Jump to content
Electronics-Lab.com Community
Sign in to follow this  
Hero999

Design new 0 to 30V power supply.

Recommended Posts

I remembered this:

http://www.elektor.com/magazines/2001/november/digital-benchtop-power-supply-(1).54823.lynkx

http://www.elektor.com/magazines/2001/december/digital-bechtop-power-supply-(2).54867.lynkx

I have both articles, maybe we can pull some ideas or tips. Let me know if interested.

Share this post


Link to post
Share on other sites

Come on Hero, if you don't stop posting I will never go to bed and it is GMT+2 here  :o Besides, I am supposed to be away for Christmas. ::)

Now bypasses with 100nF, see the edit to my previous post.


Maybe it is worth adding a capacitor in series with the gate of M2, M3. This will waste some gate turn-on volatge (depending on capacitance ratios) but will decrease the gate capacitance that U1/U3 see. Maybe we can select logic-level FETs for M2/M3.

Your email address is hidden.


I have ticked 'hide from public' in my account, does public include registered members?

Please expand a bit more.


Yes so I run the simulation with ideal sources and ones with 0.4Ohm impedance and the difference is huge. That is what I meant by heavy dependance. With the percentages I meant what is the maximum tolerable peak-peak AC ripple as a percentage of the average DC output.

It seems to be fine as long as the input voltage isn't too low


If I had to choose between more coil (higher volatge trafo) and thicker coil (higher current trafo) I would choose thicker coil as it will keep efficiency high, maintain tight regulation and not fry the op-amps (assume a primary voltage spike) at light loads.

Share this post


Link to post
Share on other sites

I remembered this:

http://www.elektor.com/magazines/2001/november/digital-benchtop-power-supply-(1).54823.lynkx

http://www.elektor.com/magazines/2001/december/digital-bechtop-power-supply-(2).54867.lynkx

I have both articles, maybe we can pull some ideas or tips. Let me know if interested.

Has I understamd this was not a very good power supply, I've never built it but others copied it and they removed there design off there web sties saying it was unstable, but you may be to understand it better than me
I love this topic I'm learning so much from you guys it's mind blowing, Thanks

Looking at the Schematic I see you have used 3 mosfets for the output current but only one to switch the main current, can you please explain the logic behind it

Woulld we get away with only using 1 mosfet ?,

I have a 30/50amp discharger which only uses 4 mosfets and they don't even get warm. but then again I may be talking nonsence has it works the other way  ???

Share this post


Link to post
Share on other sites
I've never built it but others copied it and they removed there design off there web sties saying it was unstable,
I have never built it either. Can yu point us top any reviews?

One MOSFET M1 to switch in the bridge rectifier if a high output volatge is required thus effectively adding both secondary coil voltages. When M1 is OFF, only one winding is offering power through D5 at any time. Although I haven't looked at exact numbers, since M1 is used as a switch, it will either excibit a high resistance or a very low resistance. In both cases, the dissipated power will be low.

But in the case of M2 M3, they are used as the 'pass-element' and have their resistance altered in a way that the output voltage is regulated. These FETs can dissipate significant power because of that (linear regulator action, like an LM317) hence two have been used to share the dissipated power.

There are factors that determine how warm the FETs will get, such as cooling or if they are the actual discharge element. In your charger it is possible that a huge resistor is used to dissipate the heat.

Share this post


Link to post
Share on other sites

Looking and studying the drawing more I can now see that M1 acts as a switch silly me. I understand now you are correct in saying that I've got resistors that used to dissipate the heat.

have never built it either. Can yu point us top any reviews?

