Creating a custom voltage on a toroidal transformer

[William Gaatjes]

This is a more traditional setup, the integrators are no longer connected in series with each other as in the previous schematic. This improves stability.

Now i have created an analog or port. The opamp with the lowest output voltage wins by making use of blocking diodes. Of course, the opamps outputs can now no longer source current, only sink current. To solve that issue, i have added a 12mA current source. The transistor Q5 will be a different model. Because in worst case scenario it has to dissipate around 400mW of heat. I need a transistor there that can handle that. The output stage transistors i have to change to different models as well. The output will be the MJ4502. But i will need a 60V VCE, NPN transistor with a fair amount of HFE gain . Because the 12mA current has to be amplified to a maximum of about 5A at least. But i would rather go for a 7.5A. To have a little headroom.
That is a total HFE gain of about 625. Now the HFE of the Sziklai darlington transistor is the product of the gain of the MJ4502 and the NPN transistor. If i can get the HFE of the NPN transistor to be between 50 and 100, i should be set. The HFE of the MJ4502 is between 25 and 100 at 7.5A. A higher total HFE gain will put less stress on Q5.

The opamp circuit no longer needs the -5V because the opamps are rail to rail output opamps. And the opamps will get a separate stabilized 32 to 33V supply. The control circuitry will be much more stable this way.

The toroidal transformer will supply the output transistors with voltage.

 

[William Gaatjes]

I will also be adding protection diodes to the inputs of the opamps and a reverse current diode on the output.
And i will not be using the bAT54 since it is a 30V diode.

 

[William Gaatjes]

This is an example of when to use normal diodes and when to use schottky diodes. Schottky diodes have a high reverse leakage current and can cause problems with the opamp circuits. So i used standard silicon diodes instead. The added diodes protect the inputs of the opamps if something inductive like a motor would be connected.
I added a diode to the output to prevent negative voltages from damaging the supply.

This circuit should be pretty okay.
Now i have to find the right NPN transistor for Q3.
A HFE between 50 and 100.
A VCE of 60V minimum.
Should be able to handle a watt dissipation at least, so a TO-220 casing will be needed.
R7 will be a 10Kpotentiometer.
R12 will be a 2K potentiometer.

 

[William Gaatjes]

I do feel, i have to think the current measuring circuit through again.

 

[SOFTengCOMPelec]

I do feel, i have to think the current measuring circuit through again.

Redacted.
Mistake.

EDIT:
If I'm looking at the correct datasheet. Those op-amps, are "micropower", ones.
As a result (probably because they are micropower), they have a really slow slew rate, only only about 30 millivolts, or so per microsecond. (Usually slew rates would be something like x20 faster, if not x100 on better ones).
They may/could work out ok. But I would worry that their transient response would be too slow.
E.g. At low current limit settings, a much higher current would flow, for a while, before it brought it under control. Which is undesirable, but not the end of the world.

Is there some reason why you are using micropower (usually for giving very long battery life) ones ?

 

[William Gaatjes]

Redacted.
Mistake.

EDIT:
If I'm looking at the correct datasheet. Those op-amps, are "micropower", ones.
As a result (probably because they are micropower), they have a really slow slew rate, only only about 30 millivolts, or so per microsecond.
They may/could work out ok. But I would worry that their transient response would be too slow.
E.g. At low current limit settings, a much higher current would flow, for a while, before it brought it under control. Which is undesirable, but not the end of the world.

Is there some reason why you are using micropower (usually for giving very long battery life) ones ?

These opamps have excellent characteristics with exception indeed for the slew rate specs. But that is where the output capacitor is for, to catch transients. An advantage of "slow" opamps is that the circuit is more stable, meaning it has a less tendency to oscillate. Overall, the LT1490 / LT1491 has very fine specs and is pretty precise meaning low offset. It's output can sink and source up to 20mA and is a rail to rail output. The input stage is also rail to rail and allows for voltages to be higher then the opamp supply with out strange side effects. The LT1490 / 1491 can handle a supply voltage of up to 44V.
All these specifications made me decide to use this opamp.
I will be using the LT1491 because then i have 4 opamps in a DIL 14 casing.

