 200 mV without a problem.

Building a 50 ampere buck converter isn't going to be easy. That IC is a very old design and not well suited for high frequency converters. You will probably have to use at least three or four transistors in parallel to handle the current. The example circuit is only running at 20 kHz. A physically very large and heavy inductor is going to be required. The lower frequency and comparatively low switching speed of bipolar transistors will make some of the possible noise problems easier.

#### AnalogKid

Joined Aug 1, 2013
9,189
Generally speaking, scaling up switchers does not work as well as scaling up linears. For a 5x increase in output current, the magnetics have to be completely rethought. For 50 A you probably will need multiple power transistors in parallel. Or change to one huge power MOSFET.

ak

#### nycokello

Joined Aug 10, 2009
35
Generally speaking, scaling up switchers does not work as well as scaling up linears. For a 5x increase in output current, the magnetics have to be completely rethought. For 50 A you probably will need multiple power transistors in parallel. Or change to one huge power MOSFET.

ak
Thanks Analogkid...
I av no problem with those, of course I'll use mosfet, irf3205 which is available here, I know how to drive it. Inductors and the diodes I'll also change. I'll also add other protection circuits.
But my concern is just whether we can use 0.5 v on error amplifiers instead of the 1v they are using.. .

I wish I could minimise the power dissipating in the current sensing resistor

Last edited by a moderator:

#### AnalogKid

Joined Aug 1, 2013
9,189
From this circuit, that av screenshot with my phone ,the output current is 10a and is controlled with the current sensing resistor R13 which is 0.1 ohms ; meaning it's dissipating 10w, at 1V. This one implies that my 50a will dissipate 50w, at 1v.
No, it doesn't. It implies that your 50 A circuit would dissipate 250 W in R13. Joule's Law. You have a much larger problem to solve than you thought.
But my concern is just whether we can use 0.5 v on error amplifiers instead of the 1v they are using..
Yes. If you reduce R13 to 0.02 ohm, then the sense voltage at 50 A in your circuit will equal the sense voltage at 10 A in the original circuit. This is a good starting point because it minimizes the changes to the original circuit, which we presume actually works. However, the power dissipation in R13 still will be 50 W. Better than 250 W, but still a large number.

To reduce this you will have to reduce R13 even more, and adjust the values of R1 and R2 to lower the comparator trip point. For example, reducing R13 to 0.01 ohm reduces its power dissipation to 25 W, and reduces the peak current sense voltage to 0.5 V. Increasing R2 to 9 K reduces the trip point to 0.5 V. The only problem I see is that lower the voltages at the comparator inputs are, the more the comparator's noise and offset voltage and bias current errors might affect its operation.

Separate from that, be sure to use a Kelvin connection at the sense resistor.

ak

Last edited:

#### nycokello

Joined Aug 10, 2009
35
No, it doesn't. It implies that your 50 A circuit would dissipate 250 W in R13. Joule's Law. You have a much larger problem to solve than you thought.

Yes. If you reduce R13 to 0.02 ohm, then the sense voltage at 50 A in your circuit will equal the sense voltage at 10 A in the original circuit. This is a good starting point because it minimizes the changes to the original circuit, which we presume actually works. However, the power dissipation in R13 still will be 50 W. Better than 250 W, but still a large number.

To reduce this you will have to reduce R13 even more, and adjust the values of R1 and R2 to lower the comparator trip point. For example, reducing R13 to 0.01 ohm reduces its power dissipation to 25 W, and reduces the peak current sense voltage to 0.5 V. Increasing R2 to 9 K reduces the trip point to 0.5 V. The only problem I see is that lower the voltages at the comparator inputs are, the more the comparator's noise and offset voltage and bias current errors might affect its operation.

Separate from that, be sure to use a Kelvin connection at the sense resistor.

ak
I agree with your calculations. That was my plan to reduce R13 to 0.01 ohms. And that will give error sensing voltage of 0.5v. And adjust R2 as you've shown. But I'm worried because I can't get information on noise level,offset voltage and current for the error amplifier on this datasheet. So, I decided to share this here to check if there's anyone who might av use it ; as I'm suggesting

#### nycokello

Joined Aug 10, 2009
35
I think Kelvin connection may not be necessary in my case because my power supply's current can be allowed to swing anywhere between 45 to 55A .

But I was also thinking of reducing R13 from 0.1 ohms which is equal to 100 milliohms to 2 milliohms so that I can reduce dissipating power to only 5 watts. If this is done then the sensing voltage will only be 0.1 V. Now I thought of feeding this to some sort of a preamplifier which should scale it up to 1V ; which is then fed to the error amplifier. But I can't figure out the best opamp and circuit for this! have anyone done it before...?

#### AnalogKid

Joined Aug 1, 2013
9,189
Section 7.7 of the datasheet lists the error amplifier errors. With a calibrated 25-50 A load, you can trim out the errors by making R2 a series combination of a fixed resistor and a trimpot. However, the datasheet does not say how the input offset voltage drifts with temperature.

