Jump to content
Electronics-Lab.com Community

Design new 0 to 30V power supply.


Hero999

Recommended Posts

As you all probably know, the 30V power supply project on this site has many errors and design flaws.

I'm thinking about designing a new one from scratch but before I do, I'd like to know whether anyone will test it for me?

It's not something I want to build myself because I don't need it. I could breadboard parts of the design but I won't be able to completely test because I don't have all the components so I can't fully test it.

I've already nearly designed it in my head but I haven't committed anything to paper so I'll write a list of decisions I've made so far:

The design shall use a 12-0-12V transformer, when operating at voltages below 13V the power will be taken from the centre tap in a bid to save power thus cutting the heatsink size in half.

It will be a LDO regulator with a couple of N-channel MOSFETs (e.g. IRL540) used as output transistors, I thought about using P-channel devices but gain stability can be a problem and they're more difficult to get hold of.

The current sense resistor will be a very low value <50Ohm to minimise the voltage loss.

The bridge rectifier will consist of Schottky diodes to minimise the voltage losses.

Because the output device is an N-channel, the gate voltage will need to be higher than 30V. This requirement will be met by using a voltage doubler and zener regulator to power the op-amp.

I have decided to use an MC34074 op-amp because: it can be operated from a single supply, has a high voltage rating (44V), is easy to get hold of and cheap.

Perceived risks:
The current sense resistor might prove difficult to get hold of. This can be overcome by paralleling several higher value resistors.

The filter capacitor will need to be very large 22,000

Link to comment
Share on other sites


  • Replies 129
  • Created
  • Last Reply

Top Posters In This Topic

What would you like to have tested Hero?

I would like to see a setting for the attack time of the current limiter, a feature not found on cheap or DIY power supplies.

50 Ohm still sounds large to me, how about something in the 10's of milliohms range and a low input voltage offset differential amplifier? LT does such ICs, maybe it is worth having a look there.

Link to comment
Share on other sites

I meant 50mOhm  ;D

I like the timer on the current limiter. It's probably easier to implement it as a simple RC time constant in the feedback rather than a timer.

A low offset amplifier sounds like a good idea but I would rather have something pretty standard, rather than a proprietary IC only made by one manufacturer. Perhaps I could start with Linear and find other ICs from other suppliers to suggest as substitutes.

Are you saying it doesn't need to be tested and that me designing it, simulating it and submitting it to a peer review on this site should be enough?

Ideally I would say, yes it doesn't need to be tested but I know from experience about how unpredictable things can be, especially when you factor in people building it with different components from different manufacturers.

Link to comment
Share on other sites

OK, 50mOhm is not tragic. I agree with standard components for this project. RC constant is the first thing that comes in mind here too. I dont think more than two settings will be needed, one slow and one very fast, so maybe a simple SPDT switch that connects a resistor in parallel will be sufficient.

No, it needs testing no matter who designed and reviewed it. What I meant was, what will you need for your testing? Are you going to built it on a breadboard/perfboard and have someone else test it? Submit a schematic, someone builds it and tests it? etc.

As long as people use the components in the BOM and havent messed up, it must work, and that is what we should aim for. Now, if there are some non 555-like parts, you could offer an alternative and test that config too. But you cant test every possible config. You will invest time developing the circuit around some specific parts therefore whoever builds it must invest time in finding the parts, possibly with help from us. Otherwise your efforts are in vain.

Have a look on Maxim-IC, their sampling system is like Santa Claus these days.

Link to comment
Share on other sites

I agree, it needs testing, as well as a peer review.

There are other components such as a large 24V transformer or a 3A 35VDC power supply.

I could test it at lower currents and voltages using my 1.5A 30VDC PSU but it't probably isn't worth it.

My intention is to build it, post the schematic and BoM, you lot can review it and one of you can test it, then it can become an official project or at least be put on Silicon tronics.

Link to comment
Share on other sites

I have a dual 36V, 4A PSU that can be used to simulate the transofrmer and PSU. However, we will be deviating from the idea of testing the whole prototype.

Don't worry about testing, you have me willing to do it if noone else is keen, so I would tick that on your list. I would like to test the entire prototype if possible, maybe help with PCB design?

Link to comment
Share on other sites

Alex,
The trouble with using a stabilised supply is it won't test the ripple rejection.

PicMaster,
That looks like it aught to do the job.

Attached is revision 1, it's not finished.