I will see if I can find any more info out, I did build the circuit on a test board so that I could control it with another pic but I found that under load some parts blew up can't quite remember which parts did hsa I give up with it and never looked any further into it, I may have had it wired up it wired up wrong

I'm pritty sure that this was one of the web pages that did it but they do a newer design now, may be worth a look
http://www.linuxfocus.org/English/June2005/article379.shtml

I've attached a zip file containing the other version of the elektor PSU, This is what is said about it
Hi @ll

I can mail to thoses who whant the : General purpose PIC16F876A controlled power supply, New version V3, April 25, 04 of VE2EMM from www.qsl.net

the author removed his project due to some noise ??? around the pic or at the psu output.


you guys will understand that more than me, 

PIC_Power_Supply.zip

Share this post


Link to post
Share on other sites

I know I would of used 10bit PWM myself,

Just has a thought instead of using a mosfet to switch the higher volatge what about a relay instead
Like the drawing below, Once the output reaches 12V the relay kicks in the higher voltage

post-44828-14279144037411_thumb.jpg

Share this post


Link to post
Share on other sites

If I haven't missed anything, your schematic will have the same outcome. You are also using full-wave all the time.

Your schematic is using one diode less, but is using a changeover relay which might have speed implications; during the switchover there will be no power delivered to the tank capacitor and you will see that at the output. I also think that the sudden increase in voltage (as opposed to the smoother increase as the MOSFET's gate charges will cause a current spike which will lower the transofrmer's voltage. Maybe these factors will become crtitical if your output is set for around 12V.

To do that with FETs you will need at least one more FET. The FET method will consume less power and maybe be a bit cheaper. Overall, I think this is a case of doing the same thing in a similar way using different components.

Share this post


Link to post
Share on other sites

Come on Hero, if you don't stop posting I will never go to bed and it is GMT+2 here  :o Besides, I am supposed to be away for Christmas. ::)

Yes, I noticed your post but had the self control to go to bed. ;D

Maybe it is worth adding a capacitor in series with the gate of M2, M3. This will waste some gate turn-on volatge (depending on capacitance ratios) but will decrease the gate capacitance that U1/U3 see.

I don't see how that will help.

As you said the turn-on voltage wll be higher which I think counteracts any benefit of having a lower capacitance, suppose the turn on voltage is doubled and the capacitance halved, we still have to wait just as long for the MOSGET to turn on.


Maybe we can select logic-level FETs for M2/M3.

Already done.

I have ticked 'hide from public' in my account, does public include registered members?

Yes it does hide your email from registered members. I think it's a good thing because it stops people asking questions via email; that's what the forum is for.

Yes so I run the simulation with ideal sources and ones with 0.4Ohm impedance and the difference is huge. That is what I meant by heavy dependance. With the percentages I meant what is the maximum tolerable peak-peak AC ripple as a percentage of the average DC output.

That will be the same with any power supply with rectifier and capacitor. There's nothing that can be done, short of adding active power factor correction which in that case I might as well make it an SMPS.

If I had to choose between more coil (higher volatge trafo) and thicker coil (higher current trafo) I would choose thicker coil as it will keep efficiency high, maintain tight regulation and not fry the op-amps (assume a primary voltage spike) at light loads.

That's why I've decided to go with 0 to 30V.

The output current will be limited above about 26V down to 2.4A at 30V, perhaps this can be done using software? Set the voltage above 26V and the maximu current limit is automatically reduced?


I've simulated a load transient 4.95A on and off  and a short circuit transient.

The turn on occurs at 1ms, there's some under and overshoot by about 36mV and lasts for microseconds.

The turn off transient is of a similar magnitude but there some ringing and it lasts longer.

The short circuit is odd. It immediately shoots up to about 475A, drops briefly to 350A, shoots up to 400A, decays to about 15mA for about 380

Share this post


Link to post
Share on other sites

It was only a thought has they used that in a Velleman 0-30V 0-10A psu which I did have one and never seen the voltage drop has it will be has quick has the mosfet, But if the mosfet is the better way to go then that's the way to go

how long do you reckon before we can build one for real and start playing

Share this post


Link to post
Share on other sites
how long do you reckon before we can build one for real and start playing


I was afraid this would come up any moment now.

perhaps this can be done using software? Set the voltage above 26V and the maximu current limit is automatically reduced?


