Lab PSU Design Idea/Question

[gnif]

I am currently looking at building a bench PSU that is capable of 0-30V DC 10A. Normally I know that a SMPS is not a great choice for this due to switching noise, but I was wondering if properly filtered could it still be viable?

See, the idea I have formed is to use an adjustable SMPS to feed linear regulator(s) at the required voltage + 1V, which then the linear regulators then handle the final drop to the required voltage. If this is done correctly would this improve thermal dissipation and help with the noise issues caused by a SMPS?

This will all be controlled via a MCU which will regulate the SMPS voltage to the linear regulator as well as the regulator itself. Looking at using a STM32 as I have a few laying around which contain 2x 12bit ADCs and 2x 12bit DACs (I know, overkill, but it is what I have on hand).

[c4757p]

I think a SMPS is a fine choice. I wouldn't use it for a nice, low-noise power supply, but do you really need a 10A supply to be low noise? I think the only things I've ever done that couldn't handle the noise from a well-built switcher were all under 1A.

A linear regulator after the SMPS would definitely be able to clean up some noise. You can also do a pretty good job cleaning the SMPS output with some additional L-C filtering, though you will have to work that into your SMPS calculations to get it nice and stable. And of course it would improve thermal dissipation, as even at 10A you'll be dissipating at most 10W in the linear section.

I don't think the 12 bit ADC/DAC is overkill. That gives you 10mV and 5mA resolution with a bit of room to spare for software calibration - about the same resolution you'd be able to easily dial in with a 10-turn pot. (One suggestion if you're going to use this for general purpose as well as high power - throw some switchable gain on the current sense ADC input so you can have better than 5mA resolution at low currents.)

[amspire]

It is one thing saying you can build a quiet 10A SMPS - it is another thing to successfully do it. It you build it, and there is persistent noise from the supply in your oscilloscope and other mains power devices, eliminating the noise can be extremely difficult.

The problem is often not the noise on the output, it is more often the noise induced into the mains and ground connections.

In general, linear regulators are not fast enough to handle the kind of noise a SMPS produces, but in a power supply, most of the high frequency noise filtering is done by an output cap.

If you are also considering a transformer based linear regulator approach, then if you can get hold of a good multi-tapped transformer, there is a power transistor arrangement used by HP and others that allows you to automatically switch between the transformer outputs without relays or mosfet switches. Around about 100W is a comfortable amount of power to get rid of and it is within the realm of a passive fanless heatsink. So that means something like taps for the 10V, 20V and 30V ranges. There is a cost to this method - you need separate rectifier diodes and caps for each transformer winding. The switching is instantaneous and so there is no evidence on the output to indicate you are using a multi-tapped transformer .

If you are interested in a linear design, I can describe the method HP uses.

Richard.

[gnif]

It is one thing saying you can build a quiet 10A SMPS - it is another thing to successfully do it. It you build it, and there is persistent noise from the supply in your oscilloscope and other mains power devices, eliminating the noise can be extremely difficult.

Never said I could , just that I want to use it as a project to learn more about the analogue side of things, and it is a bit of gear that I require. Yes, I would be very interested in how HP do this, thanks for offering the information.

[krish2487]

On the other hand,

Armspire volunteering to explain the method behind implementing is worth listening to!!!

If you are interested in a linear design, I can describe the method HP uses.

I vote for a "armspire's words of wisdom" as a seperate section in the forums...

I know, the documentation is out there, and HP has done a wonderful job in documenting the journals..
but armspire does a far better job of explaining the fundamentals!!

 

[digsys]

If you are also considering a transformer based linear regulator approach, then if you can get hold of a good multi-tapped transformer, there is a power transistor arrangement used by HP and others that allows you to automatically switch between the transformer outputs without relays or mosfet switches ..... If you are interested in a linear design, I can describe the method HP uses.