I did not like the fact that the current setting is dependent of the output voltage.
So i created a difference amplifier U2 and the output of the difference amplifier is fed to an integrator U3 which compares the output current with a voltage set by (R6+R12) current set.
The difference amplifier has an amplification of 1.
The voltage on the positive input of the integrator U3 is a maximum of 250mV. 250mV / by 0.05 Ohm is 5A.

 

[William Gaatjes]

I may change the current configuration. I am thinking of letting the difference amplifier amplify the difference voltage over R11, 10 times. To give a better current setting accuracy in the lower region. I would then have to adjust the current setting resistor ladder to go from 0V to 2.5V.

 

[SOFTengCOMPelec]

These opamps have excellent characteristics with exception indeed for the slew rate specs. But that is where the output capacitor is for, to catch transients. An advantage of "slow" opamps is that the circuit is more stable, meaning it has a less tendency to oscillate. Overall, the LT1490 / LT1491 has very fine specs and is pretty precise meaning low offset. It's output can sink and source up to 20mA and is a rail to rail output. The input state is also rail to rail and allows for voltages to be higher then the opamp supply with out strange side effects. The LT1490 / 1491 can handle a supply voltage of up to 44V.
All these specifications made me decide to use this opamp.
I will be using the LT1491 because then i have 4 opamps in a DIL 14 casing.

I did not like the fact that the current setting is dependent of the output voltage.
So i created a difference amplifier U2 and the output of the difference amplifier is fed to an integrator U3 which compares the output current with a voltage set by (R6+R12) current set.
The difference amplifier has an amplification of 1.
The voltage on the positive input of the integrator U3 is a maximum of 250mV. 250mV / by 0.05 Ohm is 5A.

I'm jumping to conclusions TOO quickly, and have not fully woken up yet, here.

The output capacitor(s), if present, would significantly slow down, the over/under shooting I was worried about.

It is amazing how complicated and detailed, designs like this can get.

Slow and stable/accurate is good.

 

[William Gaatjes]

Now the voltage on the positive input of integrator U3 can swing between 0V and a little more than 2.5V. The difference amplifier amplifies the difference voltage over R11 10 times. I think i will leave the circuit like this for now. And think it over later on. If you have any input, i will be glad to read it.

 

[William Gaatjes]

I'm jumping to conclusions TOO quickly, and have not fully woken up yet, here.

The output capacitor(s), if present, would significantly slow down, the over/under shooting I was worried about.

It is amazing how complicated and detailed, designs like this can get.

Slow and stable/accurate is good.

The circuit seems complicated because of the added protection diodes. If you remove them, the circuit is surprisingly simple. two integrators and one difference amplifier is all what it is. From the LT3081 basic schematic diagram, i copied the idea to use a current source to create the set voltage. But to be honest, it is no longer needed since the 33V will be a stabilized voltage, i can just use a resistor ladder. So the 3mA current source , i am going to remove again to save costs on the BOM (Bill of Material).

 

[SOFTengCOMPelec]

The circuit seems complicated because of the added protection diodes. If you remove them, the circuit is surprisingly simple. two integrators and one difference amplifier is all what it is. From the LT3081 basic schematic diagram, i copied the idea to use a current source to create the set voltage. But to be honest, it is no longer needed since the 33V will be a stabilized voltage, i can just use a resistor ladder. So the 3mA current source , i am going to remove again to save costs on the BOM (Bill of Material).

Absolutely, IT WAS NOT any criticism of YOUR circuit.
I was referring to power supply design in general.

Yours is turning out just fine, so far.

I'm being a trouble maker (sorry), by looking at it and saying stuff like, why is that such a slow device, etc etc.
Sorry.

Over the years, by asking fellow engineers, why did you use such and such a value/component here ?
I have been able to improve/increase my knowledge of electronics.

