Creating a custom voltage on a toroidal transformer

[William Gaatjes]

Thanks.
Early indications (so I can make excuses if I am WRONG) are that it will happily pass the extra current (heading towards your 5 amps), without involving the LM317's in-built current limits.

Somehow (it was probably TOO SMALL to see easily), I missed the external 0.2 Ohm resistor. So I could NOT see how the current limit was being sorted out.

How big/powerful is your case's (TO3), heatsinking ?

Is it just thin drilled sheet aluminium, or has it got a real, thick heatsinking section(s) ?

Do you know its Degrees per watt rating ?
(if not, don't worry, I'm just being curious).

Since it is a hobby/quick PSU you are building (if I understand correctly), I guess you are not too bothered.

It is 8cm * 32cm and 5 mm think and covered with 2 cm long (vertical) fins which are also 5mm but get thinner.
It is connected to a metal plate of 3 mm thick and connected to heatsinks on either side. So it has some cooling capability but nowhere near what i think is required. It has a fan as well.

And i will be adding that over temperature circuit. It will just pull the output to 0 by pulling the control voltage to < 0V.

 

[SOFTengCOMPelec]

Looking at the applications, circuit.

Surely the diode (1N457 = Small signal diode) D2, is the wrong way round ?
Connected to Pin 8 of the op-amp (LM301A) = V+ (power) supply pin.

 

[William Gaatjes]

The circuit is really easy, the LM317K is limited to a current with a maximum of : 1.3V/33 Ohm = ~39mA when the PNP pass transistor has a fully saturated VBE.
So the LM317K hardly has to do some work. But worst case, it has to dissipate around 0,039 * 33.7 = ~1.2Watts for a 35V input.

The MJ4502 does all the heavy lifting. And the opamp pulls the adjust pin down when the current through R3 is getting higher than the current level is set to.

I just have difficulty seeing the output going to 30V when the adjust pin is always 1.25Volts lower than the Vout of the LM317.
The opamp is fed with +20V. It's output should not go higher than the output saturation voltage of VCC -1V (@10KOhm load). Add the diode drop and the led forward voltage, The highest output of the circuit should be 19+0.7+1.25+1.6 = 22,55V. Oh well, i am sure i missing something. And it will reveal itself when i have build the circuit.

 

[William Gaatjes]

Looking at the applications, circuit.

Surely the diode (1N457 = Small signal diode) D2, is the wrong way round ?
Connected to Pin 8 of the op-amp (LM301A) = V+ (power) supply pin.

It is a single opamp. Pin 7 is always the positive power.
Pin 8 is a compensation pin.

 

[SOFTengCOMPelec]

It is 8cm * 32cm and 5 mm think and covered with 2 cm long (vertical) fins which are also 5mm but get thinner.
It is connected to a metal plate of 3 mm thick and connected to heatsinks on either side. So it has some cooling capability but nowhere near what i think is required. It has a fan as well.

And i will be adding that over temperature circuit. It will just pull the output to 0 by pulling the control voltage to < 0V.

I assumed there was no fan. It sounds quite good (the heatsink).
The fan is a BIG help as well (usually).

The limiting factor then (as you seem to realize), is that a single transistor, can only pass a limited amount of heat.
Maybe (very approximately) 100 Watts (because of the various thermal resistances between the junction and the heatsink + heatsink temperature under load) of the approx 175 Watts (max), that would be needed. Before exceeding its maximum junction temperature.

Also it could be unreliable and/or have a short life, as for the best reliability etc, it should be reasonably far away (lower), than the maximum junction temperature.

 

[SOFTengCOMPelec]

It is a single opamp. Pin 7 is always the positive power.
Pin 8 is a compensation pin.

Sorry, my mistake.

I quickly opened another window with its datasheet.

Probably this one:
http://www.utm.edu/staff/leeb/LM301.pdf

Then looked at the LM101 (by mistake), which has V+ as pin 8.

Sorry.

This is why the low current limit setting of bench power supplies, can save me from blowing up, freshly designed/built circuits.

 

[SOFTengCOMPelec]

It is a single opamp. Pin 7 is always the positive power.
Pin 8 is a compensation pin.

