Need a cheap dual-output, center tapped, 180VA 18V 5A transformer?

I am trying to design a nice dual-rail lab power supply and need a cheap dual-output, center tapped, 180VA 18V 5A transformer. The outputs are as follows:

Primary: 120V
Secondary 1:  9V-0-9V, @ 5A max.
Secondary 2:  9V-0-9V, @ 5A max.


Primary 120V
Secondary 1:  9V, @ 5A max.
Secondary 2:  9V, @ 5A max.
Secondary 3:  9V, @ 5A max.
Secondary 4:  9V, @ 5A max.

I prefer a transformer for the linear regulator over a switching preregulator, because a switching preregulator sort of defeats the whole purpose of making a nice, high precision, low-noise lab instrument. However, I also prefer to have more than just 2 separate 18V outputs, since that means when the output is shorted and constant current mode kicks in for either the positive or negative rail, I will have a voltage drop of 18V, at a maximum current of 5A, and thats over 90W of power dissipation!!! I think that I can deal with that if thats the only solution, since I am planning on using a large CPU active heatsink for cooling, but I prefer if I was not pushing a TO-247 device to it's thermal limits.

I have found this: http://www.antekinc.com/as-2218-200va-18v-transformer/
It is affordable, but it seems strangely cheaper than a lot of other toroidal Xformers on the market. Also, the 2 primaries are not center tapped. It is 200VA though! I have never heard of that company, and it seems like the description of it is written in chinglish. They act like the fact that they over-engineered and under-specced it is a feature (to me, thats how all ratings should be, that should be normal and standard, not a feature.).

I have also found this: http://www.digikey.com/product-detail/en/36-6/36-6-ND/1984766
It would be perfect, but it is $80!!! Very expensive, Also, I do not think it is a nice toroidal transformer, not that that matters much to be, I just like the professional look of a beefy toroidal transformer inside a power supply.

Lastly, I have found this: http://www.mouser.com/ProductDetail/Hammond-Manufacturing/1182G18/?qs=sGAEpiMZZMvwUzoUXIIvySPCJQuQgm7bYNfLdl9rdUYluBT2%2f%252bFqzw%3d%3d
It seems to be very similar to the first find, and it is from a reputable distributor. At $70 it is still expensive, and I do not want to spend over $100 for a homemade power supply. I would just get a off-the-shelf solution for that!

What about the possibility of modifying a MOT transformer? How many VA or watts can I expect from the output without active cooling? I know they are rated from upwards to 1000W, but I also know they cannot sustain that power output for a long time without overheating and requiring lots of active cooling, and from what I remember, I believe the core is really being pushed into saturation during normal operation, leading to LOTS of power loss in the transformer core, but can they output 180VA continuously and effectively? Approximately how many turns of wire would I need to achieve my requirements, and what gauge wire should I use? I am thinking 20AWG but someone who has done that before would know better than me.

sort by: active | newest | oldest
-max- (author) 2 years ago

I have sort of copied the architecture of Daves power supply design, but not use the switching preregulator at the last video, and slightly different design, but not too far off I don't think.

Wired_Mist2 years ago

There is always an ATX power supply, (I assume you've allready looked at that; Just reference for anyone else who's reading this) I can get them for $25 incd Tax.

If you really wanted to build your own I suggest the Hammond one. Good company rep and it's a bit over built for you app.

For the regulator go for a Lm-1084. It's an adjustable 5A linear regulator. It's in a T-220 package and even uses the same calculation as the LM-317 :D

-max- (author)  Wired_Mist2 years ago

I considered using a ATX power supply, but they simply do not fit my needs. I need a clean output of +-15V, I think the highest voltage rail on a ATX supply is 12V, and the -12V rail is only like a amp or something. Also, being a switchmode regulator, it has nasty RF switching noise on the output. (even the new one in my computer is noisy, I can hear the RF interference if I turn up the external speakers.)

Also, the Lm1084 is not suitable either, mainly because I want the CV and CC mode, both of which can be set with a microcontroller or DAC. I was planning on using a atmega328 or similar driving 4, 12 bit DACs, and the Vref to be 2048 or 4096mV. The issue with many of those types of adjustable regulators is that the voltage adjust on them is designed as a feedback mechanism from a resistor divider. Thats not easy to adjust with a voltage output from a DAC referenced to 5V or something. No matter anyways, since I already got a working regulator together in LTspice, with all the fully featured (except the features possible from an arduino control, I dont know how to make and program a microcontroller in LTspice yet.)

Screenshot from 2015-02-02 22:28:54.png

Wrong topology: A multiplying DAC works like a variable resistor, if you use it in the correct way.

-max- (author)  steveastrouk2 years ago

An idea did dawn on me after reading this, what if the reference was essentially connected to the output, and the Vref was on the non-inverting pin, so that the analog value output from the DAC would be able to set the "resistor divider ratio" in such a way. However, i am afraid that that will upset the loop stibility of the design, especially if the ref input and output is filtered. If this what you mean by using a MDAC?

Don't suppose you have a reference?

Sounds easyer then a Digital Pot >.>

-max- (author)  steveastrouk2 years ago

I have heard of digital POTs, but I have also heard that they are fiddly, and have lots of requirements to pay attention to to make sure to use them the correct way. I just so happen to have some really old and huge fancy 16 bit DACs from the 80's, the Burr Brown (TI now) DAC72BH's, as well as the AD574A. I have about a LOT of both of those, at least 15 of each, and they are encased into pink ceramic DIP package, the kind with the gold tops, so you know they are quality! ;)

What do you mean I have the wrong topology used here? I have set up a Sziklai pair and U1 such that there is a negative feedback from the output, and so the op amp will regulate whatever voltage is seen on the non-inverting input, times 3. (the feedback resistor ratio is 2:1, so the voltage is divided by 3) I then have a current-limiting error amplifier set up so that it's capable of pulling that voltage down lower in the case of overcurrent. This offers no benefit over having the 2 error amplifiers with common-collector outputs (or diodes on the outputs) connected up to the Sziklai pair with a pullup resistor present. That was my old topology and I could not get zero volts output with that config.

