Friday 7 February 2020

A power supply 13.8V/2A

I have been struggling to provide a suitable power supply for my ELAD FDM-DUO SDR Transceiver. The spec says it needs 2.2A @13.8V +/- 10% (or 12.4 - 15.2V). This sounds easy enough. I had been using a 10A/12V switch-mode supply giving 12.8V (the 12V supply cranked up to its max), but I read that SMPS emit stray RF noise so I determined I needed a linear PSU. And anyway this SMPS is too big to go in the fancy blackbox that I have available.

FIRST ATTEMPT
I had a toroid transformer lying about rated at 0-12V 0-12V @ 3.6A which may be enough. When bridge rectified and fed to a capacitor bank of 4 x 2200uF (diagram below is wrong) this yields an off-load of 20.8V, but an on-load as low as 14.7V, due to 2 x diode drops.

I looked around and found a low drop voltage regulator the LT1185 which is rated at 3A and has a voltage drop of < 0.75V. Great. So I built this circuit
CIRCUIT 1

And at first test this looked just fine. I set it to an output voltage to 13.8V and powered-up my FDM-DUO. The result on RX (load 0.56A) seemed excellent.

I work a lot of FT8 @ 5W on 40 & 20m and I have a very poor antenna (60cm loop on the windowsill!), so I set out to have a few QSOs. Now when the WSJT software hits TX in a digital mode the PA suddenly takes a startup current peak, but continues to ask a steady voltage. The PSU did not like this and occasionally the output jumped down to 10.8V and occasionally up to 15V. Why? Well I had omitted an output 22uF Tantalum capacitor to prevent oscillations, OK let's solder this in. But yet again the same thing happened. Now why? So far I have come up only with the idea that the sudden over current surge is tripping the LT1185's current limit. when it is hot.. maybe.

SECOND IDEA
So I need another solution. I like simple circuits, so how can I build a simple 13.8V supply? One that can deliver 3-4A or so and handle the peaks? What I found out is that to get low voltage drop the NPN pass transistor is placed in the negative rail, not the positive. If it is an emitter follower in the positive rail it will drop at least the Vbe of 1-1.5V @3A. If it is in the negative rail it can go down to C-E Vsat of 0.5-0.8V.

Various circuits on the web use op-amps for the feedback, comparing a zener voltage to the output volyage and feeding the drive to the pass transistor. But a simpler circuit can be built using just a single PNP transistor as the comparator/amplifier. So I came up with this circuit

CIRCUIT 2

It is an old trick of mine to use an NPN transistor reverse biased, forcing emitter base reverse breakdown of about 5V to use as a zener, so I will try this. Later: I tried it and it DOES NOT WORK! The 2N3904 gives a reverse VBE of 10.6V!!! Well I never. So I have ordered some 5.1V zeners...

WAIT A BIT
I ordered on Rapid (who I like a lot) a bunch of transistors (TIP41 & 42, 2N3904 & 6 and the 5.1V zener) and I await their arrival.

The parts arrived...  so I built a breadboard, like this


using this circuit.

CIRCUIT 3

And it appears to work. So I will build a final version, using a high power BD249 NPN and see if it can deliver 3-4A. On we go. (The 2N3906's emitter resistor on the breadboard, by the way, was 330R, It may need to be lower to drive the BD249 which has a lower gain... we will see. Worst comes to the worst the series transistor(s) will have to change to a darlington...

LIMIT
As you can see this circuit has no overload protection. To implement it you have to monitor the current, to do this you have to drop 0.7V or so and turn on a transistor which removes the drive. But adding 0.7V drop defeats the object of having a low drop regulator! Worst comes to the worst I will simply put an AC fuse in the 240v input... and hope.

BUT
It didn't work, the output fell to 9V on load. Due to the low gain of the BD249 series pass transistor and not enough base drive from the 2N3906's?

SO
I had another look around the web and discovered the LM338, a 5A regulator for 2-20V (or so) output. And I built circuit number FOUR

CIRCUIT 4

And this works, at least it does when the output voltage is restricted to 12.6V. I does not work at 13.8V which is what I wanted, as the transformer output after rectification is too low, and not enough for the voltage drop across the regulator. Same old problem.

You live and you learn. Anyway I will stick with this now. It is enough to drive my ELAD FMD-DUO to 5W output on FT8. But it doesn't seem much better than the SMPS I started with, except that it is in a pretty box.

LAST DITCH STAND
And then I had another brain wave. Why use a bridge rectifier when I have effectively a 12 - 0 - 12V transformer, each winding rated at 3.6A. If I use a simple rectifier on each leg then I can save one diode voltage drop. So I revised the circuit to this, number FIVE

CIRCUIT 5

And I cranked the output up to 13.8V, but this was too high and it sagged again to 13.2V on load. So I finaly settled for a Power Supply of 13.2V...