You are perfectly correct, it is pin 7. I think some people get annoyed, because I usually insist on double checking where the supply pins are, to play safe.

But (8 pin DIL, Analogue) stuff like the LM358 (dual op-amp, pin 8), or LM311 (comparator, rather than an op-amp, pin 8) have pin 8 as the V+.
That is enough to force me to always look it up.

Annoyingly most digital logic, uses e.g. pin 7 Gnd, pin 14 +ve.
But the odd device here and there, is different. E.g. Supply pins in the middle.

Again, that forces me to (sometimes) double check. (To avoid unnecessarily blowing things up).

 

[William Gaatjes]

Sorry, my mistake.

I quickly opened another window with its datasheet.

Probably this one:
http://www.utm.edu/staff/leeb/LM301.pdf

Then looked at the LM101 (by mistake), which has V+ as pin 8.

Sorry.

This is why the low current limit setting of bench power supplies, can save me from blowing up, freshly designed/built circuits.

The LM101A is what i use, But the LM101W indeed has a different pinout.
I have to use the one in a TO-99 casing. :|

I got 4 coils now.

5 more to go.

 

[SOFTengCOMPelec]

The LM101A is what i use, But the LM101W indeed has a different pinout.
I have to use the one in a TO-99 casing. :|

I got 4 coils now.

5 more to go.

I'll try and keep quiet, until you have finished the remaining 5 coils. Else you may miss-count the number of turns.

I really like the TO-99 cased ICs. It makes a pleasant change, and they kind of look cute.

Amazingly, if you google search for TO-99, it mainly comes up with one web page.

DON'T LAUGH - this is the page. I guess it shows the immense usage of the case style. Around the world.

 

[William Gaatjes]

I'll try and keep quiet, until you have finished the remaining 5 coils. Else you may miss-count the number of turns.

I really like the TO-99 cased ICs. It makes a pleasant change, and they kind of look cute.

Amazingly, if you google search for TO-99, it mainly comes up with one web page.

DON'T LAUGH - this is the page. I guess it shows the immense usage of the case style. Around the world.

I just looked at the IC, and i made a mistake, it is a LM301A.
That is a bummer, the LM101A has higher maximum voltage of 44V. I might need that.

TO-99 indeed looks cute. But i prefer dip, because dip ic sockets are easier to get. And i prefer to use dip sockets on prototyping board, for it makes easy replacing of dil ic's when something goes wrong. To be honest, i am thinking of using a LT1490A or something similar. I have to order it, but it is rail to rail out and has also a maximum voltage of 44V. And the LT1490A is sort of indestructible.

 

[SOFTengCOMPelec]

I just looked at the IC, and i made a mistake, it is a LM301A.
That is a bummer, the LM101A has higher maximum voltage of 44V. I might need that.

TO-99 indeed looks cute. But i prefer dip, because dip ic sockets are easier to get. And i prefer to use dip sockets on prototyping board, for it makes easy replacing of dil ic's when something goes wrong. To be honest, i am thinking of using a LT1490A or something similar. I have to order it, but it is rail to rail out and has also a maximum voltage of 44V. And the LT1490A is sort of indestructible.

I got a bit mixed up, when I saw your can (TO-99).

I was thinking of the ones which I have got (if I remember correctly), which (are probably usually military spec ones, so that it can easily take the wider operating temperature I think), seem to have a standard DIL 8 pin like layout.
Like this:

Yes, changing the op-amp, to a higher voltage/better one, makes a lot of sense.

I wonder why they did not use an op-amp, which has the compensation circuitry built in. If there are technical issues why that was the case (to stop the overall circuit oscillating or something, by changing/shifting the phase relationships etc), it could effect you.
But I am probably over-analysing this. It should be fine (fingers crossed!).

Anyway, your choice of the new op-amp, looks like it is a nice/good one.

If it is in a 8 pin DIL socket, it is easy enough to change, later.

 

[SOFTengCOMPelec]

The gbw (gain bandwidth product), drops by about a factor of 10, with the op-amp you suggested.

From 1 (meg) LM301A, to about 100 KHz. probably because it is intended for things like accurate voltage measurement (hence the relatively low, offset voltage, and rail to rail inputs/outputs, etc). It MIGHT matter ?

