Design new 0 to 30V power supply.

[Timmymna]

I've been keeping an eye on this thread and the old 0-30v design, which I was planning on building, but due to one thing and another haven't got round to building yet :

I was wonder how the two designs compare against each other, now there's two options which one to build?

 

[Hero999]

Using the MCU to turn the fan on and off would be a very bad idea unless you can ensure that it will fail safe i.e. if the software fails, the fan will stay on. The same applies to thermal shut down, although software controlled thermal shut down is still better than no thermal shut down.

Assuming a power dissipation of 28.333W per transistor and the thermal resistance from the die to the heat sink is 2.7

 

[redwire]

Using the MCU to turn the fan on and off would be a very bad idea unless you can ensure that it will fail safe i.e. if the software fails, the fan will stay on. The same applies to thermal shut down, although software controlled thermal shut down is still better than no thermal shut down.

I agree.  As noted  above

This can be a secondary shut down method if something goes out of wack while in development or evaluation,

  Providing safe controllable  hardware failures is  not always possible and can sometimes lead to unexpected results.  Having a second line of defense as well as a visual (LCD readout) of the problem in advance (seeing temperature rising uncontrollably,  or how the unit responds to different loads ) rather than the smell of burning parts or the flames from a burning part may be worth considering. 

 

[Hero999]

I'd probably use a comparator with an open collector output and a temperature sensor at the input. When the temperature exceeds a certain threshold the gate signal to M2 to M4 will be clamped to 0V, by the comparator, shutting the regulator down.

Another comparator could be used to switch the fan on when a certain temperature is exceeded.

I'd probably use the LM393 which is a dual comparator and will be able to do both.

A warning LED could be added to indicate the PSU has shut down due to over temperature.

The MCU could also tap the temperature sensor and do whatever it wants to do with the signal.

 

[PicMaster]

Just to let you all know due to other commitments I will not be able to fully help in building and testing the power supply,  . I have to get this other project finished that I've been working on has they have brought the completion date forward .

I will keep an eye on this topic and any spare minutes I will start and progress in getting it built up on breadboard I mean whiteboard ;D(only joking).  I have got most of the parts ready just need to get to order the op-amp and mosfets which I shall do on the next order I make. I was so looking forward to playing around with it.

 

[Hero999]

That doesn't matter someone will probably build it sooner or later.

I might even do it myself if I come across a free or very cheap 30V transformer.

 

[PicMaster]

Herro Could I use a different type of op-amp say a LM324 or some lm741's, I know that the higher input voltage would be no good, But has I have not got around to order the correct one yet, I was thinking using the above op-amps but only use a 15-18V transformer for the trial I know this will not be the 30V version but say a 20V version. Only I've got a bit of spare time to start playing with and get this topic moving again
My thinking it should prove it to a certain degree at at least we can see if the voltage/current adjustments work and see how good the ripple, Correct me if I'm wrong.

 

[canmemo]

I want to join this PSU project ,  maybe i can help you guys implement digital control unit.  But I m not good as like you on analog part.

Can you please tell me , why are we using u1d and M1 block ? and please tell what is the function on the circuit.

If im not wrong , you measure current using R5 , but how can  we found 15mR ?
In your bill of metarial list , you said (150K and 10K in series) , but i guess these words for R6 (160K) .


In a few words about my idea,

I m thingking to use ATMEGA series for digital control unit , atmega8 is cheap and some versions works on 20 Mhz . i guess its enough the handle voltage and current control , and we can use (MCP 4922 ) for DAC,  it has 12 bit resolution.
Another hand,  we can re-design VRef block , use  LM4050 shunt voltage ref. IC for it. if we can able to change Vref value to 4.096 its best for calculation in MCU . but that time we need to change Opamp gain to get 30V from output

I need your kindly response.

Thanks.

 

[PicMaster]

I want to join this PSU project ,  maybe i can help you guys implement digital control unit.  But I m not good as like you on analog part

By all means feel free.

Can you please tell me , why are we using u1d and M1 block ? and please tell what is the function on the circuit

The purpouse of this is below 15V we only use the one side of the transfomer and above we use the 30V side of the transformer, This prevents generating to much heat at low voltages

If im not wrong , you measure current using R5 , but how can  we found 15mR ?
In your bill of metarial list , you said (150K and 10K in series) , but i guess these words for R6 (160K) .