As you can see, the power supply for the op-amps is taken from a voltage doubler. This is required because the gate voltages for the MOSFETs need to be above the supply to turn them fully on.

U1 is configured as a non-inverting amplifier with a gain of 5.1, M2 and M3 connected in the feedback loop.

U2 is a differential amplifier with a gain of 66 2/3, the output voltage is equal to 1V for every Amp of current drawn.

U3 is an error amplifier which is responsible for the current limiting, when the voltage from U2 exceeds 3V, the gate signal to M2 and M3 is diverted to 0V, D10 prevents the U2 from trying to turn on M2 and M3 more when the current is under the limit. C5 and R18 provide phase conpensation to prevent oscillation.

U4 is a comparator, it's responsible for switching to the higher input voltage as explained below.

When the Vout is under about 14V, the power is taken from the transformer's (simulated by V1 and V2) centre tap via D5. When the output voltage rises above around 14V, U4 turns on M1 which connects C1 to the higher voltage side of the rectifier. R11 provides some hysteresis which prevents oscillation.

The circuit is not finished. It needs some of resistors replacing with potentiometers and a way to delay the current limiting.

What sort of delay do you thing would be acceptable?

Maybe the compensation capacitor could be moved to the differential amplifier and another capacitor added U3. A pot. could be used to adjust the delay.

Link to comment
Share on other sites

It's looking vey nice,  ;D
What op-amps are you thinking of using ?

Where does IOUT go that feeds the op-amp ?
Aslo which one aught to do the job 15 + 15V(sorry I put this down has 30V) or the 12+12V ?

Has for the delay I would think it should be fast enough to trip If I understand it correct, I let you guys think that one out.  

EDIT: I 've just reread the topic and I think I know which op-amp and transformer needed but just like to make sure

Link to comment
Share on other sites



What sort of delay do you thing would be acceptable?



A few 10s of ms I would think.
How about a capacitor between the output pin of U2 and ground through a variable resistor? No further modifications needed.

You can cover a huge range with a reasonably sized capacitor as you can always set R=0 with the variable resistor. Wiper debouncing will be taken care of by the capacitor following it.


Link to comment
Share on other sites


It's looking vey nice,  ;D
What op-amps are you thinking of using ?

I haven't decided yet.

At the moment I've used the generic op-amp model found on LTSpice.

Where does IOUT go that feeds the op-amp ?

Guess.

Hint: it isn't an input, it's an output.

Has for the delay I would think it should be fast enough to trip If I understand it correct, I let you guys think that one out. 

As drawn the RC time constant is 10
Link to comment
Share on other sites

My main concern was latch-up of switch-mode PSU controllers when a bath-tub voltage input is applied, hence the delay. I think that the actual delay time will require experimentation in the particular application. Most circuits will not mind whatsoever, maybe a few MCU design will fail to reset. A setting could exist that applies a delay as a function of current drawn, like in an inverse time overcurrent relay. Maybe this is an overkill as the currents involved are rather small and ultimately limited by the transformer.

An adjustable 0-100ms adjustable rise time on the output of the current sensing dif. amp should be enough for most applications. I expect the output resistance to be too low to allow for a reasonably small capacitor, but you can subtract it from the resistor value (tolerances?) There is a compromise here, how much time your tank caps take to charge and how much time an IC or transistor at 3A takes to blow up.

I did look at the schematic, particularly interested in topology. Would you share the LTSpice file? I can have a block-by-block look then.

How does the PSU behave with a step on it's output, say 10% to 100% current and back? How about adding some resistance/capacitance to the sine sources?

Link to comment
Share on other sites

Here's the .asc file and .asy file, just remove the .txt extension.

I realised that adding a potential divider or switch to the summing junction of an op-amp is probably a bad idea unless it's directly mounted on the PCB or connected with screened cable.

I haven't simulated any transients yet or changed the internal resistance of the sine wave generator.

0_to_30V_3A.asc.txt

Upside_dwn_Universal_opamp_.asy.txt

Link to comment
Share on other sites


Ok, we need the universal op-amp asy and the sub file it points at.
Also, the upside-down op-amp asy and the sub file it points at if different.

Zip and attach if possible.


Both the upside down and universial op-amp symbols point to UniversalOpamps.sub which should already by in the sub directory, it's a standard LTSpice library which should already be there.

I got the file name wrong earlier, it's Upside dwn Universal opamp .asy I missed the space which was probably a mistake in the first place but I'm not going to rename the file because lots of other schematics use it.