Sure, but I would like to have the digital front-end as a retrofit. We could give different SW versions for different trafos.

The short circuit is odd.


Try adding some real-life ESR to C7.

I noticed you have eliminated the pot. divider for the current threshold reference. Are we going to adjust the current threshold with the dif amp now? I know you run out of voltage. Maybe you could decrease the gain of the dif amp?

We can use those IGBTs from the other topic to simulate a short-circuit when we have a prototype.

Could you post the latest asc. file?

Share this post


Link to post
Share on other sites
how long do you reckon before we can build one for real and start playing

You can build it now if you like but don't expect it to be any good.

I'm expecting there to be problems with the first build.

I remember designing something at work. I wanted to build a prototype but the boss wouldn't let me. It didn't work properly but I did tell him. He couldn't blame me because he reviewed all the drawings and signed them off.

Sure, but I would like to have the digital front-end as a retrofit.

Perhaps there should be a mark on the voltage dial indicating the reduced maximum current at higher voltage settings?

Try adding some real-life ESR to C7.

Done 50mΩ

I noticed you have eliminated the pot. divider for the current threshold reference. Are we going to adjust the current threshold with the dif amp now?

It's just temporary.

I've added another reference to adjust the voltage, for simulation purposes which will be replaced with a pot. for the final schematic.

I know you run out of voltage. Maybe you could decrease the gain of the dif amp?

I might increase the gain of the current error amplifier slightly.

I like the idea of having a 1V/1A output because it might make it easier to add a meter. If it turns out to have been a bad idea it could be changed later.

We can use those IGBTs from the other topic to simulate a short-circuit when we have a prototype.

If you like.

I normally use MOSFETs but I can see you're itching to find a use for those IGBTs.  ;D

Could you post the latest asc. file?


See attached.

I've found an op-amp, the LT1014, it's reasonably priced, low offset and has a high CMRR but it's a little slow: the transient response is worse now.

I haven't decided on the MOSFET yet, the problem is LTSpice doesn't come with any models for TO-220 MOSFETs. I think I'll just use the IRL540 and won't worry about the model.

0_to_30V_5A.asc.txt

Share this post


Link to post
Share on other sites

I am just looking at the asc file.

Maybe you should separate Vref for the current limit and use a third source for R12/R13. In fact, how about using a constant reference for U4 as that doesnt need to be variable. How about a circuit around D8?

Current limit attack time? Are we strategically ignoring this at the moment?

Also, I noticed you used a time dependant current source as the load. I think a resistive load would be better as it is completely passive. The way I have done this before was to use a time dependant voltage source to connect low value resistors to the output through a MOSFET. I am sure there is a better way to do this.

Finally, I see you added ESR (50mOhm) to C1, not C7 as I was asking in a previous post. I was thinking that that 400+A short circuit current might have been due to the 0 ESR for C7.


EDIT: The output voltage now seems to be 6*Vref. Any ideas what happened? Also, what is the output impedance of U1 in this configuration?

Share this post


Link to post
Share on other sites
Are you talking about finding a transformer which will allow 0 to 30V @ 5A?


Yeah, that is what I meant. Since using a higher volatge trafo will require more parts, in the final project it would be nice to have a trafo that can offer 5A without output ripple.

If most people here decide it's a must, then I'll look at redesigning for a 30V transformer. This will probably mean ditching the LDO MOSFET idea and going back to a more traditional transistor based approach.


How about arranging for the zener diode regulator across +V+6 and ground by increasing the zener voltage and recalculating the resistance? This will waste more power on the zener (series diodes?) but will allow a 30V trafo and you get a regulated supply for the op-amp.

I have to justify the space they take to myself...

Put them on ebay?


I tried in the past. The problem is that most people buying from ebay can't afford them and those who can, don't buy from ebay if you know what I mean. Might try again in the future.

