DIY power supply

[MarkF]

The LM338 would work as well. 
The thread I pointed to has a few extra parts than normally shown in the datasheets. 



Here is another thread discussing bypass transistors:
   https://www.eevblog.com/forum/beginners/resistors-purpose/

That eBay 10-turn pot looks like the same Bourns one with different type pins.  I rarely buy from eBay and put together larger orders with Mouser. 
You will have to go over the design for your voltage range and pick the pot value you will need.

[Pirateguy]

do i need those transistors? if i am reading the datasheet correctly the 338 can do 5 amps, and that's about as much as i can imagine myself needing.
i have a ZVS driver kit and a stungun module that ask for 12v, and those are the only times i have needed to account for amps beside limiting resistors for leds.

if yes, anything with the right power capabilities will do?

[MarkF]

It's not the current that is the limiting factor.  It's the heat the LM338 will have to dissipate. 30W or more is going to generate a lot of heat.  You will need a big heatsink and a fan.  The worse case will be the voltage drop across the LM338 and the current drawn by the load.  You can calculate the power the LM338 will need to handle by the worse case you expect to use.  You could try with just a LM338 and see if it can handle power levels you want to use.

Even with a PNP bypass transistor, you will be transferring (I should say sharing) the power dissipation from the LM338 to the transistor.  The TIP36C transistor in the other thread comes in a TO-247 package which is almost twice the size of a TO-220.  Much better to dissipate the heat.

Most power supplies will have relays to switch to different input voltages so the regulators don't have to dissipate all the heat.

[Pirateguy]

i was thinking of attaching them to the top of the big heatsink that's in there, which is right in line with the fan.

what about the MJ2955?
it is the compliment to the 2n3055 i have and comes in a to-3 package.
also i have some ceramic plates to go in between the heatsink and transistor.

[MarkF]

Let's not jump off the cliff just yet.
Let's take a look at what you are trying to do here and what your options are.

  1)  Using the 32V 1.5A HP Supply, you could design something in the 0-30V @ 250mA to 500mA range without a bypass transistor.  It would be a simple construction and in most cases it would be all you would need.  For example, I have a 60V @ 250mA maximum supply.

  2)  Using the 32V HP Supply, you could add a bypass transistor to reach the 0-30V @ 1.5A range.  Maybe limit to 1A to make it easier.  A more involved build.

  3)  Using the 12V from the ATX Supply, you could get  in the 0-10V @ 5A range.  I think it would be rare that you would use these current levels.  In the two cases you mentioned, you would used your 12V high current output directly from the ATX supply.

To answer your question:  You could build with a MJ2955.  You could probably make almost any power PNP transistor work.  You will need to adjust the resistor values to get the desired bias voltage.

As for a variable power supply:  Is this something you want to build or buy?  How much effort do you want to do in the design and build?  Personally, this is one item I would buy.  Possibly with a LAN interface to automate some project testing.

In reference to using the ATX heatsink.  It is NOT something I would recommend.  You have already compromised the cooling and airflow by removing the case.  The saving grace is that you are NOT using anything near the capabilities of the supply.  It is still to be seen if it gets hot during use.

[Pirateguy]

i want to build my own for now, and either improve it over time or replace it with a commercial unit
if/when my projects and needs get more sophisticated. though i have a bunch of things still higher up on my wishlist.
including a DSO (currently have only a 1st gen dso nano) and moving to another house.

as for the ATX case, my intention was to put it back in the original case and build all the externals into the side.
all that will change in the end is there will be fewer wires, some room made by removing bulky connectors and soldering
wires to the board, and the back side of all the binding posts and display etc.
which, on that side should not effect airflow too much.

if i add another heat sink though that will change i think.
so right now i think maybe going with the HP one for the variable might be better.
though i don't have a case for it that can fit the additional stuff.

so for now i think i will build the ATX with the fixed outputs and a temperature display.
ill use the polyswitches for protection, and add LEDs on each output to let me know if i popped one.

and i will set aside the HP one for the variable unit, and treat that as a separate project.
prolly end up going for suggestion 1) with an option to upgrade to suggestion 2) later.

so for the ATX fixed output unit, for the load resistor on the 5v, could i use one of the ones from this thing?:

[MarkF]

Try it first without a load resistor. Some supplies don't need it. Depends on when it was made.

