Help me design a PSU

[not1xor1]

Here's  one from Banggood (it's about $6). Someone posted a schematic.. Funny drawback of producing a negative rail with a bucket load of cheapo 1N4007's is the 1 amp maximum. Real efficient heater though..

It is easy to make a much better (more stable, more efficient) PSU with about the same amount of components (dual opamp and power transistors), but if current regulation is not important even a TL431 + darlington would work.
A negative rail and another TL431 would allow to start voltage regulation from 0V.
It would be wise anyway to add a foldback current limit via an additional BJT and resistor.

image link removed  - see above for the schematic

This schematic doesn't seem to have constant current mode. Adding one will make it better since it is already easy schematic. I guess the purpose of Q1 is for short circuit protection, is it correct?

I am not a fan of having negative rails from diodes due to the heat and other issues. My initial suggestion was a TL431 + ICL7660 which is probably the best choice.

the original poster stated: no current limit needed
I anyway added a foldback current limit as I think it is the bare minimum for safe operation.

The diodes used to produce the negative rail dissipate few mWs because the circuit is just a charge pump.
There is absolutely no need for a further IC which might introduce just more noise besides the complexity and the cost.

[not1xor1]

i think that Q1 is for short circuit protection too. But i didnt understand does this circuit have negative output? Can i build this thing with opamps instead of programmable zeners and 2n2222s as transistors
Or bd138 pnps?

The negative rail is not meant for outup.
It is there just to let the output range from 0V.
It does that by bringing the "anode" of the top TL431 to -2.5V.

Of course it is possible (and better) to use an opamp to do the same, or even a dual opamp to provide proper current regulation.
In my spare time I'm testing (just on LTspice at the moment) a very simple circuit inspired by that designed by NSC during the 70:
https://www.eevblog.com/forum/projects/very-low-noise-preregulator-for-benchtop-power-supply/msg1375039/#msg1375039

My circuit is based on modern ICs (LT1013) and is simpler than the original.
I just want to make some more tests before posting it here.

[not1xor1]

There is mention of a 48 VAC transformer above and the output, when rectified, is about 62V (plus or minus the filter capacitor and current draw/voltage drop.  Nevertheless, a really big number.

When the PS is outputting, say, 1A at 1V, the pass transistor(s) drops 62V - 1V and dissipates 61 Watts (more or less).  That's also a really big number.

That's why real power supplies use transformers with tapped secondaries and some number of relays that kick in at various voltage settings.

If you want to output 5V, there is no point in having more than 8v DC going into the pass transistor (more or less).  Clearly, not 62V...

IMHO the power is still manageable by using more power BJTs and a big heatsink.
The main problem is rather the high voltage which would require a much more complicated PSU design, with a floating supply for the opamp and less common BJTs (i.e. the high voltage version of 2N3055 or something like the 2SC5200).
I think a beginner should start with a simpler design and a more manageable voltage and power, i.e. something around 15-20V 1A.

[rstofer]

There is mention of a 48 VAC transformer above and the output, when rectified, is about 62V (plus or minus the filter capacitor and current draw/voltage drop.  Nevertheless, a really big number.

When the PS is outputting, say, 1A at 1V, the pass transistor(s) drops 62V - 1V and dissipates 61 Watts (more or less).  That's also a really big number.

That's why real power supplies use transformers with tapped secondaries and some number of relays that kick in at various voltage settings.

If you want to output 5V, there is no point in having more than 8v DC going into the pass transistor (more or less).  Clearly, not 62V...

IMHO the power is still manageable by using more power BJTs and a big heatsink.
The main problem is rather the high voltage which would require a much more complicated PSU design, with a floating supply for the opamp and less common BJTs (i.e. the high voltage version of 2N3055 or something like the 2SC5200).
I think a beginner should start with a simpler design and a more manageable voltage and power, i.e. something around 15-20V 1A.

Yes, it can be managed but pretty soon you have a space heater - 60W of waste heat is a LOT.  Parallel pass transistors with the bases driven by another power transistor is fairly common.  It just needs to be considered.

[Efe_114]

After changing the lm317 it wasnt going under 1.25 when i plugged the previous one back it was stabilized at the input voltage and didnt have a voltage drop i was feeding 12v and it gave 12 volt out i think something funny is going inside the chip also i ordered them from the most unreliable seller in my country

[james_s]

It has obviously been fried, the internal pass transistor is shorted. Also if you bought them from a random China source they are probably inferior counterfeit parts.