I have inverted the current error amp, and the output is now connected to the feedback resistors, so when an overcurrent condition occurs, the voltage is pulled up, causing the output of U1 to fall, so the pass element is turned more OFFish. That made everything work nicely on a single 18V supply, except when like before when the output is shorted, oscillation of 50A starts happening.

My design is not too far off from the internals of the LT3080, with the exception of the chosen pass transistor type. I did not like the darlington pair type due to the crazy 3Vdrop when the base voltage = supply, and a single power transistor would not have worked since op amps cannot deliver the necessary current, and PNP or P-channel devices are a nightmare to make stable. (I do not know much about control theory, but my guess as to why is because a common-emitter configuration has lots of gain, then the gain of the whole feedback loop is way too high, whereas common-collector mode has unity voltage gain.)

-max- (author)  -max-2 years ago

The only thing I don't like about those A/D converters and D/A converters is that the digital rail input is literally 16 digital input pins, or 12 digital output pins, and that I will need a few shift registers to make it all work with the few pins on the arduino.

yeah 15V ai'nt gonna happen from an (unmodified) Atx, Oh well :)

Are you planning to make these digitally tune-able? that does sound cool !

The lm-1084 can do Constant voltage, or constant current, The resistor that bridges Ajst to ground can be switched out for a pair of pots(One corse one fine OR Even a single digital one is you want)

Are your trying to force more current + voltage down the line then what the device would normally accept? (just for my own knowledge, seeing you've chosen your Reg already :P) I've always gone for the simple regulate voltage and let current do as it will.

-max- (author)  Wired_Mist2 years ago

A good lab power supply will have a voltage set on it and a current limit set on it. Say I set the voltage to 12V, and the current limit to 500mA. If I goofed something up in my circuit, and there is a short or something somewhere, then immediatly when more than 500mA of current is drawn, the voltage on the output will be lowered until only 500mA of current flows, and I can make a red Constant-Current mode LED come on when that happens. Once the fault is fixed, the voltage will rise back up to the 12V.

This also works as a constant current supply, I can crank up the voltage to the highest possible rating, and set the output current, in that case, the lower the load resistance is, the more voltage sag there is on the output, because the output impedance of the power supply is effectively infinite. (it regulates current by changing the voltage) That is ideal for testing diodes, LEDs, amplifier stages, or any load that is best powered with a high impedance.

I personally do not like the idea of a coarse and fine adjust, thats just too many knobs to remember, and it's a sloppy solution. If I did the 100% analog approach, I would use a 5 or 10 turn knob, but they are $$$. A nice-feeling clickable rotary encoder is cheap and can precisely step in like 0.05V voltage increments, depending on how I program it.

yeah, I like the same type Pots to use as a fine ajst, then a second one so I'm not turning the multi turn pot a half dozen times X_x

I guess what you've constructed VS a simple regulator, boils down to the level of control and precision your looking for :P

-max- (author)  Wired_Mist2 years ago

What digital rotary encoders let be do is use "expo" when when I wind it fast, the voltage can go up exponentially fast, and fine tuning it by slowly turning it should also be easy. It is kind of the best of both. Another method would be to add a cursor to the display voltage/current readout and select which magnitude to change, like any of the values in, say: 05.00V although that would probably be difficult to program.

-max- (author)  Wired_Mist2 years ago

I might even add some sort of bluetooth to it so I can dial in the voltage and current with a app on a phone, and see the display. I will need to figure out how to setup a bluetooth connection with the arduino and make an app to talk to it. That might not be easy, but just an idea.

XD, Awesome

Look around Instructables, I've seen no lack of Documentation kicking around. Should be simple enough to send the Voltage/Current as txt or even a number to the app. Depending on what app you use you may have the option for some sort of a scroll wheel or a slide. Otherwise, you would have to set it up kinda like a cryptex. Code in a buton above and below the number to ajust the Value, and add an Update button.

-max- (author)  -max-2 years ago

The circuit still has kinks to work out, it actually parasitically oscillates when the output is completely shorted (with less than 0.05 ohms) which leads to oscillation of a few millivolts. I think real-world results will differ, due to the non-ideal nature of things. Also, I still need a slightly lower voltage on the current-limiting comparators negative rail. It is too bad I need that, but the I cannot overcome the Vdrop of the diode any other way to get 0V, and I tried useing a N channel MOSFET between that point and ground, only to get parasitic oscilation.

Cut the shunts off the MOT with a grinder, and you can improve its efficiency considerably. You should easily be able to get 180VA from one. Turns ? Measure it. Strip off the secondary, and wind 10 turns of insulated wire in. Measure the output volts. Scale accordingly. Heavier the wire the better - it will significantly improve the regulation of the transformer. Multiple windings are very simple. If you want to centre tap, its dead easy.

-max- (author)  steveastrouk2 years ago
What is the purpose of the shunts? Seems to me that they "short out" the magnetic field. Is it to increase the inductance on the input to lower the current draw?
iceng -max-2 years ago

That shunt, the air gap (yes there is one) and capacitors form a ferro resonant constant voltage like sola for the magnetron.