Yes R5 is the current resitor, Many electronics suppliers should sell them I know Farnell & RS do sell them and there world wide.

I m thingking to use ATMEGA series for digital control unit , atmega8 is cheap and some versions works on 20 Mhz . i guess its enough the handle voltage and current control , and we can use (MCP 4922 ) for DAC,  it has 12 bit resolution.
Another hand,  we can re-design VRef block , use  LM4050 shunt voltage ref. IC for it. if we can able to change Vref value to 4.096 its best for calculation in MCU . but that time we need to change Opamp gain to get 30V from output

You can use what ever MCU you like, The more designs make it more interesting and suits everybody's neeeds. I personally use PIC'S. The volatage out of the TL431 is just abouve 5volts this is the referance voltage for the voltage boost on the op-amps. So a referance voltage for voltage measurements I would use another TL431 on the digital board anyway

 

[Hero999]

Herro Could I use a different type of op-amp say a LM324 or some lm741's, I know that the higher input voltage would be no good, But has I have not got around to order the correct one yet, I was thinking using the above op-amps but only use a 15-18V transformer for the trial I know this will not be the 30V version but say a 20V version. Only I've got a bit of spare time to start playing with and get this topic moving again
My thinking it should prove it to a certain degree at at least we can see if the voltage/current adjustments work and see how good the ripple, Correct me if I'm wrong.

I think I answered this before but the site went down.

The op-amp needs to be a to work from a single supply i.e. both inputs must work down to 0V and the voltage rating must be high enough. The LM324 will be find providing the hhigh DC voltage is kept below 36V. The 741 is not suitable because the inputs won't work down to 0V.

An MC33174 could be used but the transient response isn't as good as the IC I used.

 

[PicMaster]

I think I answered this before but the site went down.

I say I never saw it or if you did answer I may have asked before but it's been a few weeks and been to bed since then many times.

I would only use the LM324 has a stand in for the trial; I thought it just be nice to see how it would perform in real life The next time I place an order I shall order the correct chip.

 

[PicMaster]

I have now obtained most of the parts for it so I'm ready to start building it, I've got the correct chip now  ;D.

has anybody built it yet means this topic was disapearing off the list

 

[Hero999]

I don't think anyone has made it but I'm certainly interested if you make it.

If you get it working and post some pictures I'll put them up on Silicon tronics if you like.

 

[PicMaster]

I will make a start this afternoon on it, It will be built up on my bread board rig, Hopefully it will work has good has it does in the simulator, I will take some pics of the tests, I have no problems where you post them, I have a load tester where I can make the current variable from 0-10amps but we will only be set to 3-5amps the 10A version can come later as well as the digital control part

So I can carry out a good load testing at different volatges to see how really well the PSU performs for long periods of time, Of course I shall use fans to keep things cooler  ;D
I will report on progress

 

[Hero999]

I hope you're aware that your breadboard will not be able to handle more than 1A so you need to connect all the high current components with wire, off the breadboard.

 

[PicMaster]

I hope you're aware that your breadboard will not be able to handle more than 1A so you need to connect all the high current components with wire, off the breadboard.

Yes I know I was going to do that any way, I've decided to layout a PCB for it, Even though I know that the design is not complete or may not work, The reason I've gone down this route is this way as long as the schematic is correct the PCB will be correct and less chance of any man made errors in hooking it up, if all goes well then we are half way there and can make minor/major changes if needed. once I spent hours trying to find out why a circuit i did on bread board did not work and it was a man made fault(I got some wires wired wrong), the PCB is almost completed, I shall use the press n peel method to make real quick and easy.

 

[PicMaster]

Update, I have finally got it built and working, Well I have only tried the voltage through the full range from 0-30 volts, Partly tried the current draw test on it, at the moment I can't do a full current test due to the fact I have not got the 15mr resistor I only have a 0.1R which only draws 1.97amps, I think this can be altered by the resistor values but not tried it yet.