If it had external frequency compensation, it would be even more, potentially compatible.

 

[SOFTengCOMPelec]

For future reference. (Obviously, you are using completely different parts, so Ignore this for now).

The 3 Amp (+1.2V to +33V) version of the (LM317K), is ONLY £0.672 each.
I.e. The LM350T.

From Farnell:

http://uk.farnell.com/fairchild-semiconductor/lm350t/v-reg-1-2v-to-33v-3a-to-220-3/dp/1417673

There are circuits available which show how to easily extend that to higher currents (using more of them, I think).

It potentially saves a lot of hassle.

But the upper voltage range of your transformer, could well be an issue.

But you can also put in a power transistor circuit (pre-regulator). Which lowers the voltage to a safe limit. Using a zener diode, resistor and power transistor (or even a power Darlington). (Obviously, the transistor would get very hot).

 

[SOFTengCOMPelec]

Sorry, please ignore my last post.

I'm just getting thrown, by the different approach to the way you are doing it.

If I was doing it. I would design it first, and make sure I am happy with the design. Potentially test it out to an extent.

Then order the parts, make test and finish it.

I'm happy to be open to other ways of doing things.

But because you/and-your-friend have already bought the bits, and your friend has already part built it. It limits my options, to suggest, changing anything.

Anyway, sorry again.

tl;dr
I'm too late to the party, so have to keep quiet about how things could have been done differently.

 

[William Gaatjes]

Sorry, please ignore my last post.

I'm just getting thrown, by the different approach to the way you are doing it.

If I was doing it. I would design it first, and make sure I am happy with the design. Potentially test it out to an extent.

Then order the parts, make test and finish it.

I'm happy to be open to other ways of doing things.

But because you/and-your-friend have already bought the bits, and your friend has already part built it. It limits my options, to suggest, changing anything.

Anyway, sorry again.

tl;dr
I'm too late to the party, so have to keep quiet about how things could have been done differently.

To be honest, that is also my way of doing it, first the design, then parts build on the breadboard , then build a complete prototype.
But when i got all this, it was supposed to be a quick soldering job. But the more i think about, the more i want to change everything . My friend is the non schematic type. He just starts and adjusts along the way. That is not my style. Hence that i am now doing a few quick fixes that do not cost a lot of money. At first i did not put much thought in it. I just want to get on with my own projects. But i also feel i cannot let my friend down.

But anyway, here is the transformer with 4 14.4V AC coils @6.44A. And one auxiliary coil of 9V 100mA. The auxiliary coil is made of thin insulated copper wire, but it can drive a 12V/20W halogen lamp. The voltage drops from 9,1V AC to 7,5V AC. It will be only supplying about 90mA, so it will work fine.

I just started adding turns and then fine tuned the 14.4V AC coils by measuring the voltage and adding turns until the coils have the same voltage. Then made a pair of coils by soldering them together. Since i always started winding in the same direction, the phase is always correct.

It does not look as neat as it used to be and i am only partially finished with the plastic covering. But it will suffice. Fun thing is that the copper windings, are not as tight wound along the core as they used to be. So whenever a lot of current is drawn, say 10A, the copper windings will rattle a bit.

 

[William Gaatjes]

I should note that i just got the parts he bought and he asked me to solder them for him on a prototyping board. It was all thought up without me being a part of it. I only came into the picture later. Hence the rather messy way of solving this.

 

[SOFTengCOMPelec]

I should note that i just got the parts he bought and he asked me to solder them for him on a prototyping board. It was all thought up without me being a part of it. I only came into the picture later. Hence the rather messy way of solving this.

The skills of doing it like, your friend does it, are still potentially very useful and good, in their own way.

E.g. Two days before mass production starts. Testing shows that there is a major problem, and solving it is going to be rather tricky, as it is way too late to make significant design changes and/or change the PCB (manufacturing is at too late a stage). They could be better (in theory), at suggesting modifications to fix the situation.

Sometimes it is fun, to just mess about with electronics stuff, and see what turns out.

But I think you are too close to the limits of the upper voltage range, as it was meant for 35V (according to your schematic). That could cause damage and/or excessive current to be drawn (zener diode like effect).