I've decided on the IRLZ34N

Anything. That is one cheap MOSFET btw..

RS is cheaper for the 10,000μF

Anything than a single 22000uF/63V cap.

It's probably a good idea to put footprints on the PCB for two capacitors as well as one and different diameters if convenient.


Good idea since this is a kit.

I would have thought simulation would be good enough since the simulator performs loop analysis anyway.


Can LTSpice do Bode plots? I am asking, I don't know. We can optimise some of the components this way but as you say good enough. We can simulate some reactive loads too later on.

Surely it's more cost effective to use one voltage reference IC and obtain other voltage references using potential dividers?


Sorry, I get confused with multiple sources on the diagram.

The original idea was to use an LM78L05 but then I decided the tolerance is too wide.

Good thing you rejected that.

Now I'm thinking of using an LM431B which has a tolerance of 1%. I could use the C version which is 0.5% but it's probably harder to get hold of and will be more expensive.


It is not like we have much choice in TO-92 at this tolerange range... I looked on RS and Farnell, most products are the 431 from different manufacturers ::) 1% sounds reasonable. We should consider how much effect on the output a 1% reference error will have compared with using 1% resistors for the potential dividers. That said, TI's TL431 says A=1%, B =0.5%. The TL431A costs only 30p on Farnell and you dont have to buy a bag of 50 like on RS.

How about the LM317L? Just an idea. Never mind, the 1.25V bandgap reference is at +/- 0.05 V i.e. 4%.

I'm thinking 1% for all the gain and reference setting resistors and 5% for everything else.
Close tolerance is nice but <1% starts to get expensive and harder to get hold of.


Fully agree, 5% can also go for R16,17. 1% resistors are used more and more frequently so if we say 1% throughout I dont think the cost difference would be prohibitive, if any. It saves the effort of saying this 5% that 1% etc.

Share this post


Link to post
Share on other sites
How about arranging for the zener diode regulator across +V+6 and ground by increasing the zener voltage and recalculating the resistance? This will waste more power on the zener (series diodes?) but will allow a 30V trafo and you get a regulated supply for the op-amp.

Done.

A 39V zener such as the BZX85C39 or 1N4754A.

I've simulated this with three zeners in series giving 40V which is near enough.

I might as well upgrade the voltage to 33V as the simulation clearly shows it's possible.

Can LTSpice do Bode plots? I am asking, I don't know. We can optimise some of the components this way but as you say good enough. We can simulate some reactive loads too later on.

It does do Bode plots but it's not something I've used before so I'll need to have a play.

It is not like we have much choice in TO-92 at this tolerange range... I looked on RS and Farnell, most products are the 431 from different manufacturers ::) 1% sounds reasonable. We should consider how much effect on the output a 1% reference error will have compared with using 1% resistors for the potential dividers. That said, TI's TL431 says A=1%, B =0.5%.

I'll say TL431A or LM431B on the BoM

I'll say <1% for all of the reference and gain setting resistance and <5% for the rest. I don't like setting unnecessarily tight tolerances for anything. If it were me building it, I would just use 1% for everything but it isn't going to be me building it. I'll make it clear that 1% can be used instead of 5%.

I'm also sticking to E12 values where ever possible as I know lots of people only stock E12 values.

The TL431A costs only 30p on Farnell and you dont have to buy a bag of 50 like on RS.

RS are a bit more expensive but you can buy just one of them.
http://uk.rs-online.com/web/search/searchBrowseAction.html?method=getProduct&R=0428436

EDIT:
Here's the latest schematic.

Share this post


Link to post
Share on other sites
I might as well upgrade the voltage to 33V as the simulation clearly shows it's possible

I would say limit it to 30V has the extra 3 volts woudl not make that much difference

I'll say TL431A or LM431B on the BoM

I totaly agree has it's simple and an easy to obtain the part, Cost does play a apart in slightly but it want want to build a quality PSU then cost does not matter to much, After all if you had

Share this post


Link to post
Share on other sites
I would say limit it to 30V has the extra 3 volts woudl not make that much difference

In theory that's fine.