[coldfiremc]

Sorry but I can't resist

[MarkF]

Fantastic 
But the sword is jealous

[MarkF]

I do like these little HP Supplies.  Not a lot of power but without a fan, they are quiet and don't take up much space.
   HP 6216A Power Supply

Ditto on the little HP's. The 6216A is good for about a half amp at 24 volts. Has both voltage and current modes, each having coarse and fine adjustments. Very well documented which is easily available online. Operation, service & schematics. With a little patience they can be had in nice shape for $50 shipped on ebay. I bought three last winter like that for my bench in a spare bedroom. Quiet and compact, built like a tank and easy to repair.

Great minds think alike Old Printer.
I also bought three of the HP-6216A power supplies for my bench in a spare bedroom. In addition, I bought a HP-6217A which is 50V@200mA.
I also built a +/-15V@1A using the PCB I posted above for an analog project needing clean power rails.

[Pirateguy]

oh for crying out loud!

lol ok i get it, can't beat em...

so wat do you guy think of these modules?:

[plazma]

oh for crying out loud!

lol ok i get it, can't beat em...

so wat do you guy think of these modules?:

I already suggested them. I favor the DPS5005 module because it is the most powerfull of the compact ones (pcb in display module).

[MarkF]

I watched Dave's video a while back.  But that is all I know about them.
They look promising if you want to buy something.  One would probably be a better choice to pair up with your HP power supply or the ATX depending on the voltage range you're interesting in.

Edit:
After watching the videos again, my only concern would be that the output is noisy.  It may or may not be an issue depending on what you're using it for at the time.

[Pirateguy]

i don't mind some noise, it will just force me to learn about putting filters in my projects n stuff when it comes up.
(does that make sense?)
i like the panel type setup, leaves a lot of freedom over the placement of the binding posts n stuff.
ill keep these in mind for later. for now ima diy it. i feel like this is all stuff i'm gonna need to know about
anyway to progress in my hobby. so itll be worth it just for the experience.

speaking of dave's videos, i found the perfect one for me and this project:

i haven't watched the whole thing yet, but it looks like this will teach me the math i need
for the adjustable one.

anyhoo, wat do you guys mean when u talk about 'double bananna plugs'?
and does the 3/4" measurement have to be exact? or is it a minimum?

[Old Printer]

About a year ago when I was needing my first decent power supply I ran across this series of videos where Dave designs a PS. I had a look inside some good quality supplies and was amazed at the complexity of the circuit and components. I realized I was not going to be able to come up with anything like that on my own and started watching Dave's. I don't recall which video number it was, but 3 or 4 into the series he started running into complications. As I recall he kind of lost interest in finishing, or at least overcoming all of the problems to his satisfaction. This is very loose recollection on my part so don't quote me, but it was at that point I set my sights on a good commercial PS. The nice HP's were too expensive so I headed to the used market. Seven PS's later I think I have enough for now They can be an addiction.

[MarkF]

anyhoo, wat do you guys mean when u talk about 'double bananna plugs'?
and does the 3/4" measurement have to be exact? or is it a minimum?

There are a lot of varieties.  The spacing is 3/4".  Just Google them.  Here are a few:

[MarkF]

speaking of dave's videos, i found the perfect one for me and this project:
      _{Video removed}
i haven't watched the whole thing yet, but it looks like this will teach me the math i need
for the adjustable one.

Yes.  Dave's video would be a perfect learning experience.  Follow it with two LM338's and get it working on a breadboard.
Then layout a PCB and get it made.  You can expand on the layout I showed above by adding the current limit stage.
The 4 diodes form a full-wave bridge for an AC input.  Or, you can bypass them and feed in DC via a wall-wart plug.
You can get five 100mm x 100mm PCBs made for $5.00 US at Elecrow.

     

[Pirateguy]

lol but postage would be $8
for now i think ill just use some perfboard.

so i had an idea:
using this calculator: http://diyaudioprojects.com/Technical/Voltage-Regulator/
i determined that i need 5.9k to get exactly 32v. so i was thinking
what if i just put a 10k multi turn pot parallel with a small 20k pot on the pcb, and use the 20k to tune the max resistance
till it gives me 30V?