E.g. If the mains voltage fluctuates a bit, and temporarily rises up by +10%, or something. Unloaded, it could be too much for your input stage (consisting of the LM317K etc).

Also the 10A desired total output current, may be too much for a single pair of power transistors.

One solution, at this late stage. Would be to reduce the specification. By going for a max output current, less than 10 Amps. Which then can (if reduced enough), keep the output transistors, below their 200 degrees C (max). (Junction temperature. The external case temperatures, would be a lot less than this, because of the thermal resistances, such as junction to case thermal resistance etc)
(I prefer staying well below 200 deg C, as I worry about the life expectancy/reliability of the output power transistors).
If it is still not viable to put in additional power transistors, because the heatsink is already constructed and drilled out.

 

[William Gaatjes]

I know, everything is to close to the maximum ratings. I negotiated with my friend that i would come with a proper schematic. I will be using the LT1490A dual opamp for voltage and current regulation. So i decided to not use the LM317K. But the LM317K has a TO-3 casing. Just as the MJ4502. So, now i can use two transistors for each supply (It will still be a dual bench supply in one housing) . This reliefs the power transistors. Switching (with a relay and a comparator tracking the output set voltage) between 28.8V AC and 14.4V AC halves dissipation in worst case scenario's. And the dissipation is spread over two transistors. The LT1490 is fed with about 33V and the MJ4502 get the full voltage. I may not even need a negative voltage anymore.

I got a rough schematic that i have to calculate through. After that, i am going to simulate it in LTspice for as far as possible.
I think i will make use of sziklai darlington transistors. 2 Mj4502 in parallel with emitter resistors and one npn drive transistor. Maybe a BD911. I am not sure yet. This way, i can reuse the MJ4502 and this transistor configuration has less voltage drop. Oh well... Now i can use my creative side and produce the best possible within limits.

I am thinking of creating a setup similar to the LT3081. The LT1490 allows for this because of the superior high voltage performance. The setvoltage will also be monitored by the relay comparator, switching between 14.4V and 28.8V. The current control opamp will also be able to influence the set voltage. Lowering it when the output current is higher than the set current level. This way, the relay will switch automatically (The comparator will have some hysteresis of course).

 

[SOFTengCOMPelec]

I know, everything is to close to the maximum ratings. I negotiated with my friend that i would come with a proper schematic. I will be using the LT1490A dual opamp for voltage and current regulation. So i decided to not use the LM317K. But the LM317K has a TO-3 casing. Just as the MJ4502. So, now i can use two transistors for each supply (It will still be a dual bench supply in one housing) . This reliefs the power transistors. Switching (with a relay and a comparator tracking the output set voltage) between 28.8V AC and 14.4V AC halves dissipation in worst case scenario's. And the dissipation is spread over two transistors. The LT1490 is fed with about 33V and the MJ4502 get the full voltage. I may not even need a negative voltage anymore.

I got a rough schematic that i have to calculate through. After that, i am going to simulate it in LTspice for as far as possible.
I think i will make use of sziklai darlington transistors. 2 Mj4502 in parallel with emitter resistors and one npn drive transistor. Maybe a BD911. I am not sure yet. This way, i can reuse the MJ4502 and this transistor configuration has less voltage drop. Oh well... Now i can use my creative side and produce the best possible within limits.

I am thinking of creating a setup similar to the LT3081. The LT1490 allows for this because of the superior high voltage performance. The setvoltage will also be monitored by the relay comparator, switching between 14.4V and 28.8V. The current control opamp will also be able to influence the set voltage. Lowering it when the output current is higher than the set current level. This way, the relay will switch automatically (The comparator will have some hysteresis of course).

That sounds much, much better.

Also if it does not end up quite right. You will have much more scope to adjust/mess/alter it, to make it better.

The combination of doubling the number of power output transistors and (approximately) halving the differential voltage by switching secondaries, via the relay circuitry, should also make it much more robust and reliable, by up to about a factor of X4.
(i.e. Halves the max heat generated/dissipated AND shares it between 2 rather than 1 power devices, per PSU output stage).

I think application circuits give you the gist, of how to use their ICs, in specific applications.
But real life circuits, may need a lot of additional parts, to make it into a realistic and practical device.