The only thing that I did notice is that the current pot has to be turned up slightly to get any output voltage.The good thing is that the ouput voltage goes right down to ZERO  ;D

I shall carry out lots more investigstion on it, Has there is lots more to be tested, I think that it would be a good idea to to put an extra 1L pot for fine tune on the voltage ?

Herro do the pots have to be 100K or could we get away with 47K pots also which are the best pots to use LOg or LIN ?

Here are some Picutres of it so far.
What other tests do I need to look at carrying out ?

View attachment 40910

View attachment 40911

View attachment 40912

 

[Hero999]

Great, I'm really glad my design works.

Thanks to all that helped me design it and you for building and testing it.

The only thing that I did notice is that the current pot has to be turned up slightly to get any output voltage.

That's to be expected.

If you set the current limit to zero then the output voltage will drop to zero because it won't allow any current to flow and Ohm's law states that at zero Amps, you have zero volts.

I shall carry out lots more investigstion on it, Has there is lots more to be tested, I think that it would be a good idea to to put an extra 1L pot for fine tune on the voltage ?

It's possible, I would recommend a 10k pot. in series with the negative side of the voltage adjustment pot.

I don't think it would be a perfect solution though, a multi-turn pot. would be ideal but they are expensive.

Herro do the pots have to be 100K or could we get away with 47K pots

You can replace R21 and R22 without making any changes, but if you replace R20 with 47k, C5 will need increasing to 220nF and R23 will need reducing to 470R. This is to keep the current limiting delay time constant the same.

also which are the best pots to use LOg or LIN ?

I'd recommend linear because the voltage change will be the same for a given adjustment of the pot. regardless of whether it's at the high or low voltage setting. A logarithmic pot. will be more sensitive at higher voltage settings than at lower settings: some people might want this but I certainly wouldn't.

What other tests do I need to look at carrying out ?

Measure the V+ and adjust the output voltage, it should suddenly increase when the output voltage is increased to above about 12V. It's probably best to do this with a load connected, 1A should be enough.

You could test the transient response.

You'll need a 'scope, a pulse generator a, N-channel MOSFET and a power resistor.

Connect the power resistor between the MOSFET's drain and the positive rail, the source to 0V and the gate to the output of the pulse generator set for a high enough voltage to turn the MOSFET on.

Turn the power supply on and set the voltage and current limit to a level where the load will pass near the full 5A (or 1.6A if you can't do 5A yet).

Look at the waveform on the output of the power supply, adjust the frequency of the pulse generator until you can see pulses on the PSU output.

When the load is turned on there'll be a small voltage drop, when it's turned off there'll be a small voltage spike. The smaller the spike and dip and the shorter the lenght of time the last for, the better the power supply is.

Then there's the overcurrent transient response: repeat the above experiment but reduce the load resistor so the current exceeds the limit by a factor of two, and measure how long it takes for the voltage to fall to zero. Adjust R20 and you should be able to observe a change in how long the supply takes for the overcurrent protection to kick-in. The higher the resistance, the longer it will take.

You might need a digital 'scope for the last test because the supply is likely to take awhile to recover from a short circuit and you might not be able to see the short circuit part of the waveform on an analogue 'scope. With a digital 'scope, replace the MOSFET with an SCR and the signal generator with the gate connected to a suitable supply via a push button switch in series with a 1k resistor. Connect the 'scope's trigger to the SCR gate. When the push button switch is pressed, the SCR and 'scope will be triggered, the PSU's current limit will be exceeded but it will take a finite amount of time for the voltage to fall to the current limit multiplied by the load resistance.

A can post a schematic along with expected waveforms if any of this isn't clear.

I would be interested to see if a negative supply is really needed. LTSpice wasn't happy until I connected U1 to -V instead of 0V. Try disconnecting U1 from -V and reconnecting it to 0V. If the supply works just as well, then the following components can be removed: D14 to D16, C12 and C13 and R19.

I hope the whole circuit can be fine tuned and all bugs eliminated so it can replace the other crappy power supply.

 

[PicMaster]

I would be interested to see if a negative supply is really needed. LTSpice wasn't happy until I connected U1 to -V instead of 0V. Try disconnecting U1 from -V and reconnecting it to 0V. If the supply works just as well, then the following components can be removed: D14 to D16, C12 and C13 and R19.