Yes zipping is probably better for large files, code tags are only good for very small files. I've renamed them to .txt and edited my previous post.
Link to comment
Share on other sites

Right, I will give them a try. In the meanwhile, what are the output specs of the project? 0-30V 0-3A? I was thinking, since you have engaged into designinmg a new one why not juice it up a little bit, say 50V 5A?

I was also thinking how nice a digital front-end would look like, a small LCD with up/down buttons to set current limit, attack time and output voltage. I am very keen on a portable powerful psu with a fancy front panel.

Link to comment
Share on other sites


Right, I will give them a try. In the meanwhile, what are the output specs of the project? 0-30V 0-3A? I was thinking, since you have engaged into designinmg a new one why not juice it up a little bit, say 50V 5A?


30V@3A was the original design spec.

Increasing the current capacity won't be a problem but the a voltage would be. The op-amp needs to be powered from the output voltage plus 6V to turn the output MOSFET on. The trouble is most op-amps are rated for 44V or less.

I'm sure it's possible to increase the voltage to 50V but it'll need a total re-design: a 40V centre tapped transformer (does such a beast exist?), P-channel MOSFETs (this will make it less stable), and moving the current sensing resistor to the low side.

I was also thinking how nice a digital front-end would look like, a small LCD with up/down buttons to set current limit, attack time and output voltage. I am very keen on a portable powerful psu with a fancy front panel.

Unfortunately I don't have enough experience with micro-controllers to do that but adding such an interface shouldn't be hard for someone with the know how.

At the moment, the regulator uses a 5V voltage reference, see Vref on the schematic I haven't decided on an IC yet.

The output voltage is simply Vref*5.1 giving a voltage of 30.5V. I was intending to use a potentiometer but there's no reason why the Vref can't be the output of a DAC.

The curent limit is set by R14 and R15 which form a potential divider and is simply equal to the voltage across R15. At the moment it's set for 3A, I'll probably increase the maximum current limit to just above 3A to allow for tolerances. Again, I was intending to use a potential divider and it'll be easy to connect to a DAC.

Iout outputs a voltage proportional to the current, just connect it to an ADC.

A programmable current limit delay will probably me more tricky, maybe you could use a digital pot. but I don't know if you can get them which can be interfaced with an MCU?


I have associated all the file names and all the symbols load correctly. However I get the following error. See attached. Any ideas? Could you post your universalopamp.sub file?

See attached.

UniversalOpamps.sub.txt

Link to comment
Share on other sites

How much current can you draw without a total redesign of the topology? By upgrading all components in the main current path we can easily get >5A. This will be very useful. 10A would be ideal.

I think you meant Vout = Vref*5.1 + Vref = 6.1*Vref = 30.5V

I was thinking of digipots as a retrofit but DACs will probably be cheaper.

I have noticed you changed all op-amp supplies to +V+6V. Are you considering using a quad op-amp to keep the pcb compact?

I still get the SPICE error. Not sure what is wrong.

Link to comment
Share on other sites

Unfortunately I don't have enough experience with micro-controllers to do that but adding such an interface shouldn't be hard for someone with the know how.

I can help out there, I have the code working to a degree, I send out a smooth stable volatge of 0-5V with a coarse and fine settings, This can be sorted out once we have a working power supply, we can just include it on the PCB this way people can a have a choice what they like to build.

A programmable current limit delay will probably me more tricky, maybe you could use a digital pot. but I don't know if you can get them which can be interfaced with an MCU?

Yes they can be interfaced with MCU but can only handle the same voltage has MCU which is 5V, I've also code some code some where to adjust them.

I've never has the need for 50V supply and not sure that many others would either, The PSU I'm running at the moment has never been above 13.8V really well once it went to 28.8V this was just to set up a battery charger timer card, If you think about it you could just build 2 and you can get 60V,

Has for the display we can either use a LCD/GLCD or 2 x 4 7 segment displays or even both the leds will read actual voltage/current the LCD could display the max voltage/current settings and Temp, These are just ideas and can be added later if needed

Link to comment
Share on other sites


How much current can you draw without a total redesign of the topology? By upgrading all components in the main current path we can easily get >5A. This will be very useful. 10A would be ideal.

More and more MOSFETs can be added.

The limitation is that the increased gate capacitance will slow things down.