E.g. Someone may want to add fuses, thermal fuses, output isolation switches, power on and CV/CI mode Leds, V and I meters, etc etc.

Some/all of the above, may NOT be included in the application circuit.

 

[William Gaatjes]

Well, i got an preliminary version here. I have to find a way to improve the performance of the current source.

R4 will be a potentiometer to calibrate the voltage set current to 3mA.
R7 will be a 10K potentiometer to set the output voltage.
R12 will be the current set potentiometer. But i have not calculated it yet.
Diode D4 blocks when the current control output is positive.

I need the -5Voltage after all in this first version to get the output to zero when the current control circuit becomes effective.
But i have one more trick up my sleeve for that.
If i exchange the inverting input and the non inverting input of the current control opamp, i might get it to work by using a mosfet with an open drain.
I then no longer need a -5V supply.

The transistors are not the ones i will be using but this is a first attempt.
The relay circuit i have not added yet.

LTspice is wonderful for trying things out. And later on by making an actual prototype, the wrinkles can be ironed out.

 

[SOFTengCOMPelec]

Well, i got an preliminary version here. I have to find a way to improve the performance of the current source.

R4 will be a potentiometer to calibrate the voltage set current to 3mA.
R7 will be a 10K potentiometer to set the output voltage.
R12 will be the current set potentiometer. But i have not calculated it yet.
Diode D4 blocks when the current control output is positive.

I need the -5Voltage after all in this first version to get the output to zero when the current control circuit becomes effective.
But i have one more trick up my sleeve for that.
If i exchange the inverting input and the non inverting input of the current control opamp, i might get it to work by using a mosfet with an open drain.
I then no longer need a -5V supply.

The transistors are not the ones i will be using but this is a first attempt.
The relay circuit i have not added yet.

LTspice is wonderful for trying things out. And later on by making an actual prototype, the wrinkles can be ironed out.

That's looking pretty good, for a first shot at it.

Ignoring the fact that you are going to alter/improve it anyway. I hope you don't mind if I say a few quick things about it, given the way it is at the moment.

The voltage control potentiometer (R7), can be slightly noisy and intermittent, for a tiny fraction of a second. Usually much less so, if it is a quality, multi-turn one. But even so.
Because you have a wiper, e.g. Metal, in contact with the resistive material. As it moves along it, it can briefly disconnect.

E.g. Think of worn out volume controls on some equipment, where it makes loud pops and noises, as you move the volume control up and down.

This brief (e.g. 3.5 millisecond) disconnection (which if you are lucky does not occur. But it can and might), would cause the output voltage to rise to maximum (ignoring the relay centre tap selector) voltage.

If there were brief 30V (approx), voltage spikes, for a few milliseconds, when it is set to say, +5V for a PIC project. It could easily break your external devices.

In the circuit you have got now, putting in a capacitor to minimise that effect, would mess up the constant current circuitry. But there are ways of doing it, which I will leave you to consider.

Old single turn, carbon potentiometers, were notoriously bad at doing this. So I stand corrected, if modern, quality ones, don't need this capacitor and/or other solutions.

I think expensive equipment in general (VERY top end), may use optical turn detection devices (going WAY over the sophistication of your project, and WAY too expensive/complicated). Which largely eliminates that problem. Since a microcontroller, simply watches the up/down pulses (or whatever), from the optical turn device.
There are other ways of doing it.

Q3 2N5550, may be in danger of having its Emitter/Base junction, damaged by exceeding the 6V maximum, "reverse" voltage, applied to it.

E.g. Imagine the output is peacefully at exactly 15 Volts.

There would be 15 Volts, almost directly connected to Q3's Emitter. So if the Op-amp U1, suddenly turned off (ok, swings towards -5V/negative rail). E.g. The output current, suddenly exceeds the set limit.

U1's output would drop to about -5 volts (rail to rail), which connects to the base (via a resistor) of Q3.

So Q3 would (briefly), be at 21 Volts between the Base and Emitter, in the "wrong" direction. As well as being more than the +6V allowable on the datasheet. It would probably rapidly, or sooner or later, damage/destroy the device.

If it was me, I would probably put in a small signal diode, to eliminate the problem. Being careful to NOT effect other circuit functions.

