Updated schematic for a popular DIY linear bench power supply

[Aramp]

The other design flaw of the "green" power supply, as was mentioned by  Kleinstein above, is that the amperemeter is placed after the voltage probe connection point. Thus, the drop in it will not be compensated. This means that the effective output resistance is increasing, suggesting voltage drift as the load grows. The original schematic had the amperemeter in series with the shunts to eliminate that problem; however, the standard cheap digital ammeters will not work that way as they are expected to be connected in series with the negative line. On the other hand, the shunts on these ammeters have resistance in the range of 0.01 ohms, which is lower than the resistance of most wires one may use to connect to the power supply.

If the output resistance still matters, the best way would be to have a custom-designed indicator that can be placed in the positive line or use an old-school mechanical meter measuring the voltage across R11.

[enut11]

Hi Aramp
The original circuit is confusing as both voltage and current pot connections are shown as 2-pin instead of 3.

My suggestion is to correct it on your updated circuit as there are no instructions with the kit on how to wire the pots to the plugs anyway. Better to present it the right way.

As I was only using a 2-tap, ie 3 wire transformer, I had to disable the second relay by cutting a track going to the coil otherwise the circuit switches to a non-existing winding and shuts down the power supply momentarily only to repeat the whole process.

I have just finished load testing my PS and all is well.

This PS is missing a Constant Voltage (CV) indicator. I believe this may be as simple as inserting an LED in the U1B pin 7 line either in series with or instead of D17.
enut11

[enut11]

The other design flaw of the "green" power supply, as was mentioned by  Kleinstein above, is that the amperemeter is placed after the voltage probe connection point. Thus, the drop in it will not be compensated. This means that the effective output resistance is increasing, suggesting voltage drift as the load grows. The original schematic had the amperemeter in series with the shunts to eliminate that problem; however, the standard cheap digital ammeters will not work that way as they are expected to be connected in series with the negative line. On the other hand, the shunts on these ammeters have resistance in the range of 0.01 ohms, which is lower than the resistance of most wires one may use to connect to the power supply.

If the output resistance still matters, the best way would be to have a custom-designed indicator that can be placed in the positive line or use an old-school mechanical meter measuring the voltage across R11.

Not a problem for me as I have decided to use analog meters anyway. The cheap digital meters inject noise into the PS output.

 R11 won't work for this mod as there is almost no current flowing through it. The ammeter shunt would have to be after the emitter resistors (R24 and R25).
In my case it will be simple as I need 2A output so only one output transistor (Q4). The analog ammeter will be wired across it's emitter resistor( R25).
enut11

[magic]

Isn't it this kit?
https://www.eevblog.com/forum/beginners/shitty-chinky-psu-kit/

The thread is mostly a few pages of spoonfeeding an utterly incompetent OP, but I recall that there was some important discussion there, including fixing potential problems with the design.

[enut11]

Thanks @magic. I just did not think to search for"shitty chinky psu kit"!

I have not had any serious problems yet but will go through that thread.
enut11

[enut11]

The Constant Voltage (CV) LED mod was easier than expected. The indicator was wired in series with D17 to pin 7 of U1B.

Now new indicator is lit in CV mode but is extinguish if CC mode is invoked.

EDIT: I no longer recommend this mod due to the possibility of LED failure. See Reply #38 for a better solution.

[Aramp]

R11 won't work for this mod as there is almost no current flowing through it. The ammeter shunt would have to be after the emitter resistors (R24 and R25).
In my case it will be simple as I need 2A output so only one output transistor (Q4). The analog ammeter will be wired across its emitter resistor ( R25).

Of course, you are right, as the opamp has nanoamp current at its inputs. If two transistors are used, the good compromise would probably be to put the ammeter, as in the original version of the circuit, after the junction of two shunt resistors, and connect the voltage regulation circuit after that point. I have used a digital voltmeter for this one. I measured the level of digital noise with an oscilloscope. It's clearly there, but it seems to be below 100 uV. However, am I planning analog indicators for the other one, as I also need a silent PS for radio equipment that has no noise.

[Aramp]

enut11,

One question about your mod: is it safe from wiper disconnecting? Will the 1M resistance in parallel suffice to protect from the opamp misbehaving?  Did you test that?

[magic]

Wiper disconnection is one of the issues I remember ranting about in that old thread.

I think the other one was what happens when voltage on input pins of the opamps goes below their negative supply under fault conditions. LM324/358 may get phase reversal if this is allowed to happen.

[enut11]

Hi @Aramp. No I did not. Also found another thread from @magic that highlights the danger of LED failure.

A safer option suggested by @xavier60 in that thread:
"D17 could be replaced with an LED also but LEDs can and do go open circuit.
A safer option is an LED with series dropper resistor from the +15V control rail to the output of the CV opamp, pin 7"

In our case that would be the +12v rail. This is the mod that I will be going with.
enut11

[enut11]

Wiper disconnection is one of the issues I remember ranting about in that old thread.

I think the other one was what happens when voltage on input pins of the opamps goes below their negative supply under fault conditions. LM324/358 may get phase reversal if this is allowed to happen.

@magic, would you tell me more on this please.

[magic]

1. The wiper of a pot may lose contact with the track. Whatever PSU you are building, it's a good idea to ensure that this doesn't result in the output jumping to maximum. LM358 has input bias current flowing out of the pins, so never leave IN+ floating under any circumstances. R13 on your schematic should prevent problems in such case, make sure it is installed. The original diagram in your post above (unlike the red lines drawn over it) is safe even without R13.

2. Under some circumstances, if the voltage on either input pin of LM358 goes ~0.5V below the negative rail, the output may (or may not) jump to the positive rail. The details are a bit murky, but several manufacturers caution against such operation. Make sure negative voltage is never applied there. Here I think this could happen if output voltage of the PSU is still hight due to charge stored in output caps but the CV pot is quickly turned down to zero.