As drawn, the total gate capacitance is 3.74nF with R4 = 10k the time constant is 37.4

0_to_30V_3A.asc.txt

Link to comment
Share on other sites

I agree with PICmaster that most projects stay below 12V or so, but the higher the output voltage range, the better. That said, I do realise that with a clean design like this, increasing the output voltage is quickly limited by factors such as op-amp supplies. This design must remain clean to be attractive and cost-effective therefore more than one PSUs can be connected in series. However, as they are regulated, connecting them in parallel will cause problems so we should try to squeeze as much current as possible, without having an array of MOSFETs or a brick of a transformer.

In my opinion 3A is on the low-side. 5A would be nice, 8A would be excellent, 10A and you have an outstanding PSU, capable of powering most projects (eg an audio ampilfier).

Hero, I was not able to find the MOSFETs on the IRF website, they are not real parts, are they?

I would prefer using a tad more expensive MOSFETs with lower RdsON than using more MOSFETs to keep it compact. Regarding gate current, a driver IC can be used (non standard?) or maybe an op-amp connected as a buffer. But then you need an extra device on the PCB as you already have 4 op-amps.

Here it is with the standard symbol, it doesn't look as good but it should work.


Nop. :(


I think 10% regulation is quite high for a power supply where a constant voltage source is important. Anyway, I calculated the no-load RMS output voltage as 13.3V for a 12V transformer rated at 3A with 10% regulation. Assuming sine wave, that is 18.86V peak, not quite sure where the 0.2V discrepancy is coming from, maybe rounding error? We both used a calculator for sqrt(2) so 0.2V is quite high.

One thing to bear in mind here, is that the regulation will change as the power factor from the rectifiers changes, you are switching between half and full bridge rectification.

(13.3VRMS-12VRMS)/3A=0.43 Ohms, so the secondary impedance calculation is right, you didn't over-estimate it.

What I realise is that I used the word impedance and not resistance. The secondary will also have inductance that, unlike it's resistance, will not waste energy as heat but send it back to C1. I am not sure as to what range of secondary inductance such a transformer will have. So what I did is assume that the secondary inductance exhibits an inductive reactance that is equal to the secondary resistance and therefore equal to Zsec / sqrt(2) = 0.43 / sqrt(2) = 0.3 Ohm

The inductive reactance of an inductor is |Z| = ω L [Ohms]. Plugging 50Hz and 0.3 Ohm into that and rearranging, I got an inductance of 9.67 E-4, i.e 967 μH.

So, Hero, try simulating the transformer with two 18.8V peak sine sources each with an inductor of 967uH in series with a resistor of 0.3 Ohm. Sorry I can't help with simulation my LTSpice has a will of its own.

I once again say that 10% is rather high. With such high(bad) regulation you can improve performance by increasing the output voltage of the trafo and 'sacrifice' it under load. But you will have housekeeping PSU problems when under light load. Otherwise, you can maybe increase the current rating thus decreasing secondary impedance and improving regulation?

You could sacrifice efficiency by using only the full bridge rectifier to drastically reduce the peak rectifier current. But I quite like that topology, and considering all of the above I think a higher current trafo is needed.

To clear your mind from phazors, I think we must concentrate on the analogue part which is what will define the performance of the whole PSU. Digipots, LCDs, DACs, ADCs are there and will allow a digital front end but will not affect things like closed-loop stability and output ripple, which are what the load will care about.

Hero, could you zip and email me your lib folder and its subfolders and files?
Link to comment
Share on other sites

In my opinion 3A is on the low-side. 5A would be nice, 8A would be excellent, 10A and you have an outstanding PSU, capable of powering most projects (eg an audio ampilfier).

Each to their own I suppose. I thing 1.5A is good enough for most applications, for high powered stuff a regulated power supply becomes less important so I make my own unregulated supply using a Variac, a mains transformer, a rectifier and a filter capacitor.

Now as this supply isn't going to be for me, it's whatever everyone here needs.

Maybe I could design it for 5A and include some modifications to increase it to 10A?

Don't forget even 5A would require a 250VA transformer which will be quite large.

Hero, I was not able to find the MOSFETs on the IRF website, they are not real parts, are they?

It's just a MOSFET in a TO-220 package I could find on LTSpice. The datasheet can be foud on Google easilly enough. I don't know how easy it is to buy though.

I would prefer using a tad more expensive MOSFETs with lower RdsON than using more MOSFETs to keep it compact.

I would agree in general but I think common and easy to get hold of is more important here, besides power dissipation is the main problem here, not on resistance or current capacity.