In practise, I'll always design the output voltage to be a little be higher than 30V because it's possible that component tolerances could limit the voltage to less than 30V. I could include a trimmer so the user could trim to exactly 30V but I don't see the point.

I'm glad that you've stuck with the 5.6V reference this makes it easier to make it a digitally controlled PSU with no extra parts like bossting the voltage,

Would the 0.6V differnce make that much difference in the max output voltage/current has you can only get 5V out of the MCU, I now you could use a DAC but just aads to the cost and makes the a little harder in the design of it has I've never used DAC's so not reallt sure

I could easily go back up to 5V but I'll have to increase the gain of the current and voltage amplifiers slightly.

The problem boils down to tolerances again, with a 5V reference and the current gain of 1A/1V it's possible the current limit will be less than 5A.

Will there be a LED for the visual indication for over current, Would it be to hard to add some LEDs so you can see if the PSU is supply constant volatge or constant current only I've seen some of these on other deisgns

That's a good idea, I'll look into it.

I've just looked at the links for the cap's you posted is 40V cap going to be big enogh ?

That was selected when I was going to use a 24V transformer, with 30V it's going to be increased to 50V.

I was also thinking would it be better to have the bridge rectifier and cap mounted of the PCB, This would save people hunting the right Cap & rectifeier and aslo reduce the size of the PCB.  This is the way I would consider doing it.

That's a possibility. I'm going to leave that up to the person designing the PCB.

Another edit:Now I'm not sure if it's something over looked but when you have got the current pot turned down to zero(min current) you get no output voltage I have to set the pot to about 2% before I get voltage that's 0-15v then it will not go any higher than 20volts until I take the current pot to 3% to obtain 30Volts so to me the current limit is active even with no load.

That's pretty normal for power supplies I've seen except the current limit pot. only has to be adjusted to slightly above zero and the output suddenly jumps to the full voltage setting.

I suppose it's possible for it to be unstable if the user selects really low currents <1mA due to noise.

Share this post


Link to post
Share on other sites

The velleman power supply I have got had  2 pots that you had to calibrate min and max, the had to be set for about 2-6ma and the max was for 5amps, I suppose that's how they got around it so when the main current limit pot turned turned to zero the calibration pot prevented the volts from droping to zero

In practise, I'll always design the output voltage to be a little be higher than 30V because it's possible that component tolerances could limit the voltage to less than 30V. I could include a trimmer so the user could trim to exactly 30V but I don't see the point.

I never thought about that bit, It maks sense now  ;D
The problem boils down to tolerances again, with a 5V reference and the current gain of 1A/1V it's possible the current limit will be less than 5A.

I had a play around with the values of the TL43B to make vref 5.01 and the max current went to 5.01 where has before it only went to 5.63 has that what the vref voltage was, Would it possible just to alter the gain slighty so we can control it with 0-5V (4.95V), may be this where the min and max pots could be used just incase you overshoot 5amps and at the same time this will cure the volts problem with the current pot turned to 0

Share this post


Link to post
Share on other sites

LTSpice is annoying me.

I'm testing the stability of the over-current protection by connected it to a constant voltage linear ramp.