To add a preamp to the current sense signal, consider a part designed for this exact task. I don't know what is available in Kenya, but Linear Technology, Analog Devices, and Maxim make current sense ICs, including some that have the sense resistor built-in.

If you want to go with a generic precision opamp instead, pick one that has an input common mode voltage range that extends below the negative rail and an output that swings down very close to the negative rail. In your application, low input errors are more important than bandwidth or open loop gain.

ak

#### Dodgydave

Joined Jun 22, 2012
9,940
This chip is used in some Atx psus, they use a current transformer to monitor current output.

#### ebp

Joined Feb 8, 2018
2,332
I would expect the amplifier input offset voltage to be 10 mV or less, amounting to 5% error for 200 mV sense voltage (10 W at 50 A).

This circuit places the current sense after the output filter, so a transformer can't be used. A transformer could be used in either the switch or diode current path. The expected duty cycle is such that it should perform quite well.

If an N-channel FET is used a transformer might also be the easiest way to drive it, though if the switching frequency is to be kept low it will require rather high inductance to keep the magnetizing current reasonable. Again the expected maximum duty cycle allows lots of time to reset the core, but care would be needed at power up and power down.

I would be considering a molybdenum-peramalloy powder (MPP) core for the inductor. This would keep losses low, but more importantly it gives the choice of higher permeability which means fewer turns. MPP is expensive! A gapped ferrite core would be cheaper and much easier to wind, but ferrite saturates very abruptly which could be a serious problem with a controller that does not do cycle-by-cycle current limiting. Kool Mu cores from Magnetics might be a good compromise between MPP and powdered iron. I don't think the high flux product line from Micrometals offers any higher permeability than their normal powdered iron products, but they are better suited for high DC bias.
50 amps of DC bias is a lot to cope with. I'd probably target 30-40% reduction in effective permeability with a powder core.

#### nycokello

Joined Aug 10, 2009
35
Section 7.7 of the datasheet lists the error amplifier errors. With a calibrated 25-50 A load, you can trim out the errors by making R2 a series combination of a fixed resistor and a trimpot. However, the datasheet does not say how the input offset voltage drifts with temperature.

To add a preamp to the current sense signal, consider a part designed for this exact task. I don't know what is available in Kenya, but Linear Technology, Analog Devices, and Maxim make current sense ICs, including some that have the sense resistor built-in.

If you want to go with a generic precision opamp instead, pick one that has an input common mode voltage range that extends below the negative rail and an output that swings down very close to the negative rail. In your application, low input errors are more important than bandwidth or open loop gain.

ak
It's not easy to know which components you may get from our electronics shops because they majorly stock ics spare parts for replacement on damaged or failed ones! I feel like aborting my preamp plan!

I'll try my power supply this week by scaling it down to 200millivolts set point and also adding a trimpot as you are saying .And I'll calculate it so that it can vary the reference voltage from 150mv to 1v ;so that I build it on the safer side. I'll also create a room for changing current sensing resistor from 20milliohms to 3milliohms .

#### nycokello

Joined Aug 10, 2009
35
I would expect the amplifier input offset voltage to be 10 mV or less, amounting to 5% error for 200 mV sense voltage (10 W at 50 A).

This circuit places the current sense after the output filter, so a transformer can't be used. A transformer could be used in either the switch or diode current path. The expected duty cycle is such that it should perform quite well.

If an N-channel FET is used a transformer might also be the easiest way to drive it, though if the switching frequency is to be kept low it will require rather high inductance to keep the magnetizing current reasonable. Again the expected maximum duty cycle allows lots of time to reset the core, but care would be needed at power up and power down.

I would be considering a molybdenum-peramalloy powder (MPP) core for the inductor. This would keep losses low, but more importantly it gives the choice of higher permeability which means fewer turns. MPP is expensive! A gapped ferrite core would be cheaper and much easier to wind, but ferrite saturates very abruptly which could be a serious problem with a controller that does not do cycle-by-cycle current limiting. Kool Mu cores from Magnetics might be a good compromise between MPP and powdered iron. I don't think the high flux product line from Micrometals offers any higher permeability than their normal powdered iron products, but they are better suited for high DC bias.
50 amps of DC bias is a lot to cope with. I'd probably target 30-40% reduction in effective permeability with a powder core.
On driving the n-channel mosfet ,I always add another isolated DC power source.it works for me best.
On transformer saturations, have been using inductors in parallel to solve it, and av never encountered any problem.

#### nycokello

Joined Aug 10, 2009
35
I would expect the amplifier input offset voltage to be 10 mV or less, amounting to 5% error for 200 mV sense voltage (10 W at 50 A)
I'm going to try this . And I can be more than happy if reducing it to 200millivolts can still be a safer stable operating zone ; coz dissipating 10w only would be a great save on energy and even heat sinks !