Regarding gate current, a driver IC can be used (non standard?) or maybe an op-amp connected as a buffer. But then you need an extra device on the PCB as you already have 4 op-amps.

Perhaps I could add a transistor to the current limiting section but it'll probably need more phase compensation as it would increase the gain.


I think 10% regulation is quite high for a power supply where a constant voltage source is important.

I was talking about a typical 100VA transformer. I hope the power supply will have a much tighter regulation factor than 10%.


Anyway, I calculated the no-load RMS output voltage as 13.3V for a 12V transformer rated at 3A with 10% regulation. Assuming sine wave, that is 18.86V peak, not quite sure where the 0.2V discrepancy is coming from, maybe rounding error? We both used a calculator for sqrt(2) so 0.2V is quite high.


12V + 10% = 13.2V

If a 3A load is applied, the voltage will drop back to 12V, 13.2 - 12 = 1.2V

The impedance is 1.2/3 = 400mOhm.

One thing to bear in mind here, is that the regulation will change as the power factor from the rectifiers changes, you are switching between half and full bridge rectification.

No, it always uses full-wave rectification. When running from the centre tap the rectifier is configured as a biphase rectifier, it's still full-wave the transformer ensures that, when switched to the higher voltage, it's configured as a standard bridge4 rectifier. I wouldn't even consider drawing such a large load from a transformer using a half-wave rectifier because it will cause core saturation and meltdown.


What I realise is that I used the word impedance and not resistance. The secondary will also have inductance that, unlike it's resistance, will not waste energy as heat but send it back to C1. I am not sure as to what range of secondary inductance such a transformer will have. So what I did is assume that the secondary inductance exhibits an inductive reactance that is equal to the secondary resistance and therefore equal to Zsec / sqrt(2) = 0.43 / sqrt(2) = 0.3 Ohm

The inductive reactance of an inductor is |Z| = ω L [Ohms]. Plugging 50Hz and 0.3 Ohm into that and rearranging, I got an inductance of 9.67 E-4, i.e 967 μH.

So, Hero, try simulating the transformer with two 18.8V peak sine sources each with an inductor of 967uH in series with a resistor of 0.3 Ohm. Sorry I can't help with simulation my LTSpice has a will of its own.

That doesn't look right to me.

I would have thought the only inductance would be the leakage inductance which will be negligible.

The primary and secondary resistances will dominate. The resistance seen a the secondary will be equal to the Rs + Rp/TurnsRatio.

Even if part of the impedance is inductive it'll still drop the voltage by limiting the current pules, it's true that power won't be wasted but that doesn't make that much difference for the purposes of this exercise.

You could sacrifice efficiency by using only the full bridge rectifier to drastically reduce the peak rectifier current. But I quite like that topology, and considering all of the above I think a higher current trafo is needed.

I think tap switching is the only sane way to build a high powered linear power supply. I would like to use a transformer with as many taps as possible but it wouldn't be a standard part.

Hero, could you zip and email me your lib folder and its subfolders and files?

Done.

EDIT:
I couldn't email you so I've uploaded it to Silicon Tronics.
http://www.silicontronics.com/Stuff/lib.zip

Don't expect it to be there forever, it's just a temporary thing.

EDIT:
Here's a 35V 5A power supply.

The effect of the extra MOSFET capacitance has been reduced by bypassing R4 with 100nF which probably has slowed the over current limit response time.

The transformer model is 225VA which has a regulation factor of 8%.

It just about manages 35V and 5A simultaneously but I woundn't bet on it in real life; the ripple valley is very close to the output voltage.

I also wouldn't bet on my make shift high voltage zener (Q1, D8, R14 & R15) to have a votage drop below 44V, the maximum rating of the op-amps.

I think we need to accept that the output voltage and current won't be available simultaneously.

I'm tempted to go back down to 30V as the extra 5V isn't worth the trouble.

I can still keep the 5A requirement, it's just that beyond about 27V the voltage will drop and there'll be ripple, at high currents.

0_to_35V_5A.asc.txt

Link to comment
Share on other sites


5A it is then.

I have obviously looked on Google and IRF.The closest matches are the IRLR3915PBF from IRF, which  comes in a D-pak and the IRLU3915PBF which comes in an I-pak. Where did you get the TO-220 pack? Meh, I am sure we will find a candidate.

You're right, I thought it was TO-220 but the datasheet clearly states SMT.


37.4
Link to comment
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...

×
  • Create New...