[enut11]

@magic
I was about to change the 7912 for a 7905 on the control circuit. This would not be a good idea given your explanation above?
enut11

EDIT: I did not go ahead with this mod as it is unnecessary.

[enut11]

I have completed the 'Green PCB' kit often labelled as 35v/5A or 30v/5A purchased from Aliexpress.
I used a digital panel meter, despite my previous comments but added a disable switch if needed.
The plastic Jiffy control box (200x60x110mm) contains everything except the main power transformer which sits in its own box connected by 3 wires the the control box.

My need was for only 1A so I eliminated one output transistor and one 0.22R emitter resistor.
The transformer I used had 17v and 11v windings so I also eliminated one of the switching relays.

I modified the circuit in 2 other ways, one I consider very important:
1) I added a constant voltage LED indicator as per @xavier60 - nice to have. I no longer recommend the mod in Reply #30.
2) The voltage pot was re-wired by swapping the hot and wiper leads. This  significantly improves stability at lower voltage outputs.
The voltage pot wiper now feeds into the opamp ( U1B pin 5) as per normal and no longer adversely affects the Vref line from TL431 at low output voltages.

enut11

[enut11]

Noise measurements at 10v were made with 70mA and 500mA loads. They were 40uV pp and 70uV pp respectively - good figures given the low cost of this kit (<$30AUD).

The 500mA recording is shown below. The Y scale spans 1mV.

Thank you @Aramp for your up-to-date schematic. Much appreciated.
enut11

[Aramp]

@enut11,

Great job! The device looks and performs very well. I'll make both mods and add a switch for the display. They are indeed very useful.

From my experience, the use of the PS is satisfactory. It sits on the shelf on my bench and just works. Mine uses a heavy transformer and produces all the 5A without problems, and I have never seen the fan turning. I have forced it by heating the thermistor during the test. I have installed a multi-turn pot for voltage regulation to make it less touchy and more precise and to avoid very fast swings that may potentially lead to the device misbehaving.

[enut11]

@Aramp
Agreed, 10T pots are worth installing. When I was fiddling with my kit I tried an open-circuit pot wiper to U1B and the supply shut down. Good to know that the 1M resistor from pin 5 to ground works.

I have ordered a fan as I am using a small heatsink.
enut11

[Aramp]

BTW, I have just modeled in LTspice the transformer taps switching circuit to help analyze its workings and found out that it is very cleverly designed. If you look at the schematic, there is a D7-R26 in series going from the collector of the transistor T1 (and R27 and D8 from T2) that introduces hysteresis. The values of R26 and R27 are different to compensate for the higher threshold voltage for Stage 2, so the hysteresis value remains roughly the same for both stages. The idea is that whenever the voltage at the non-inverting input of the opamp becomes larger than the threshold defined by the pot RV3, the opamp opens the transistor, and the collector is grounded. This immediately drops the voltage on the inverting input even further to secure the switching and avoid oscillations. Then, when the output voltage is regulated downwards, it should achieve a lower value for the opamp to switch back, and then the same effect happens in the opposite direction. C17 is there because whenever the switching happens with high load connected, the current going through R7-R8 may fluctuate, so it makes sure the voltage at non-inverting inputs is not jumping around. The Zener diode D11 is needed to provide a reasonable voltage on the base of the transistor.

[xavier60]

@Aramp
Agreed, 10T pots are worth installing. When I was fiddling with my kit I tried an open-circuit pot wiper to U1B and the supply shut down. Good to know that the 1M resistor from pin 5 to ground works.

I have ordered a fan as I am using a small heatsink.
enut11

Because of the way wire-wound multi-turn Pots can fail, the PSU's output can go high, regardless of how it's wired. This is due to the wire breaking rather than loss of wiper contact.
I have found the  cheap 3590S style Pots on ebay and Aliexpress to be very unreliable.
In 3 years, I have not had any WXD3-13-2W style Pots fail. But they are huge.
There is a shorter 5 turn variant.

[dietert1]

It's fairly easy to implement a configurable voltage safety limit.
Just get a rotary switch with let's say 6 positions and wire 5 different zeners to it. Connect that between ground and the control signal at the base of Q2. A 6.8 V zener will result in about 5.5 V maximum output voltage. Each safety limit will be about 1.3 V less than the nominal zener voltage. One could implement 5 V, 9 V, 12 V, 15 V, 18 V limits and the 6th switch position remains for "no safety limit".

Regards, Dieter

[xavier60]

With my build of the design from the other large thread, I implemented Pot failure protection by monitoring for a rapid increase in wiper voltage exceeding 300mV. This idea can only be used with designs where the Pot voltage divides the reference voltage.

[enut11]

@Aramp
Agreed, 10T pots are worth installing. When I was fiddling with my kit I tried an open-circuit pot wiper to U1B and the supply shut down. Good to know that the 1M resistor from pin 5 to ground works.

I have ordered a fan as I am using a small heatsink.
enut11

Because of the way wire-wound multi-turn Pots can fail, the PSU's output can go high, regardless of how it's wired. This is due to the wire breaking rather than loss of wiper contact.
I have found the  cheap 3590S style Pots on ebay and Aliexpress to be very unreliable.
In 3 years, I have not had any WXD3-13-2W style Pots fail. But they are huge.
There is a shorter 5 turn variant.

@xavier60
I accept that multi-turn pots can be unreliable and it happened in one of my previous projects. However, in this case, I deliberately disconnected the wiper going to opamp pin 5 on my PS and the output went to zero. This can only be due to the grounded 1Mohm resistor and appears to be fail-safe.
enut11