The first thing I noticed is it was taking ages. I attributed this to it being unstable when the power supply voltage is really low. To get round this I set the phase on the voltage sources to 90

0_to_30V_5A.asc.txt

Share this post


Link to post
Share on other sites

I'm simulating mine in ISIS and everything seems to work OK, To get it to Sim correct I've had to remove M1 and replace with relay has I could not alter the output voltage has the way it is drawn it always stayed at max output voltage, The way it is set up at the moment to me is working great according to the Sim, The voltage goes from 0-30.2(this is because I got veref set to 5V), I then close the switch on a 0.1R resistor and then alter the current from 0-5.01 amps the voltage drops right off to 0.16V, I have also tried bigger value resistors and gone through the range and the voltage/current remain stable,

I've tried to add the 2 100k resistors as per your latest drawing and  it will not Sim so I know that by adding these 2 resistors that cause's the problem, I removed them and everything back to working, Now this where I get confused and how much faith do you put in these Sim programs, Normally they are spot on if it does not work in ISIS then it does not work in the real world. Some times I get problems in simulating but in the real world everything OK  ???
My drawing attached, I've not really used LTspice at all, it is all new to me so i'm just playing and learning how do you alter the pots in LTspice

new_design_psu.pdf

Share this post


Link to post
Share on other sites

So much information...I was only away for like 3/4 of a day!

I would overrate the outputs (I, V) slightly to ensure that despite tolerances we have 30V, 5A.

This has shaken my confidence in LTSpice.

Welcome to the club.

Will there be a LED for the visual indication for over current...

I think adding LEDs and indicators is trivial and could be left for later on.

I expected the minimum current setting to be 25mA maximum

Pretty high?

I was also thinking would it be better to have the bridge rectifier and cap mounted of the PCB, This would save people hunting the right Cap & rectifeier and aslo reduce the size of the PCB.

IMO the best of both worlds would be to put multiple cap sizes on the PCB. The same for some other components.

On a different note, I was thinking of the digital front end. An I2C bus would be nice as we can hang all digipots on it, plus say a temp sensor or a fan controller.

Thanks for the latest asc file. PICMaster, you have a relay in your schematic ??? Sorry I can't spend more time on this despite wanting to.

Share this post


Link to post
Share on other sites
Thanks for the latest asc file. PICMaster, you have a relay in your schematic  Sorry I can't spend more time on this despite wanting to.

Yeah this is for testing purpose only, It was the only way that I could get to simulate it because with the mosfet in there i was only getting max voltage out and no variation in the output. I know that simulators can sometimes play up causing all sorts of problems let's just hope this does not happen real life.

IMO the best of both worlds would be to put multiple cap sizes on the PCB. The same for some other components.

Yes may be but we would have to complete the bom and source the components and the data sheets to get the correct sizes so the footprints can get made.

On a different note, I was thinking of the digital front end. An I2C bus would be nice as we can hang all digipots on it, plus say a temp sensor or a fan controller.

Why digital pots, You could just use a op-amp as a buffer/protection and read the the voltage/current,the temp and control a fan all at the same time, But if you like the idea of digital pots then may be we could have 2 interface add on's  ;D

think adding LEDs and indicators is trivial and could be left for later on.

Yes I know it's sounds trivial but needs to be thought of, Has for the LED's to state constant current or constant voltage is not so important but we need to add all the stuff soon before we come wup with a working prototpye.
Next year  ;D (week) I shall get the op-amp and a few other bits ordered ready to start biulding that's providing we stick with the chosen op-amp,

Well this has been very interesting and would love to continue working on this project but the wife telling me to get ready to go out,

Happy new year to you both and We'll speak next year  ;D ;D

Alex what language do you write your code in ?

Share this post


Link to post
Share on other sites

Pretty high?

I agree, I would have hoped it would be lower but I looked at the datasheet and the worst case is 25mV, 10mV  is typical and 5mV typical with 600R resistor.


I'm simulating mine in ISIS and everything seems to work OK

It doesn't look right to me, the PDF you posted says the voltage before the MOSFET is 88.8V which is much too high.

I've tried to add the 2 100k resistors as per your latest drawing and  it will not Sim so I know that by adding these 2 resistors that cause's the problem, I removed them and everything back to working,

So you're having problems with both of the pot's?

I'm only having a problem with R24 on my schematic. It works fine without it but goes wrong as soon as it's added.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
Sign in to follow this  

×
×
  • Create New...