I will try that and let you know.

That's to be expected.

If you set the current limit to zero then the output voltage will drop to zero because it won't allow any current to flow and Ohm's law states that at zero Amps, you have zero volts.

I thought that but if I remmember in the sim it does not do it. I know the sim to real world might not be the same.

You can replace R21 and R22 without making any changes, but if you replace R20 with 47k, C5 will need increasing to 220nF and R23 will need reducing to 470R. This is to keep the current limiting delay time constant the same.

The only reason I asked becuase, A velleman PSU I had has a 47K & 1K pot for the voltage adjust, The 1K pot gives you a fine tune over 1V has it went up in 0.01V steps which was good so you could set it to 13.84V easy.

Measure the V+ and adjust the output voltage, it should suddenly increase when the output voltage is increased to above about 12V. It's probably best to do this with a load connected, 1A should be enough.

That's wasy enough done.

You could test the transient response.

You'll need a 'scope, a pulse generator a, N-channel MOSFET and a power resistor.

The only scope I have at the moment is a velleam 10MHZ one, My other old analogue one packed up some time ago and I've never got around to replace. (well the wife put her foot down. may be I need to earn some more browning points  ;D), Has for the pulse generator I can hook up a pic to create the pulses(I think).N-Fet and power already have set up for another job I did(load testing).
My skill of understanding scopes are limited may be if the velleman scope would be up to it I could take pics of the screen if that would help.
The rest I have copied and printed for bed time reading and to get it to sink in, attached is my schematic I created to make the PCB, I know some of the vaules in the schematic may not be correct but on the real PCB they are, Once we get closer in the final design I will correct all values and make the modifications then creat the BOM for it.mosffet_power_supply.pdf

 

[Hero999]

The only reason I asked becuase, A velleman PSU I had has a 47K & 1K pot for the voltage adjust, The 1K pot gives you a fine tune over 1V has it went up in 0.01V steps which was good so you could set it to 13.84V easy.

That won't work well with this circuit. Suppose you put the 1k pot. in series with the positive side of R21: altering the 1k pot will make a large difference when R21 is set to the high setting and not much difference when it's set near 0V. If you put the 1K in series with the negative side of R21, the situation will be reversed. This is why using a multi-turn pot. is the best solution.

That's wasy enough done.

Sorry, I don't understand, is that a yes or a no?

The only scope I have at the moment is a velleam 10MHZ one,

That might be able to measure the transient response, but it won't be able to measure the current limiting delay.

I think you might be able to do it another way: with a PIC and a comparator.

Set the PSU voltage and current limit known values. I know that sounds vague but remember it doesn't matter what the current or voltage settings are, just as long as you record them.

Connect a load which will draw double the set current via an SCR. For example if the output voltage is set to 25V and the current limit is 2.5A use a 5R resistor which will try to draw 5A.

When the SCR is triggered you need to monitor how long it takes for the current to drop to half the current limit setting plus 36.8%. You can use Ohms law to work out what the voltage will be, for example, for a 5R load, a 2.5A limit and a voltage setting of 25V (as above) the current is 2.5+36.8% = 2.5*1.368 = 3.42A so the voltage is RI 5*3.42 = 17.1V.

Here's where the comparator and PIC comes in. The SCR can be triggered from the PIC which will then wait and time how long it takes for the voltage to drop below 17.1V which the comparator can detect.  I'm sure you can figure out something.

Adjusting the value of R20 should alter the current limit delay.

The rest I have copied and printed for bed time reading and to get it to sink in, attached is my schematic I created to make the PCB, I know some of the vaules in the schematic may not be correct but on the real PCB they are, Once we get closer in the final design I will correct all values and make the modifications then creat the BOM for it.

I noticed that you've substituted a few of the resistors for other resistors connected in series. Did you just do it with the ones I suggested on the BoM or were there other resistors you couldn't get hold of and had to replace with two in series?

I just suggested the substitutions for the E24 values that I know are harder to get hold of than E12 values.

Maybe the final PCB could include space for the extra resistors and they could be replaced with wire links if not needed? Another option is to provide space and bypass the pads with a track which can be drilled if the resistor is required.