[fsr]

You can make an LM317 go to 0v even without a negative transformer. The trick is to add two diodes between GND and the (-) output of the PSU (so, your outpur is aprox 1.4v above the real GND). The diodes must be able to handle the max output current and the power dissipation. The resistive divider that controls the LM317 output voltage is also referenced to this point.
You will need to consider the power dissipation of the LM317 in the worst-case scenario and use a suitable heatsink. The datasheet has a very good explanation on how to calculate this under the "Heatsink Requirements" section.

EDIT: i'm sorry, the circuit i posted before won't work, it was just a part of a circuit i did which had a current limit that could go all the way down to 0v, but for the CV control to go down to 0v, you need some way of getting 1.25v below your main diode bridge's negative output. I should have checked the circuit carefully, instead of sketching it and posting.

Anyways, if you use the LM317, remember to check the "Heatsink Requirements" section, and forget about using a 48v transformer. The power dissipation for the chip is about (input_voltage - output_voltage) * current. Try to use a transformer/bridge/capacitor combination that will result on a voltage that just goes over the requirements for the LM317 for the output voltage you need (check the dropout voltage), and whatever additional drop that the final circuit could introduce.

[VEGETA]

You can make an LM317 go to 0v even without a negative transformer. The trick is to add two diodes between GND and the (-) output of the PSU (so, your outpur is aprox 1.4v above the real GND). The diodes must be able to handle the max output current and the power dissipation. The resistive divider that controls the LM317 output voltage is also referenced to this point.
You will need to consider the power dissipation of the LM317 in the worst-case scenario and use a suitable heatsink. The datasheet has a very good explanation on how to calculate this under the "Heatsink Requirements" section.

that act itself creates a negative voltage. Remember that voltage is relative, so it is negative relative to ground which is 0. There is no way lm317 could output 0v without negative voltage.

In your case, you did it by splitting the rail using diodes, and in my case by creating another rail using 7660. Same result.

[fsr]

You can make an LM317 go to 0v even without a negative transformer. The trick is to add two diodes between GND and the (-) output of the PSU (so, your outpur is aprox 1.4v above the real GND). The diodes must be able to handle the max output current and the power dissipation. The resistive divider that controls the LM317 output voltage is also referenced to this point.
You will need to consider the power dissipation of the LM317 in the worst-case scenario and use a suitable heatsink. The datasheet has a very good explanation on how to calculate this under the "Heatsink Requirements" section.

that act itself creates a negative voltage. Remember that voltage is relative, so it is negative relative to ground which is 0. There is no way lm317 could output 0v without negative voltage.

In your case, you did it by splitting the rail using diodes, and in my case by creating another rail using 7660. Same result.

I mean that he doesn't need a center-tapped transformer, or something like that for this to work, just a voltage drop greater or equal than 1.2v under the negative output terminal, reference the voltage divider to that point and that's it. Bob's your uncle

Something like this (see attachment):
EDIT: Sorry, wrong circuit. Check my previous post for details.

[spec]

I want to make a small psu that could go up to 15 volts and DOWN TO 0 VOLTS (because i like hard projects) .no current limit needed. i want to make it out off jellybean parts and not copy it from a random website. And most importantly im a beginner(i have been in hobby for 3 years but since 90% of my projects fail i call myself beginner)Can anyone help me about this? Also this is my first post in this forum so forgive my mistakes please

Hi Efe,

The attached schematic may be what you are looking for. It adjusts down to zero volts using the principle (already discussed) of generating a -1.25V line to offset the 1.25V reference voltage of the LM317, but without an additional winding on the transformer or voltage dropping diodes.

If you are interested in this approach I can post a practical schematic showing decoupling capacitors and describe the best layout.

(4m7 on the schematic is 4.7 milli Farads or 4700 micro Farads)

[not1xor1]

You can make an LM317 go to 0v even without a negative transformer. The trick is to add two diodes between GND and the (-) output of the PSU (so, your outpur is aprox 1.4v above the real GND). The diodes must be able to handle the max output current and the power dissipation. The resistive divider that controls the LM317 output voltage is also referenced to this point.
You will need to consider the power dissipation of the LM317 in the worst-case scenario and use a suitable heatsink. The datasheet has a very good explanation on how to calculate this under the "Heatsink Requirements" section.