Is this a magnetic amp design ? I work on those and linear multi-tap designs a lot. IF you understand DC controlled AC, then it is
literally the "perfect" system. I agree with amps, you'll never get lower noise than a good linear, even with series chokes / filters, PI networks - unless it's synchronous resonant, which is quite awesome. Out of interest, can you post, or IM the method, amps?

[eevblogfan]

hey

What Richard talks about is AWSOME Circuit called "pre regulator" But made of SCR's , since I can not explain better then Richard , I'll let him explain and all the credit , Just wanted to give the name In order to give you something to google until Richard post the more detailed and specifice explanation:P

Google: "SCR pre regulator "  for more info about the Idea

plus - LT PDF :  http://cds.linear.com/docs/en/application-note/an32f.pdf

Cheers

[digsys]

Thanks, but that technique has been around for quite a while, and has it's own issues. I am hoping that it's maybe something "novel" or new. :-)

[amspire]

What Richard talks about is AWSOME Circuit called "pre regulator" But made of SCR's

Actually that is another good approach, but I am going 100% linear - for better or for worse.

The question is this - say you have four DC voltages from the transformer/rectifier circuit. How do you automatically switch between the four different voltages without any relays?

Here is the circuit. It is one of those circuits that you look at and say "How can that work", so I did it as a LTSpice project so people can play with it.



The devices are just ones I picked quickly from LTSpice. You would need MOSFETs rated for high current linear operation - most are not. The NPN darlington pair does all the regulation of the output. For reasons I have mentioned previously, I think transistors are very much superior power supply regulator devices then MOSFETs.

V5 is replaced by the driver transistor from the regulator circuit.
It is simple reducing or increasing the stages - just follow the pattern. This approach does need extra mosfets, diodes and electrolytic capacitors which is why most modern budget power supplies use other circuits, but it does switch supply voltages instantly and I just think it is a neat idea. No relays. No weird glitched as you wind up the voltage - just a smooth transition.
All three MOSFETs can be mounted near each other as when one winding is active, the others are either hard on or hard off.

The plots show that there is a seemless transition of current from the appropriate winding:



If you have two center tapped windings on the transformer, here is how you get 2 DC voltages out from each winding:



If anyone needs an explanation, let me know, but see if you can work out how the circuit works. The LTSpice file is below.

Richard.

[BravoV]

Richard, I'm so happy, thank a lot ! 

I have this beast and its gathering dust for years waiting for this kind of circuit. 

Great... no more relays .. 

[amspire]

If the mosfet gates are rated at 20V, it may be possible to eliminate D7 and D5.

Here is the power of the different power devices with 10A output :



The graph exaggerated the Darlington power a bit, as my load is a current source, so the graph is starting at a negative voltage. The peak darlington power looks to be under 85W.

[C]

Richard,
I remember a couple circuits in some old equipment you may want to use in your PS.

1. The basic idea was the output of the Power Supply was controlled by the voltage on a cap. That's it, a cap with a buffer amp controlled the power supply output. They added a circuit to counteract the cap's leakage. The control circuit to set the cap's voltage was isolated almost to extremes. Instead of one path to charge or discharge the cab, they used two to control the charge. The charge control circuit default state was balanced to counter the leakage from the cap. The ends of the two resistors not connected to the cap could be varied almost to Power Supply rails. For fast changes, the control moved both to the same rail. To kill noise, opposite rails, while trying to get back to a state to just counter the leakage. A very high impedance input buffer connected to the cap supplied the output to the control circuit and the power supply output control. From what I can member it was very stable over a very long term.

---

2. I remembered the second circuit when you think of a power supply in the right frame of mind. Try this, "A power supply is a power operational amplifier with a lot of parts missing or reduced to almost nothing". The equipment I am thinking of used tube based operational amplifiers. Dating my self a bit a bit, but they wanted easy to repair equipment, so they used the same tube opamp all over the place. For example, to convert a single ended signal to differential signal, they would tie the opamp's output to the voltage level the differential signal should be centered on.  The Differential Signal they then picked up off the resistors they inserted in the supply leads to the opamp. They would often use this to shift from between the power supply rails processing to a signal referenced to one or both of the power rails. If you think for a second, to get that very high power supply isolation specification on an operational amplifier in a chip, the designer needs to make all the currents static or have matching currents changing in opposite directions. In a single ended opamp they could do this for all but the final output.

For example, Instead of one opamp controlling the output, you might gang all 4 together in parallel with different gains so that combined output controls the power supply output. Insert a resistor in the supply lead to the chip to get a combined output of all the opamps.

In looking at Your LM324, it may have some hidden benefits. The 4 opamps may  be constructed in side the chip in such a way that they are almost matched. Not a good idea to count on it, but it still could have advantages. 

C       

[M. András]

for a pre-regulator, a buck converter with voltage tracking feature is suitable or there are some more suitable types for this application? ohh btw voltage or current mode?

[LaurenceW]

WOA! Be careful about sticking a nice (final) linear regulator on the back end of an SMPS, and then expecting it to filter out any high frequency noise that escapes from the SMPS.

Linear regs may regulate to the required DC voltage, but their power supply noise rejection can be quite poor at SMPS switching frequencies, unless they have been specifically designed or this. Noise just goes straight through 'em like a good curry!

I have seen some other articles around here on this very point (especially in relation to fixed three terminal regs), but cannot find it now.

[BravoV]

The devices are just ones I picked quickly from LTSpice. You would need MOSFETs rated for high current linear operation - most are not.
... <snip>
All three MOSFETs can be mounted near each other as when one winding is active, the others are either hard on or hard off.
... <snip>
If anyone needs an explanation, let me know, but see if you can work out how the circuit works.

Richard, interesting circuit.

From the plots of 3 mosfet's V_gs and their dissipations, now ... questions .. 

Looks like the mosfets are not driven really hard aren't they ?

Is this the reason you said we have to use those beefy & expensive "linear" mosfets ?

[amspire]

Looks like the mosfets are not driven really hard aren't they ?

Hard enough. When you start with a 200W device and derate it for temperature, you will still have to have a good heatsink to get about 80W dissipation.

Is this the reason you said we have to use those beefy & expensive "linear" mosfets ?

The reason you need linear mosfets is they are specified not to blow up when used in a linear mode. Most mosfets can have great current capability when fully on, but when in linear mode, the current sharing between the mosfets cells is not good enough to to stop excess current going to one of more individual cells. In a switching mosfet, there is thermal runaway, so the hotter a cell gets in the mosfet, the greater a share of the current flows to it. That cell blows bringing down the rest of the mosfet. The best linear mosfets will probably have a higher resistance and lower current capability then a switching mosfet with the same die size.

If the mosfet does not have a DC Safe Operating Spec that shows it can be used at 10A at the maximum voltage (About 9V in my example), then it cannot be used in this supply even if it is a 200A mosfet.

In designing a high current switching mosfet, the idea is to minimize the resistance of each mosfet cell in the mosfet. In a linear mosfet, you have to add enough resistance to each cell so that the cell resistance dominates in the current sharing over the cell gate sensitvity. If the resistance dominates, then when a cell heats, the resistance increases and the current reduces.

So to make a great linear mosfet, you have to deliberately make a poor switching mosfet. It will probably need a higher gate voltage to drive the linear mosfet as well compared to the switching mosfet.

[BravoV]

Linear mosfet is quite pricey, about $50 a pop. 

Since this kind of circuit is mostly for enthusiast for one off project, is there any other trick in your sleeve  as such to drive the mosfet harder (much lower dissipation) even it means that this solid state transformer taps switcher circuit requires another separate transformer windings ? Or even separate smaller transformer ?

[DLWarr]

Now maybe the Original Posters have moved on to other things... But 1000's of new people are just starting there search for circuit ideas like this... This is a vary Unique circuit, with vary little info on the web... I'm sure it would interest a lot of people,,, it's to bad It's hidden away with the title " Lab PSU Design Idea/Question" ... That said,, if anyone has other Ideas or can point to info on switching the taps for linear supply's please post them....