Please rate this circuit, a DC 0 to 500V 10-mA power supply

[BravoV]

Stumbled upon this circuit at EDN -> http://www.edn.com/design/power-management/4403836/Regulate-a-0-to-500V--10-mA-power-supply-in-a-different-way-



I have many of those quality linear based AC to DC adapters from cell phone charger back from 90's, really intrigued to build one.

Really aware of the dangerous when working with high voltage stuff, like current go thru the heart when handling it with two arms, or creepage distance and etc.

Pretty straightforward question, is this "circuit design" it self is good and safe enough ? Or any other EE comments are welcome as well.

TIA

[Bored@Work]

I have a general problem with such wide-range power supplies. It is just too easy to make a mistake when using them and instead of dialing in a lowish voltage dialing in a high one. I am more happy with supplies that can be adjusted from "dangerous" to "very dangerous", instead of from "harmless" to "very dangerous". So when I use one I know it is always dangerous. Anyhow, that is fixable in that design with one more resistor, but ...

There is very little in it in terms of active protection from overload, just a current limit. There is nothing in it in terms of active protection from catastrophic failure. For example, no over-voltage protection.

Isolation of the operator, i.e. you, relies on the DIY (meh ...) photo-coupler, the feedback resistor chain and not the least the second transformer T2 with C1 in front. There is noting in the low voltage part protecting you as a last resort if any of these isolations break. And think, for example, not only about the 500V, but what could happen if a kV or few kV are injected from the outside? Think inductive load.

I also miss basic protection of the transformer (fuse) if something in the low-voltage part goes wrong.

[penfold]

It looks like a rather nice design. The use of the two transformers at the input will give you a reasonable degree of protection (assuming that touching the terminals would be on an accidental basis). Adding some proper input protection circuits would take it a lot further.

Unless Q1 is designed with a descent of overhead in its safe operating area if it were to fail to short then you would lose all regulation.

For this kind of application (relatively low currents, regardless of voltage) I normally prefer shunt style regulators, in this case it would eliminate the need for the opto isolator.

The use of the opto-isolator in the feedback, whilst its quite neat, I would only use them as a last resort in a feedback loop, they tend to kill a lot of the feedback bandwidth, but in this application, sure it would work.

Now if the result is to make an AC/DC cell phone charger adapter - its not really the circuit for you, its a bit too novel for its own good,

[SeanB]

Very nice, like the isolation between drive and control element. If you cannot get the photodiodes the simplest way is to take some flat faced LED's and couple them together with black heatshrink tubing to exclude external light and use them like that. A simple optocoupler, though with a very poor current ratio, but it can be improved by using clear blue leds on the driven side and using a near UV led to drive it. Higher output voltage of around 3V means you can use 4 in series ( with a 100k load resistor to provide switch off) to drive a high voltage mosfet to get higher current rating. Slow turn on ( charging gate capacitance from high impedance source) but will respond fast to short or overcurrent though it will be slow to drop when turned down.

A good idea as well will be a 2A slow blow fuse in the first transformer secondary, and a neon indicator on the output to show high voltage is present, along with a 480V varistor across the output to clamp overvoltages. C1 should not be an electrolytic, but a 10uF 400V motor run/PFC capacitor, as the reverse voltage will destroy it in short order. It will need to be connected off board, and is available either with flying leads ( lighting PFC capacitor) or with spade terminals.

[Jay_Diddy_B]

Hi,

I saw this schematic a few days ago on the EDN website.

The traditional way of doing this is to rearrange the circuit so that the control circuit is connected directly to the positive output. The feedback resistor extends to ground.

The traditional circuit is quite similar to LM317 based circuits, where the error amp reference are floating. This removes the need for isolated base drive.

You could also consider using a SMPS, a boost or a flyback.

Jay_Diddy_B

[Jay_Diddy_B]

Hi,
Here is a LTspice model to illustrate the concept:



This shows the effect of adjusting the variable resistor in 500k Ohm steps:



Note: This circuit has only been modelled in LTspice. It has NOT been built and tested.

Jay_Diddy_B

[BravoV]

I have a general problem with such wide-range power supplies. It is just too easy to make a mistake when using them and instead of dialing in a lowish voltage dialing in a high one. I am more happy with supplies that can be adjusted from "dangerous" to "very dangerous", instead of from "harmless" to "very dangerous". So when I use one I know it is always dangerous. Anyhow, that is fixable in that design with one more resistor, but ...

Good idea, thanks, for sure I will not use it for low voltage usage, say like powering a MCU or a led , c'mon, that sounds so ridiculous and stupid.
Probably will set the feedback voltage divider to minimum at 100 V instead of zero volt , a 3 digits minimum voltage should looks scary enough.

There is very little in it in terms of active protection from overload, just a current limit. There is nothing in it in terms of active protection from catastrophic failure. For example, no over-voltage protection.

Assuming you're talking at output part, does putting a MOV across the output is suffice ?

And think, for example, not only about the 500V, but what could happen if a kV or few kV are injected from the outside? Think inductive load.

Again, does MOV at the output is enough ?

[BravoV]

It looks like a rather nice design. The use of the two transformers at the input will give you a reasonable degree of protection (assuming that touching the terminals would be on an accidental basis). Adding some proper input protection circuits would take it a lot further.

Yep, the 1st thing that interested me at this circuit is the double transformers isolation idea, really like it.

Unless Q1 is designed with a descent of overhead in its safe operating area if it were to fail to short then you would lose all regulation.

That Q1 Vce max rating is at 1500 volt, should be enough right ? Also I do have other options with higher voltage transistor if needed.

For this kind of application (relatively low currents, regardless of voltage) I normally prefer shunt style regulators, in this case it would eliminate the need for the opto isolator.

How is that look ? Mind do some scratch on the circuit ?

Now if the result is to make an AC/DC cell phone charger adapter - its not really the circuit for you, its a bit too novel for its own good,

As I answered BAW above, the minimum voltage will be at 100 Volt, and thanks for the concern though, if I need low voltage PSU, trust me, I'm fully covered and don't need this circuit. The one I have is more than enough, click My bench psu to get the idea.

[BravoV]

Very nice, like the isolation between drive and control element. If you cannot get the photodiodes the simplest way is to take some flat faced LED's and couple them together with black heatshrink tubing to exclude external light and use them like that. A simple optocoupler, though with a very poor current ratio, but it can be improved by using clear blue leds on the driven side and using a near UV led to drive it.

Thanks, great idea and I have lots of blue and UV leds at my disposal, can't wait to try this out. 

Higher output voltage of around 3V means you can use 4 in series ( with a 100k load resistor to provide switch off) to drive a high voltage mosfet to get higher current rating. Slow turn on ( charging gate capacitance from high impedance source) but will respond fast to short or overcurrent though it will be slow to drop when turned down.

Mosfet at the driver (low voltage side) ? Sean, please draw the circuit.

A good idea as well will be a 2A slow blow fuse in the first transformer secondary, and a neon indicator on the output to show high voltage is present, along with a 480V varistor across the output to clamp overvoltages. C1 should not be an electrolytic, but a 10uF 400V motor run/PFC capacitor, as the reverse voltage will destroy it in short order. It will need to be connected off board, and is available either with flying leads ( lighting PFC capacitor) or with spade terminals.

Yes, thats the plan for C1 & C2, I will use those big aluminium can cap that is rated at 1000 V, and for sure there will be no electrolytic caps at the high voltage part, except for the 7805 regulator.


Also I've thinking to add better security at the isolation part with this setup below, any comment ?
Got this idea using LM317 for AC regulation in the National (TI) LM317 datasheet, and I'm aware that the secondary winding at the left transformer needs higher voltage for the regulator drop out to reach the required voltage for the primary winding at the right/second transformer, but its not a big deal, at least for me.

[BravoV]

Hi,
Here is a LTspice model to illustrate the concept:

This shows the effect of adjusting the variable resistor in 500k Ohm steps:

Note: This circuit has only been modelled in LTspice. It has NOT been built and tested.

Jay_Diddy_B

Thanks Jay, really appreciate your help, but please explain further cause I'm a bit lost here, why you omitted the opto-isolator part in the feedback loop ? I thought the voltage divider (the high resistance resistor from RF1  downto RFG) which is the only part that is exposed to high voltage is scary enough, while your circuit is now making the low voltage part exposed at many points to high voltage. 

I'm not an EE, but please, what make your circuit is safer or better than the original one ?

Also I don't think a 5 Meg Ohm potentiometer is easy to find. 

[SeanB]

Mosfet in place of the BUZ transistor, use 4 blue leds in series then you will eventually  put enough charge on the gate to get it to conduct. Same current limit transistor and discharge resistor but the ability to have a higher current ( dissipation dependant on the mosfet of course, unless you use the mosfet to drive the bipolar as well). You can drive the 4 transmit units in parallel via a current sharing set of resistors so a 10V opamp supply will do. Using the lm317's as voltage.current limiting will not work, as there is no diode to bypass the regulator when it is off. A bridge rectifier with the 317 in the DC leads will work, but you will lose a lot of voltage across it as static loss.  Better to use just a fuse.

[Jay_Diddy_B]

Hi,

I have made a small modification so that the circuit no longer needs a 5 Meg Ohm Pot.
Here is the revised schematic:


The circuit also shows two loads resistors R5 and R6. R6 is in series with a switch. The load is stepped from 2.5mA to 5mA by the switch.
By using the stepped load like that we can observe the control loop stability.

All regulated power supplies are servo systems there is a reference that drives the demand, error amplifier and a feedback circuit.

Here is the results showing load current steps and the output voltage. You can see that the system is stable.



Opto-couplers present challenges in designing the control loop. The gain of the opto coupler, for a given part number, can vary from 50% to 200% of the nominal value. In addition the gain changes as the opto coupler ages. These may not be a concern if you are building one power supply for yourself, but if you are are planning on building several it is better to avoid opto-couplers.

Yes, the circuit is floating at the output voltage. I would strongly recommend finding a potentiometer with a plastic shaft and mounting all the control circuitry, including the body of the potiometer, inside the enclosure with only the plastic shaft sticking out.

There are many ways to do anything in electronics. I just wanted to show the forum members the traditional approach to high voltage power supplies. This idea of the floating bias supplies can be found in most power supplies greater than 20V. Below 20V emitter or source followers can be used to boost the output of an op-amp.

Regards,

Jay_Diddy_B (Not my real name, I am in disguise  )

[mikeselectricstuff]

On minor flaw is there is nothing to make sure the caps are discharged when powered off - some bleed resistors would be a good idea.

[BravoV]

@SeanB,

About the LED driven part, even 4 or even 5 in series using the right leds even properly shone, will they have beefy enough juice to supply the gate capacitance ?

This is really new for me and apart from this circuit, using this led technique to supply the mosfet's gate is now making my brain spinning & thinking to implement it at other purpose.   


@JDB,

Thanks for the circuit, truly appreciate it, and understand the advantage such as tight loop control and also has a really good load transient response, but again, your design will be 2nd option just in case the "led" thingy failed miserably. 


@Mike,

Simple yet important, I guess we all missed that part, thanks. 

[SeanB]

The gate is a capacitor, it will take a while to charge with the limited current, but it will charge up eventually. Not going to be a fast response, probably under 10Hz max switching speed.

[BravoV]

Wow .. 10 Hz !!