Last edited:

#### nycokello

Joined Aug 10, 2009
35
Dear members,
You are saying that current can also be sensed through a transformers or diodes etc. Can you help me with any circuit diagram of a buck converter having it ; so I may study. But let it not be a synchronized buck! Coz I can't get those ics here!

Last edited:

#### AnalogKid

Joined Aug 1, 2013
9,189
Sensing the current through a transformer is very different from the circuit you are starting with, and the TL494 is not designed for it. Of course, an external circuit van be designed to convert the pulsed current through the transistor switch into a DC level for the TL494, but that is a significant extra effort.

ak

#### nycokello

Joined Aug 10, 2009
35
Sensing the current through a transformer is very different from the circuit you are starting with, and the TL494 is not designed for it. Of course, an external circuit van be designed to convert the pulsed current through the transistor switch into a DC level for the TL494, but that is a significant extra effort.

ak
Ok fine ...
But it's good to learn new things.
I'm self-learned engineer ; and nowadays I can follow everything in electronics even if I never attended lectures on the subject!
Just give me a referral, if you don't mind...

#### ebp

Joined Feb 8, 2018
2,332
I don't know what is available these days because I haven't looked in a long time, but there should be lots of good info on switchmode power supplies at the Texas Instruments site. TI acquired Unitrode several years ago and Unitrode was a leader in SMPS controllers suited for mains-powered supplies but also produced many controllers suitable for DC-DC use. Unitrode published a great many applications notes and put on an annual power supply design seminar. Each year the seminar covered different new topics but the book also included lots of basic information. I know at least some of the seminar books were on the TI site. Lloyd Dixon was the author of many of the articles. There is info on using current sense transformers in there somewhere, but I can't remember exactly where.

Philip Todd from Unitrode came up with a very interesting way to sense DC current using a small ferrite toroid, but it takes some added circuitry. You may find that at TI too.

A company called Coilcraft makes several types of current sense transformers and I believe has some applications information on their website. I have used their transformers in several industrial switchers.

Many of the "current mode" controllers can work very well with a current sense transformer because the inner control loop controls the peak current in the inductor, usually by sensing the peak current in the switch.

I definitely recommend reading everything that Jim Williams ever wrote during his days at Linear Technology if you are interested in analog electronics. In my opinion he wrote some of the best applications notes in the industry. He came up with interesting ways to do things and was very good at writing in a way that helped in understanding the circuits and the issues that are important. I think EDN magazine also maintains an archive of his work. He didn't write a great deal about switchers but he did write some that is very helpful for beginners and people with moderate experience. (Linear has some amazing controllers and fully integrated devices for DC-DC converters)

Bob Pease, who was at National Semiconductor for many years, also wrote lots of good things. You'll be able to find things by him at TI and more at EDN.

Several good authors have produced a large amount of excellent material that you can find on the Analog Devices website, though very little on switchers.

Also have a look around the web for papers by Bruce Carsten. He was (still is ?) a private consultant who specialized in SMPS.

#### nycokello

Joined Aug 10, 2009
35
I don't know what is available these days because I haven't looked in a long time, but there should be lots of good info on switchmode power supplies at the Texas Instruments site. TI acquired Unitrode several years ago and Unitrode was a leader in SMPS controllers suited for mains-powered supplies but also produced many controllers suitable for DC-DC use. Unitrode published a great many applications notes and put on an annual power supply design seminar. Each year the seminar covered different new topics but the book also included lots of basic information. I know at least some of the seminar books were on the TI site. Lloyd Dixon was the author of many of the articles. There is info on using current sense transformers in there somewhere, but I can't remember exactly where.

Philip Todd from Unitrode came up with a very interesting way to sense DC current using a small ferrite toroid, but it takes some added circuitry. You may find that at TI too.

A company called Coilcraft makes several types of current sense transformers and I believe has some applications information on their website. I have used their transformers in several industrial switchers.

Many of the "current mode" controllers can work very well with a current sense transformer because the inner control loop controls the peak current in the inductor, usually by sensing the peak current in the switch.

I definitely recommend reading everything that Jim Williams ever wrote during his days at Linear Technology if you are interested in analog electronics. In my opinion he wrote some of the best applications notes in the industry. He came up with interesting ways to do things and was very good at writing in a way that helped in understanding the circuits and the issues that are important. I think EDN magazine also maintains an archive of his work. He didn't write a great deal about switchers but he did write some that is very helpful for beginners and people with moderate experience. (Linear has some amazing controllers and fully integrated devices for DC-DC converters)

Bob Pease, who was at National Semiconductor for many years, also wrote lots of good things. You'll be able to find things by him at TI and more at EDN.

Several good authors have produced a large amount of excellent material that you can find on the Analog Devices website, though very little on switchers.

Also have a look around the web for papers by Bruce Carsten. He was (still is ?) a private consultant who specialized in SMPS.
Thank you so much for the good literature ; I'll call it a dissertation.
I'll get time to Google them...