That is not a trick but a really dumb solution as, besides the additional 1-2W of wasted power,  the output voltage would drift according to the temperature of the diodes and the load current.

As I showed in the previous schematic, with just a couple of diodes and capacitors you can get an additional negative rail from the transformer (charge pump) and with a cheap voltage reference and a resistor you can get -1.25V as stable as the inner reference of the LM317.

And if the unregulated supply is coming from an AC-DC converter (switching PSU) you may use a 555 or even a cheap CMOS logic IC (which are usually cheaper than a dedicated IC) to get a negative rail via a similar charge pump circuit.

[not1xor1]

I mean that he doesn't need a center-tapped transformer, or something like that for this to work, just a voltage drop greater or equal than 1.2v under the negative output terminal, reference the voltage divider to that point and that's it. Bob's your uncle

Something like this (see attachment):

That would not work.
You have to connect R1 to Adj and the potentiomenter to the negative of the rect. bridge.
It is anyway a bad solution.

[not1xor1]

Hi Efe,

The attached schematic may be what you are looking for. It adjusts down to zero volts using the principle (already discussed) of generating a -1.25V line to offset the 1.25V reference voltage of the LM317, but without an additional winding on the transformer or voltage dropping diodes.

If you are interested in this approach I can post a practical schematic showing decoupling capacitors and describe the best layout.

(4m7 on the schematic is 4.7 milli Farads or 4700 micro Farads)

You can even use a cheaper 1.25V shunt regulator (LMV431, LM385, etc.). Of course you must properly bias it so that the total current of the negative rail, at the worst condition (lowest transformer voltage) should be greater than the minimum shunt regulator voltage + the quiescent current of the LM317 (aproximately 12.5mA in the schematic you attached).

[spec]

Hi Efe,

The attached schematic may be what you are looking for. It adjusts down to zero volts using the principle (already discussed) of generating a -1.25V line to offset the 1.25V reference voltage of the LM317, but without an additional winding on the transformer or voltage dropping diodes.

If you are interested in this approach I can post a practical schematic showing decoupling capacitors and describe the best layout.

(4m7 on the schematic is 4.7 milli Farads or 4700 micro Farads)

You can even use a cheaper 1.25V shunt regulator (LMV431, LM385, etc.). Of course you must properly bias it so that the total current of the negative rail, at the worst condition (lowest transformer voltage) should be greater than the minimum shunt regulator voltage + the quiescent current of the LM317 (aproximately 12.5mA in the schematic you attached).

I know that, but you can't get much cheaper and ubiquitous, than an LM337. Besides, it looks nice and symmetrical with two three terminal regulators.

[not1xor1]

Hi Efe,

The attached schematic may be what you are looking for. It adjusts down to zero volts using the principle (already discussed) of generating a -1.25V line to offset the 1.25V reference voltage of the LM317, but without an additional winding on the transformer or voltage dropping diodes.

If you are interested in this approach I can post a practical schematic showing decoupling capacitors and describe the best layout.

(4m7 on the schematic is 4.7 milli Farads or 4700 micro Farads)

You can even use a cheaper 1.25V shunt regulator (LMV431, LM385, etc.). Of course you must properly bias it so that the total current of the negative rail, at the worst condition (lowest transformer voltage) should be greater than the minimum shunt regulator voltage + the quiescent current of the LM317 (aproximately 12.5mA in the schematic you attached).

I know that, but you cant get much cheaper than an LM337 and besides it looks nice and symmetrical with two three terminal regulators.

You're right. I just checked. The price difference is less than what I remembered... just 20-30 cents and only if one carefully selects the cheaper shunt regulator.

[VEGETA]

I think you need to see this: https://www.aliexpress.com/item/50PCS-TL431-TO92-TL431A-TO-92-431-new-voltage-regulator-IC-free-shipping/32694971062.html

I recommend getting it since you can use it in lots of other purposes and it is dirt cheap.

Here is ICL7660 if you wanted it too, which I recommend having for lots of different purposes:

https://www.aliexpress.com/item/5pcs-ICL7660SCPA-DIP8-ICL7660-DIP-ICL7660S-7660SCPA-new-and-original-free-shipping/32493575139.html

Both of them won't cost you 5$.