[MarkF]

If you do two LM338 in series (1st doing current limit, 2nd setting the voltage), you are probably only going to get 28V for the output. That's assuming 2V drop-out volts for each LM338.  Designing for 25V would be more realistic.

Using that calculator, you have R1=240 and R2=5K.  I would only use a 5K ohm 10-turn pot mentioned earlier for the voltage control.

I built an Electronic Load with a 10K 10-turn pot and it gave me plenty of sensitivity to set the current.  One 10-turn pot for each LM338 should be enough for the current and voltage adjustments.  You would probably only use two pots if they were the 3/4 turn type. If you want two pots, they would be in series with one big (say 5K) and one small (say 500).  Parallel pots would be a NO NO.

You will need to determine the value for the current limit pot.

Dave's video mentioned that the LM338 needs a minimum load and to adjust R1 and R2 to provide that load.  I haven't worked through the math to determine the best values to use.


I know the postage would cost more than PCBs from Elecrow if you only get one design made.  I try to order more than one PCB design to make the postage worth while.  But, I suggested it so you have the experience of going through the whole process and learn how to do a PCB layout and get it built.  Maybe even a second board with all your polyfuses and whatever meter switching you want to do.

[Pirateguy]

yeah i'm not quite at the point where i want to order commercial boards like that yet (though i am excited about these options we have now).
i want to be better at designing circuits first so i can have some confidence of not having to order the same board 5 times due to tiny mistakes.
and also i want to transition from THT to SMD first, at least to the point of having confidence in being able to solder a board full of 0805.

y no resisters in parallel?
with what i had in mind the 2nd one is just for fine tuning the range and then maybe fixing it in place.
the idea was that i would not be bound by standard values for resistors...

anyway i totally spaced on the 2nd 338 for the current
back to the drawing board...

still, you sound like no pots parallel is a bad idea in general?
if so, how so?

also the minimum load, isn't that what the 240ohm resistor is for?
my circuit would have been the diode protected version that is in the datasheet, only adding that sencond pot...

[rstofer]

Pots in parallel:  Spend some quality time with Excel.  Compute, say, every 10% of the first pot in parallel with every 10% of the second pot.  Rerun for different pot values...  Remember:  The parallel combination of resistors is always less than the smallest value.

'Fine' and 'Coarse' adjustment pots are often in series.  It might be marginally faster to adjust two single-turn pots than a single 10 turn pot but just how often does the voltage need to change?  Three supplies may be set for +-15V and 5v and never changed.  Ever!

I suppose an adjustable supply could be used as a signal source and speed dialing might be handy but I would still prefer the 10 turn pot.

[Pirateguy]

what i mean is that i only mount 1 pot on the panel and one inside the case that cannot be accessed.

i would turn the panel pot up all the way and then dial the pcb pot in to the max voltage i want.
after that i would never touch the pcb pot again and only use the panel pot to change the voltage.

[ricktendo]

I already suggested them. I favor the DPS5005 module because it is the most powerfull of the compact ones (pcb in display module).

And also there is a open source version firmware called OpenDPS for the DPS5005 (also controllable via WiFi (ESP8266) or via a serial port)

https://johan.kanflo.com/upgrading-your-dps5005/

https://johan.kanflo.com/hacking-the-dps5005/

And it seems like the OpenDPS firmware is also compatible with my DPS5015

https://gojimmypi.blogspot.com/2017/04/opendps-with-dps5015.html

Edit: I wish somebody could post some instructional videos on the whole compiling and flashing process!

[MarkF]

I just ran across this video that is of interest in doing the current limit.

   


Which would look something like:
R_SC=10,  R2=110,  R3=1K  ==>  125mA<=I_OUT<=1.26A
Remembering that our minimum voltage is going to be 1.25V because we don't have a negative voltage source.
I'm afraid that the voltage drop across R_SC will not make this a viable solution.

   

   

[MarkF]

I forgot all about this:
  You might consider this power supply by Peter Oakes: