Switching Pre-regulator - GP Power Supply

[amspire]

As promised, I will post details of the progress of an isolated switching regulator to use with the General Purpose Power Supply design.

https://www.eevblog.com/forum/projects-designs-and-technical-stuff/general-purpose-power-supply-design-7488/

It seems to be becoming a project in its own right, and there is no reason it couldn't be used with other low dropout regulator circuits, so I am making it a new thread.

Here is the test circuit I am using that is based around a standard IC - the UC3842.  Not perfect for the final application, but it is fine to test out the other components in the circuit.



The problem is not getting it to work - I had it working fine over a week ago.

The problem is choosing the specifications and parameters that best match the available parts in the market. The biggest problem is that there are cheap MOSFET and Schottky diodes with amazing specs as long as I keep the maximum voltage rating under 40V. If I go above 40V, the parts are more expensive, and the specifications are not as good.

The problem is if I go after a big range, then choosing a transformer that keeps the MOSFET voltage well under 40V means that the schottky diode has to handle way over 40V. If I keep the schottky diode voltage under 40V, then the MOSFET has to handle way over 40V.

Also I have to make decision about the input range. Ideally is would go from 3V to 20V, but this is unlikely to happen. For one thing, to output 25V at 1A with a 3V input, I would need over 10A of input current.  If I reduce maximum output current to 250mA at maximum voltage, then the current in at 3V is reduced to 2.5A which is a better value.

I will get into the test later, but first some notes on the test circuit. Th UC3842 only starts up at 15V and shuts down at 10V. I could go for the UC3843, but I prefer to find a modern device that does not need a current sense resistor. (They use the voltage drop across the MOSFET instead).

I would love to find a device with integrated switching FETS, but I think it is impossible unless I narrow the range specs a lot.

The secondary is fully isolated.  The feedback circuit keeps the switching regulator output at one IR LED drop + one V_be drop above the linear supply input reference. If the linear supply output drops below this (in current limit mode), the switching regulator will follow the output down. I do not want to drive the switching regulator from the output all the time, as I do not want it to increase the voltage if the load is supplying an overvoltage.

A second rectifier on the secondary captures the negative voltage swing of the transformer, as this is exactly proportional to the input voltage. So the micro running the supply (if you are using a micro) knows what the switching supply input voltage is, and can adjust the maximum currents to suit.  More soon ...

Richard.

[amspire]

I will go into the details of the circuit and the waveforms, but I had to put together a constant current load. It is almost essential for the job as I can set the current to, say 1.1A, and then vary the output voltage over the range I need for the linear supply input.

For anyone interested, here is my load:



The current control pot goes the wrong way - it is from a PC fan speed control, and one of the potentiometer leads was cut off.  I will probably get a 10 turn POT sometime to replace it.

It works very well. It is intended for up to 2A, but I could increase this by lowering the current sense resistor value. The IRFP640 will go up to 400VDC as long as I watch the power consumption.

I didn't bother with a switch - I just unplug the battery. It is built on a laminated bakelite base so it is pretty solid.

I haven't bothered with a display - it is easy adding a multimeter in series with the load.

Richard.

[alm]

One neat feature I've seen on an ELV dummy load design was an external modulation input. I'm not sure about its bandwidth (probably fairly limited), but this is a nice way to test transient response. Just set it to the max output current of your power supply and feed it a 10% to 100% square wave at 1 kHz or so and observe any ringing in the output voltage.

An interesting test would also be to set the max current beyond the current limit on your power supply, and see how the linear/SMPS combination responds to switching between constant voltage and constant current with a very low voltage drop. I imagine the linear regulator might get fairly hot if the flyback supply is unable to respond fast enough. The linear regulator would also have to discharge the capacitors so many times per second.

[amspire]

One neat feature I've seen on an ELV dummy load design was an external modulation input. I'm not sure about its bandwidth (probably fairly limited), but this is a nice way to test transient response. Just set it to the max output current of your power supply and feed it a 10% to 100% square wave at 1 kHz or so and observe any ringing in the output voltage.

I have been thinking of some means to do transient switching, and I still have 3  spare LM324 opamps. I was thinking of just a switch, but a repetative cycles allows the use of an analog scope. Not a bad idea. I hadn't thought of that.

An interesting test would also be to set the max current beyond the current limit on your power supply, and see how the linear/SMPS combination responds to switching between constant voltage and constant current with a very low voltage drop. I imagine the linear regulator might get fairly hot if the flyback supply is unable to respond fast enough. The linear regulator would also have to discharge the capacitors so many times per second.

Yes, that is about the worse case load for the power supply. It is possible that with just the right duty cycle, the switching regulator can be still pretty much full volts, and half the time the linear regulator is near zero volts at full current. That would mean instead of dissipating 4W, it is dissipating over 10W. If this is the case, I may be forces to add thermal protection to the linear regulator.

I have seen real loads actually behave exactly like this.

For example you are debugging a faulty switching regulator that has a delayed startup. It tries to start and a shorted turn in the inductor (or similar) overloads the power supply dragging the voltage down. When the regulator's minimum voltage is reached, it turns off and waits a period before trying to turn on again.

In the micro controlled version of the linear regulator, it would be possible to add some code to monitor for this, but the potentiometer controlled version needs to be safe as well.

I can actually work out the worse case conditions. If the switching regulator has a 220uF output capacitor. Then if the linear regulator is set at 25V output voltage and the load is shorted for about 0.22 mS and then goes back to normal for about another 0.22 mS in a repeating cycle, you could get as much as 15W dissipated. I may be forced to add a thermal foldback on the current limit circuit.

Definitely I will have to think about that. Dave's supply will have exactly the same problem. It is easy to think that the LT3080 that Dave is using has thermal protection so it will be fine, but this is not the situation that protection was designed for. I wouldn't like to bet on the reliability of the LT3080 when the IC junction is at 10 deg C above the absolute maximum temperature rating for an extended period (as happens in thermal limiting).

It is a rare situation, but it can happen, and the supply needs to be able to survive it. I will have a linear pre-regulator option (needing a big heatsink) and it will not have this problem at all.

Can I reduce the size of the capacitors on the switching regulator output by going to low ESR 10uF ceramic caps? Do I deliberately slow down the switching regulator's rise time after a current limit episode? More questions.

Richard.

[Rerouter]

hmm, would there be some way to make use of the feedback to the switcher to lower its voltage so that while shorted, e.g. 3V, so that the linear section is only dissipating ~3W,

i would have to think it would somehow tie in with the constant current mode, but my mind hasnt quite woken up enough to play with the schematic.

[amspire]

hmm, would there be some way to make use of the feedback to the switcher to lower its voltage so that while shorted, e.g. 3V, so that the linear section is only dissipating ~3W,

i would have to think it would somehow tie in with the constant current mode, but my mind hasnt quite woken up enough to play with the schematic.

That is what will happen, but the problem is that the switches has a reasonably big output cap, and if the overload time is short enough, the voltage down not have time to drop on that capacitor. The Switching regulator can raise the voltage on the filter capacitor, but it cannot lower the voltage - the only thing that can lower the voltage is the current drain into the linear regulator circuit. At 1A output, this means the maximum discharge rate of a 220uF capacitor is 1v per 0.22mS. So if the capacitor is at 27V and the linear regulator is shorted for 0.22mS, the voltage only has time to drop to 26V. If a short is applied and removed at a 2KHz rate while the output voltage is at 25V and the current limit is at 1A, the voltage on the switching supply capacitor will only vary between 27V and 26V.

The two things that usually sets the size of the capacitor on the switching regulator output is you want the noise on the output to be low, and the capacitor has to be big enough to handle the large ripple current from the flyback transformer output.  If I can get low ESR ceramics handling most of the ripple current instead, then perhaps I can get away with much less capacitance, that means less chance of this problem occurring.

Richard.

[Rerouter]

or like you mentioned before, if the switching frequency is increased, a lower value cap can be used, though i do not know what difficulties you may find with that, both a chip capable and if your 7c toroids can handle higher frequencies, i suppose if those 2 items where compounded it could help to lower the pulsed short loading :/

[amspire]

or like you mentioned before, if the switching frequency is increased, a lower value cap can be used, though i do not know what difficulties you may find with that, both a chip capable and if your 7c toroids can handle higher frequencies, i suppose if those 2 items where compounded it could help to lower the pulsed short loading :/

At high voltages out with low voltages in, the ripple current can be very high. If we go to extremes, with 3V in and 30V out with a 1:1 transformer, the current pulses in the secondary are 10A for 10% of the time at best. It could be as high as 20A peak if I use a low inductance transformer. The problem is finding a small cap that can handle the ripple current. A small electrolytic, but several ceramic caps in parallel might. It is not just the ESR - you also want to smooth those 10A pulse out into something like 20mV of noise, so you do not want the capacitance too low.

The 10A pulsed current is 3.1A RMS. The 20V peak current with the low transformer is probably about 4-5A RMS ripple. These are extreme values, but I definitely can get half this. If I go for the best low ESR packages in the 50VW rating, I can just get a useable ripple handling at about 270uF. My test board is using a 470uF cap right now.

Note that a buck converter does not have this kind of problem, as you can get a pretty constant current supply coming from the transformer. This is a problem that I get trying to isolate the supply.

The joys of switching power supplies.

Richard.

[Rerouter]

hmm, been digging about, without a transformer it would be ~2A RMS, been digging about how about a charge pump based isolator? i can assume there is problems with transient response and some additional cost, but just trying to think of other ways around the problem :/

[Zero999]

You could use an unregulated DC-DC converter to provide isolation from the mains, then add a switching regulator to the secondary side.

An example is given on the TC33167 datasheet. You should be able to get the parts by hacking an electronic halogen lamp transformer.
http://www.datasheets.org.uk/dl/Datasheets-310/85248.pdf

[alm]

Definitely I will have to think about that. Dave's supply will have exactly the same problem. It is easy to think that the LT3080 that Dave is using has thermal protection so it will be fine, but this is not the situation that protection was designed for. I wouldn't like to bet on the reliability of the LT3080 when the IC junction is at 10 deg C above the absolute maximum temperature rating for an extended period (as happens in thermal limiting).

In Dave's supply it would be easy to fix in software. Detecting it and shutting the supply down until the pass transistor has cooled would be an acceptable solution, in my opinion. The user could just switch to a traditional, big, robust linear supply. Not should how to do this in the purely analog design, however, possibly just some way of temperature sensing near the pass transistor?

[amspire]

Definitely I will have to think about that. Dave's supply will have exactly the same problem. It is easy to think that the LT3080 that Dave is using has thermal protection so it will be fine, but this is not the situation that protection was designed for. I wouldn't like to bet on the reliability of the LT3080 when the IC junction is at 10 deg C above the absolute maximum temperature rating for an extended period (as happens in thermal limiting).

In Dave's supply it would be easy to fix in software. Detecting it and shutting the supply down until the pass transistor has cooled would be an acceptable solution, in my opinion. The user could just switch to a traditional, big, robust linear supply. Not should how to do this in the purely analog design, however, possibly just some way of temperature sensing near the pass transistor?

It is hard to do a reliable fix in software without some extra hardware to detect the problem.  The best solution is probably to add temperature monitoring as then you can avoid taking any action until the temperature reaches a dangerous level. Temperature monitoring is probably a good idea anyway.

The output capacitor sets an upper limit to the frequencies of the switching load that can cause the problem - a 10uF cap means that the problem goes away above about 20Khz. A big enough capacitor would make the problem go away totally, but one of the advantages of a linear supply is you can use a small output capacitor which is is often a good thing. If you know you have a switching-type load like a PWM motor control or something, you can add a 2200 uF cap across the supply outputs but sometimes you do not know what a load will do until you try powering it up.

How far you go is a matter of design choice and that is often the difference between buying cheap supplies and the really good supplies. In my mind, it is the job of a pwoer supply not to blow up under a difficult load, and the cost of the parts to protect the supply is a tiny fraction of the cost you could cause in damaging a load after the regulator IC or transistors short sending maximum volts to the load.

Richard.

[BravoV]

Richard, how did you wind that transformer ? Any specific values on that both identical windings ?

Btw, I assume you're still working on the linear part and have you made the decision either using mosfet or bi-polar for the high current pass element ?

Again, thanks for all your effort & time making this to this forum.

[amspire]

Richard, how did you wind that transformer ? Any specific values on that both identical windings ?

I went for a 1:2 primary"secondary ratio winding for a few reasons. One was I wanted to be able to use a 40V mosfet. I found one that is about 20C - Mouser had over 1,000,000 in stock. 10 milliohm on resistance. If I have to go for higher voltages the price doubles and they are not nearly as good.

The 1:2 winding ratio does mean that I need an 80V or more schottky in the secondary which again costs double the 30/40V mosfets, but I couldn't keep both parts at 40V with a flyback switching circuit.

Now I am not sure I will succeed with an isolated switching circuit and if I do not, I will have to revert to the much easier Boost/Buck converter without the isolation. The reason I wanted the isolation is so I could attach several boards to the one power source and end up with several different supplies. For example, you could use a standard PC supply. The supplies can be run in master slave mode in parallel or series.

The problem is if I can get the flyback converter efficient enough and quiet enough. Initial efficiency was about 75 %, but that was with non-ideal parts. 85% efficiency would be a good result.

Btw, I assume you're still working on the linear part and have you made the decision either using mosfet or bi-polar for the high current pass element ?

Believe it or not, I have a design with a PNP pass transistor fed from a current source on the opamp output.  It is my Mark III design and I think it is a lot better then the previous two. I meant to get back to it sooner but I was hoping to come up with a full design instead of another partial one. If I cannot wrap up the switching supply soon, I will get back to the main regulator board very soon.

The actual choice of the regulator "power transistor" is a surprise. Let's just say it is very cheap, about as common as a transistor can be, but not what you would expect at all. Can be discreet of SMD mount with no heatsink.

Richard.

[fmaimon]

Richard,

Any news on the switching pre-regulator?

While googling around, I found this: http://www.candlepowerforums.com/vb/showthread.php?61748-Power-Supply-with-Switching-Pre-Regulator&p=709927&viewfull=1#post709927


The switching pre-regulator have some good ideas. First, it use a "Simple Switcher" from National (now TI), as it does everything, except isolation.
The drop voltage is "regulated" by the Vbe of the 2N4403 plus the 1N5226 (3.3V) zener diode. The 1N4148 seems to be protecting the PNP for reverse breakdown.
There is another protection circuit built around a 2N4401 and another zener. Quoting the designer:

The new parts (incldg. 2N4401) is to swamp the FB pin in case the neg supply goes away before the pos supply ...  This will kill the output voltage
 in case input power goes down while powering something critical, so the output wont jump up unexpectedly.

You don't seem to be using a negative supply in your design, but if you do, here is an idea.

For isolation, it may be better to use a single 24V power supply from ebay for each unit. They are cheap at about $15 each.

[Mechatrommer]

just notice this thread. is there anyway to make the picture in the first post bigger without "digital resampling"?

[amspire]

I was trying to make a very cheap isolated switching regulator, and hit a bit of a brick wall.

It was just very hard finding an economical way to make a transformer using easy-to-get cheap core and former parts.

The solution I was hoping would work was to use powered metal toriods - they type you can recover from old PC power supplies. The cores can be purchased at prices as low as 5 to 10 cents each which sounded great.

The problem was that I found that leakage inductance was a big problem if I tried to insulate the secondary winding. The only practical solution seemed to be to use some fairly exotic and expensive 4 layer polymide (Kapton) enameled wire for one of the windings (1500V AC rated insulation) to allow the close proximity between the windings, or alternatively go to a much bigger core running at a much lower frequency (like 20-40KHz). The leakage inductance still looked big enough to cause more secondary noise then I wanted without the pre-regulator becoming far more expensive then the supply module.

I would probably have to find a commercial transformer, but a fairly brief search didn't find anything with a great supply source and price. Without an insulated pre-regulator, many of  my design options for using multiple supply modules disappeared.

A non-insulated pre-regulator is considerably easier to design, as you do not need to worry about leakage inductance, but the idea does not excite me at all.

I have several other projects going on that are of more immediate interest, and this one has more or less been put to the side until I get some inspiration.

Richard.

[T4P]

You can also pick a same size core but i see most of the PC power supplies run at 1MHz at least ... nowadays 

[amspire]

You can also pick a same size core but i see most of the PC power supplies run at 1MHz at least ... nowadays 

PC power supplies can get custom designed transformers using the best core for the job. The supply of cores and hardware to the hobbyist is really bad.

If someone can tell me a supplier with a good range of cores, at proper prices, please let me know. I do not think Mouser, Digikey, Element 14, etc want to know about ferrite cores.

It is not a matter of a supplier selling a small range of cores. It is about a supplier selling the full range of cores so that the hobbyist can buy exactly the same core that the designer specifies. I hate it when I see a pair of small E core worth under 10 cents selling for $2 to $5, and on top of that, they then ask you to pay a dollar or two for the bobbin.

Now about the 1MHz operating frequency, the problem there is also that the transformer has to be constructed perfectly. This can be extremely hard for the average hobbyist to do. I have had transformer designs subcontracted out to transformer manufacturers to supply, and it often takes these specialty companies with all the right equipment several attempt to get the transformers perfect. If you make a 1MHz flyback transformer, and the leakage inductance is 50% more then it should be (very easy to do if you are hand winding the transformer), that may kill the switching transistor or cook the snubber components.

Richard.

[robrenz]

Richard, have you ever visited this site?  http://www.surplussales.com/Inductors/FerPotC/FerPotC-2.html

[HLA-27b]

Some useful reading material about transformer cores:
http://www.mag-inc.com/design/technical-documents#



A few from within the above place:

Inductor Design in Switching Regulators
www.mag-inc.com/File%20Library/Product%20Literature/General%20Information/sr-1a.pdf



Cores for Switched–Mode Power Supplies
www.mag-inc.com/File%20Library/Product%20Literature/General%20Information/sr-1a.pdf



Magnetic Cores for Switching Power Supplies
http://www.mag-inc.com/File%20Library/Product%20Literature/General%20Information/ps-02.pdf



A Critical Comparison of Ferrites with Other Magnetic Materials
http://www.mag-inc.com/File%20Library/Product%20Literature/Ferrite%20Literature/cg-01.pdf