[Efe_114]

I found out something funny was going on in the lm317 later it just shooted up to 12v and i wasnt able to change the output considering it is from the worst supplier locally i think it was broken

[VEGETA]

I found out something funny was going on in the lm317 later it just shooted up to 12v and i wasnt able to change the output considering it is from the worst supplier locally i think it was broken

I guess this is similar to what happened with me, sort of.

Probably this is due to limited total output power of the source, it could be say 10W total... So you could have 10V 1A, but when you draw 2A it will drop down to 5V regardless of how high you dial it. This is a well-known phenomena in all industry.

This occurs even on industrial supplies. One time we had a diesel generator which worked perfectly but then we connected a huge pump or motor which made the generator's voltage to drop down significantly.

Solution: use a better power sources, test it with DC-DC switching modules to verify its capability, then use it with lm317.

[David Hess]

the original poster stated: no current limit needed
I anyway added a foldback current limit as I think it is the bare minimum for safe operation.

I took that to mean no adjustable current limit needed.  I would still include short circuit protection.

I would use a single supply operational amplifier to drive an output transistor or likely Darlington with a Vbe current limiter on the output.  This configuration is the same as that of a 723 but the NPN differential amplifier of the 723 cannot operate down to zero volts.

SiliconWizard's design is close to what I would do but I would add a reference and Vbe current limiter. (1)

Not1xor1's design requires a negative supply.  If you are going to do that, then just use a 317 with the modification to operate down to zero volts.

(1) Who am I kidding?  I would drive the output transistor using the positive supply lead of the operational amplifier and have the current limiting go into the *output* of the operational amplifier.

[spec]

Of course, if you want the ultimate in simplicity for a 0V to nV output PSU, go for the LT3080 family of current reference regulators (take care with the layout and capacitors to ensure frequency stability though).

https://www.analog.com/media/en/technical-documentation/product-selector-card/LDO.pdf

https://www.analog.com/media/en/technical-documentation/application-notes/an142f.pdf

__________________________________________________________________________________________________________

https://www.analog.com/media/en/technical-documentation/data-sheets/3082f.pdf

https://www.analog.com/media/en/technical-documentation/data-sheets/3085fb.pdf

https://www.analog.com/media/en/technical-documentation/data-sheets/3080fc.pdf

https://www.analog.com/media/en/technical-documentation/data-sheets/3081fc.pdf

https://www.analog.com/media/en/technical-documentation/data-sheets/3083fa.pdf

[JoeO]

I decided to use LT1080 before but it isnt available in my country.

Why don't you turn on your country flag so that the people helping you will know what parts are available in your country.

[spec]

The day Al Gore was born there were 7,000 polar bears on Earth.
Today, only 26,000 remain.

[not1xor1]

The day Al Gore was born there were 7,000 polar bears on Earth.
Today, only 26,000 remain.

Unfortunately polar bears, whose only guilt is being much more clever and handsome than any politician  (although seals might disagree), live most of their lives and almost exclusively feed on fast disappearing sea ice (Ursus maritimus does mean sea bear) and so they are listed as vulnerable. 

I would rather laugh when there will no longer be a guy needing a blonde dog to hide his empty head. 

Probably I won't have to wait much as the U.S., although being on top for health expenditures, are one of the worst places (among developed countries) regarding life expectancy and one of the few where it has been decreasing for several consecutive years.

And subsidies to coal and oil and most of other absurd overturns of previous EPA regulations are quite unlikely to help.

[nemail2]

all you need is a DAC, VREF, uController, opamp, a series pass transistor and some resistors and caps. take a look at daves uSupply and then at my PSU. it is not as complicated as it looks like.
https://github.com/mamama1/LabPSU_Darlington/blob/master/Hardware/schematics.pdf?raw=true

[VEGETA]

all you need is a DAC, VREF, uController, opamp, a series pass transistor and some resistors and caps. take a look at daves uSupply and then at my PSU. it is not as complicated as it looks like.
https://github.com/mamama1/LabPSU_Darlington/blob/master/Hardware/schematics.pdf?raw=true

But don't you need stabilization? are you sure your circuit won't oscillate?