 

[William Gaatjes]

This should be a good version for prototyping. The large electrolytic capacitor C8 and R22(100k) are to prevent transients during power up.

 

[William Gaatjes]

Absolutely, IT WAS NOT any criticism of YOUR circuit.
I was referring to power supply design in general.

Yours is turning out just fine, so far.

I'm being a trouble maker (sorry), by looking at it and saying stuff like, why is that such a slow device, etc etc.
Sorry.

Over the years, by asking fellow engineers, why did you use such and such a value/component here ?
I have been able to improve/increase my knowledge of electronics.

You are no trouble maker to me. When designing, it is human to always forget something. So any input is greatly appreciated.
I might find out after building the first circuit, that the opamps are indeed too slow. I sure hope not, though.

I now need to figure out how to get that independent 33V stabilized supply.

 

[William Gaatjes]

Well, i had an opamp left when using the LT1491 because it has 4 opamps in one package. So i decided to make a novel discrete voltage stabilizer on 33V with the circuit comprised of U4 and other components.

The opamp is first fed through R26. But when the voltage is high enough, the opamp U4 with transistor Q4 will start to regulate the 33V line.

It needs enough output capacitance of a minimum of 10uF , or it will become unstable. It seems to work. I connected a load which alternates between 10mA and 30mA. and the 33V output voltage is stable. I may exchange the 0,6V diode for a cheap but accurate voltage reference.

I have some insulated copperwire left with a diameter of about 0.1mm, so i am going to add another secondary coil to the toroidal transformer to create the needed 35V AC. But that coil will be independent of the other coils. It should work. I have to add about 40 to 50 winding turns.

EDIT : I do not know why, but the forum seems to scale down my pictures. They should be at full resolution of 1920 * 1080.
They are full scale at photobucket.com

 

[SOFTengCOMPelec]

It is normally STRONGLY recommended, to never use one of the op-amps themselves, to control its own power supply regulation.
Apart from the risks of datasheet violation (including transient effects, and what happens during first power up, breaking common mode range and/or exceeding output voltage limits).
It could easily have circumstances which will make it badly misbehave.

I can point you to at least one significant electronics text book. Which gives it as an example, of bad circuit ideas.

Even if you can meet all datasheet limits, and show it works perfectly. It does not sound like a good idea.

Power supplies to the op-amps, should be extremely robust and reliable. Getting the op-amp to control its own voltage regulation, can cause all sorts of weird things to happen.

 

[William Gaatjes]

Well, that can be an issue indeed. But i think R26 solves most of the issues.
The problem is which voltage comes first. With R26 and the use of a PNP transistor, The supply voltage comes first and the input rise in voltage later.

But i can also create a voltage stabilizer with a zenerdiode and an extra transistor.

 

[SOFTengCOMPelec]

But i can also create a voltage stabilizer with a zenerdiode and an extra transistor.

Much better idea.
Possibly doing it in two stages. The above (down to max of 36V approx. i.e. a safe zone for the standard regulator, which follows) and followed by a normal/cheap linear regulator, to make the 33V.
33V ones probably not available, but can usually use a zener (or sometimes a resistor), on the adjust (or whatever it is called, e.g. common) pin. Some datasheets, will explain it.

 

[SOFTengCOMPelec]

Well, that can be an issue indeed. But i think R26 solves most of the issues.
The problem is which voltage comes first. With R26 and the use of a PNP transistor, The supply voltage comes first and the input rise in voltage later.

But i can also create a voltage stabilizer with a zenerdiode and an extra transistor.

Part of why you generally can't/shouldn't, is because the op-amp, goes from fully +Ve, to -VE (or vice versa), with a tiny fraction of a milli-volt, change in inputs (when it happens to be switching). Because of the huge gain.