And, hopefully, that's where the saga ends.

NO, IT DOESN''T
After a few minutes of operation the display flapped about 11.2-13.8V!!! I turned the output down to 12.3V, but the same problem. Why?

I had a check and the input to the regulator was 15.8V, the data sheet says it has a drop of 0.75V, but it also says it could drop out if the Vin - Vout is less than 3V, so that means I should expect an output of no more than 12.8V, and even then it could dropout. It did.

ON WE GO
So I move on. The next trial will be with an LM741 opamp driving a series PMOS transistor, in theory the PMOS will have a voltage drop of Iout x Rds(on), and Rds(on) can be very low, say 0.1R at 5A, so the drop should be < 0.5V.

I await the PMOS IRF9Z30 devices to try it out. This is the circuit, simple huh?

CIRCUIT 6

One background concern is the stability of the "reference" 5.1V zener, as the current through it will vary depending on the input voltage, which could flap about from 20 to 15V on load.

The IRF9Z30's have arrived! Build time

I wired it up wrong, connecting the input onto the output and vice versa, twiddled the pot and got either 20.8 or 18.9V out. Then I spotted my mistake, fortunately no damage done. The voltage adjustment is a bit sensitive, but I set it for a measured 13.8V output - the LED display says 13.6V (it is wrong). I also added a 22uF Tantalum cap on the output, not shown here.

On RX load 0.56A the output is 13.6V indicated, and on TX load 1.8A it falls to 13.4V. Not a wonderful load regulation, but I suspect this is due to 1) the rectifier output falling to just 15.5V on load, 2) the zener voltage varying as it is supplied from the input voltage which ranges from 20.8-15.5V and 3) the supply to the op amp coming from the input side with a range of 20-15V.

Later, voltage display again started flapping around 11.2-14.5 on TX. Back to square one.

ONWARD AND UPWARD - 15V transformer?

I must get to grips with power supply design, eventually. What I will do next is to buy a 15-0-15V transformer and a couple of higher voltage electrolytics 4700uF/63V and try again... with circuit SIX.

The transformer arrived, fortunately it is the same size as the 12V one so fits, and I have updated the 4 x 2200uF caps to 2 x 4700uF / 63V ones.

And now, IT WORKS. Only that the input voltage to the IRF9Z30 is up at 22-24V and so the MOSFET power dissipation is a lot higher (9-11V x 2A = 18-22W), hope the heat sink is big enough...

WATTS
I now have more power than I need for the FDM-DUO, now and increasing the audio drive from WSJT I can easily reach 8W RF output. Wonder what the FDM-DUO limit is???

CRASH
I mentioned above that the display flaps around 11.5-14.8 or so. Now it is happening again with my latest circuit SIX! I have discovered it is due to a faulty voltage display module! - so all those early circuits may have been OK - Grrr.

While changing the display I accidentally short circuited the supply and phut! it failed, output at 24.5V and uncontrolled (fortunately my FDM-DUO was not connected). I have changed the MOSFET and the Op-amp and the Zener, but it still outputs 24.5V. I have no idea why. Bother.

BACK TO SQUARE ONE - or CIRCUIT 5 anyway
Being the simplest, and because I had a few LM338 available, I decided to knock up CIRCUIT FIVE again, with the 15-0 0-15V transformer.

OK, check the output voltage with my meter, and set it to 13.8V. The display now shows 13.7V. But wait it is flapping around again! But my meter is not - Oh ho, fault in the display? I change it for another, this time wiring it differently - power input from the rectified 24.5V, sense from the output 13.8V. But again it flaps around. (naughty words). Later it seems to stabilise.

Decided to live with it, knowing my meter is not showing any voltage variation on RX or TX. Now the only concern is the increased power dissipation of the LM338, with 24.5V in and 13.8V out, or about 10V @ 2A = 20W. The heat sink is not that big... so...but it gets fairly warm on TX.


Badly exposure photo, can't see the FDM-DUO display, but I assure you it is there, tuned to 7074kHz for FT8 ops. And actually, although the display says 13.8V, the metered output is 14.0V. 

So much for cheap and cheerful displays.





1 comment:

73 from Ian GM3SEK said...

To really "get to grips with power supply design", download PSU Designer 2 from Duncan Amps (duncanamps.com/psud2) and start simulating before you build.
It doesn't simulate voltage regulators, only transformers-rectifiers and C-R-L smoothing circuits, but you can adjust the current demand and watch what happens to the output waveform going into the regulator.
It certainly taught me a lot, including the fact that the sag in regulated output voltage will be due to the negative ripple peaks poking through the dropout threshold of the regulator.