 

[K7SN]

Wow - what a great thread. Wish I had something to add; I learned tube theory and passed the stupid 1st and 2nd class licenses just as RTL logic came about by mentally solving the problem using tube knowledge then applying it the simple transistors of that day. 1970 an electrical engineer could only get a job where you would say "would you like fries with that" so I took my math knowledge, 4 years of the GI Bill and became a computer scientist. Haven't built anything with transistors since about when with a 555 timer chip (least that what I remember it was called). sometime in the late 70s or early 80s. .

Don't let me interrupt your wonderful discussion; just wanted to know that someone else is enjoying your thread.

 

[SOFTengCOMPelec]

Redacted.

 

[William Gaatjes]

That's looking pretty good, for a first shot at it.

Ignoring the fact that you are going to alter/improve it anyway. I hope you don't mind if I say a few quick things about it, given the way it is at the moment.

The voltage control potentiometer (R7), can be slightly noisy and intermittent, for a tiny fraction of a second. Usually much less so, if it is a quality, multi-turn one. But even so.
Because you have a wiper, e.g. Metal, in contact with the resistive material. As it moves along it, it can briefly disconnect.

E.g. Think of worn out volume controls on some equipment, where it makes loud pops and noises, as you move the volume control up and down.

This brief (e.g. 3.5 millisecond) disconnection (which if you are lucky does not occur. But it can and might), would cause the output voltage to rise to maximum (ignoring the relay centre tap selector) voltage.

If there were brief 30V (approx), voltage spikes, for a few milliseconds, when it is set to say, +5V for a PIC project. It could easily break your external devices.

In the circuit you have got now, putting in a capacitor to minimise that effect, would mess up the constant current circuitry. But there are ways of doing it, which I will leave you to consider.

Old single turn, carbon potentiometers, were notoriously bad at doing this. So I stand corrected, if modern, quality ones, don't need this capacitor and/or other solutions.

I think expensive equipment in general (VERY top end), may use optical turn detection devices (going WAY over the sophistication of your project, and WAY too expensive/complicated). Which largely eliminates that problem. Since a microcontroller, simply watches the up/down pulses (or whatever), from the optical turn device.
There are other ways of doing it.

You are right, In V2 i already added the capacitors because i had the same idea.
I tried it with the mosfet, but i think the loop gain becomes unstable. The mosfet has a certain amplification as have all stage and some phase shift.
So in the simulation it started to oscillate. I will be going to a more tradional setup. The voltage and current opamp will both get diodes in their outputs.
Both outputs will be like an analog or port. The output with the lowest voltage will win.
And a current source will deliver a 10mA current for the drive transistor since the opamps only can sink current when the diodes are added..
Both opamps will be able to pull the base of the npn drive transistor to a point lower than is needed to conduct. I can then stop using the negative power supply. The comparator that will switch the relay will sense the output instead directly and switch with hysteresis. This is a more tradional setup. But it will work more reliable.

Q3 2N5550, may be in danger of having its Emitter/Base junction, damaged by exceeding the 6V maximum, "reverse" voltage, applied to it.

E.g. Imagine the output is peacefully at exactly 15 Volts.

There would be 15 Volts, almost directly connected to Q3's Emitter. So if the Op-amp U1, suddenly turned off (ok, swings towards -5V/negative rail). E.g. The output current, suddenly exceeds the set limit.

U1's output would drop to about -5 volts (rail to rail), which connects to the base (via a resistor) of Q3.

So Q3 would (briefly), be at 21 Volts between the Base and Emitter, in the "wrong" direction. As well as being more than the +6V allowable on the datasheet. It would probably rapidly, or sooner or later, damage/destroy the device.

If it was me, I would probably put in a small signal diode, to eliminate the problem. Being careful to NOT effect other circuit functions.

The transistors i will not be using. They where available in the LTspice library.
I will have to calculate some stuff manually for the MJ4502 and the drive transistor i will use. The drive transistor i will use depends on the dissipation i will have to deal with. I will post an updated schematic when i got time to simulate. My goal is to not use the -5V negative supply.

 

[SOFTengCOMPelec]

Getting rid of the -5V (negative) supply, is good, if you can achieve it.

I'm glad it's progressing.

Looking forward to seeing your later, updated results.