So if the op-amp was hovering between fully +ve, and -ve, it could change the power supply usage current, by 10 milliamps (as the output swings), or whatever. Since the supply would be going to both the supply pins AND the input(s), of the op-amp. This could now change the inputs of the op-amp, enough for it to change its mind again.

i.e. It would/could oscillate, under some circumstances. Not all of which you will necessarily realize and/or be able to test.

tl;dr
Best to design out such potential weaknesses, rather than try potentially risky stuff, just to save 1p on the components cost.

 

[William Gaatjes]

I will have a look in my Horowitz/Hill electronic bible and see what i can come up with. I have some idea's but it has been a decade at least since i calculated on that kind of circuits.

I am sure you heard of this ?
I am guessing you have a copy at home ?

https://en.wikipedia.org/wiki/The_Art_of_Electronics

 

[SOFTengCOMPelec]

I will have a look in my Horowitz/Hill electronic bible and see what i can come up with. I have some idea's but it has been a decade at least since i calculated on that kind of circuits.

I am sure you heard of this ?
I am guessing you have a copy at home ?

https://en.wikipedia.org/wiki/The_Art_of_Electronics

That's where I remember seeing it, under one of their examples of "bad circuit" ideas.

As regards, solutions.
Something like this (is what I meant):

If that is not good enough, you could follow it by a conventional 3 terminal voltage regulator (not shown in schematics). As it would have been reduced in voltage enough, to NOT exceed the voltage regulators max input voltage.

 

[William Gaatjes]

That's where I remember seeing it, under one of their examples of "bad circuit" ideas.

As regards, solutions.
Something like this (is what I meant):

If that is not good enough, you could follow it by a conventional 3 terminal voltage regulator (not shown in schematics). As it would have been reduced in voltage enough, to NOT exceed the voltage regulators max input voltage.

I was thinking along in the same way, but i wanted a negative feedback loop.
And i remembered a circuit but was not sure. So i looked in my Horowitz/Hill bible and found the circuit comprised of Q4 and Q6 and components :

Calculation is very easy, as it turns out.

 

[SOFTengCOMPelec]

I was thinking along in the same way, but i wanted a negative feedback loop.
And i remembered a circuit but was not sure. So i looked in my Horowitz/Hill bible and found the circuit comprised of Q4 and Q6 and components :

Calculation is very easy, as it turns out.

I like that circuit. It is a very good idea.
With not much more in the way of components or cost, you should get better regulation from it, than the resistor/transistor/zener circuit.

That circuit will be better at ignoring (rejecting) the ripple voltage, from the transformer/filter-capacitors.

 

[William Gaatjes]

By reducing C9 to 100nF, there is still some filtering and makes the circuit less susceptible to EMI noise but the overshoot during startup is gone.

 

[William Gaatjes]

In my simulation, it does not perform very well. Almost 600mV delta at the 33V line when using a load change between 10mA and 90mA at the 33V line. In reality current consumption will be around 30mA. But i do not like it.

 

[SOFTengCOMPelec]

In my simulation, it does not perform very well. Almost 600mV delta at the 33V line when using a load change between 10mA and 90mA at the 33V line. In reality current consumption will be around 30mA. But i do not like it.

That does not make a lot of sense. Assuming you have the raw input voltage high enough.

But anyway, I thought the 33V rail (without extensively analyzing it), was only going to use a few milliamps. Higher currents during switching transitions, would hopefully be coped with by decoupling capacitors etc.

Also 30mA seems rather high (assuming no load connected). Surely it should be more like around 12 milliamps ? (4K7) + output ones.

EDIT:
Actually probably less quiescent current. I can't easily make out (see) some of the resistor values, because they are too small and/or out of focus, when I blow it up.

EDIT2:
I correct myself. I can believe that it can change a fair bit. Because of movements, in the Hfe curves, as the output current (collector current), changes in the top transistor (Q4).

EDIT3:
That's why I would have put a tiny, TO-92 linear regulator in (or bigger case,, if too much heat dissipation), after the initial input voltage, is dropped to safe levels. Such as 36 or 37 Volts.