DIY project for constant voltage 5v, 2amp

[bitman]

I've watched eevblog on you tube for quite a while, and one of the series I've watched more than once is when Dave creates a bench power supply from scratch.  Very informative and an inspiration to repeat. HOWEVER I'm trying to find something "small" I can build to generate 2amps 5volts to power some TTL. According to the bench supply it currently draws between 1.15 and 1.35amps depending on "state" so I think 2amps would be a safer thing than 1.5.  That leaves out the basic 7805 circuits - or rather I'm experimenting with two 7805 in series.  The problem is I need to calculate cooling requirements and a few other ares I don't feel comfortable in right now.

So I'm looking to see if there's a DIY project spec out there creating a basic 5v PSU (from 120v) which protects against shorts and reverse polarity. It doesn't have to be a Dave's level of teaching but something that explains the circuit would be great since I'm on a learning path here.  It should include a gerber or similar file so I can get the circuit created permanently.

I've created several LM317 and LM7805 circuits from different templates I've found - but they all seem to crack around .5 -> 0.7 amps. I then found an old CPU heat-sink drilled a few holes in it and used it with the LM7805 - that got me closer to 1.5AMP but not all the way to 2AMP. It seems odd to have to keep adding 7805s - in particular because the voltage seems to drop to around 4.8 volts under load.

I know the problem - it's me   I don't know which components are best for these kind of efforts. Initially I didn't think I needed a simple arduino or other micro-processor but the more I see small projects using a simple LCD the more it appears to me that it's the right way to go. Anyway - I can find a lot of projects out there but well, they show their age. Some are even malware sites today - most have pictures of circuits not the actual data I can use with software like KiCad. I am NOT planning to etch out a PCB at home - those chemicals don't belong in my closed work-space, so being able to get it printed somewhere is important.

Anyway - the short version: Looking for a project site to create simple PSUs 5v and at least 2amp output (from 120v).

[Benta]

Quite frankly - for 5 V, 2 A from 120 VAC I'd simply buy a USB wall-wart. You won't get anything easier or cheaper. 10...12 W types are plentiful.

[bitman]

Quite frankly - for 5 V, 2 A from 120 VAC I'd simply buy a USB wall-wart. You won't get anything easier or cheaper. 10...12 W types are plentiful.

Ahh yes. But you see, part of this is to learn. Ie. figuring out how much thermal cooling I need etc.  I'm not doing it because I cannot find a PSU that just plugs in (if fact I have plenty lying around).

[Ian.M]

Simple linear regulators above about 1.5A are a P.I.T.A.  and best avoided unless you cant tolerate the noise of a switcher, or *really* need the learning experience!     

Perhaps the easiest option is simply put a separate 7805 on each board, or if its all one big board break it up into sections, such that each 7805 load current is under 500mA, and distribute a higher voltage between 8V and 9V to all the regulators.  That splits up the dissipation into reasonable 2W 'chunks' which can be handled by a small heatsink on each regulator.

However, if you must go above 1.5A:

L78S05 is a 7805 that's rated for 2A load current.  Its input voltage and heatsinking are *CRITICAL* - like any other linear series pass regulator, the heat dissipated is equal to the product of the load current and the voltage differential, and if you don't get it out of the die quickly enough it goes into thermal limiting and the output collapses.  The dropout voltage *should* be under 3V differential, but it rises at low temperatures, so figure on 8V in at the troughs of the ripple at 2A load current.  If you give it a smooth 8V in, you've got 6W to get rid of.  That's a 30 deg C junction temperature rise above the heatsink temperature.  However 4V pk-pk ripple (not unusual for an unregulated rectified supply) will give you an average of 10V in and you've now got 10W to get rid of and a 50 deg C junction temperature rise.   That's on-top of whatever temperature rise the heat-sink is running at, and ideally you want to run the die at no more than 100 deg C with 25 deg C ambient to give you enough margin to cope with elevated ambient temperatures and still get a long life out of the regulator.   

If the transformer output voltage is a bit too high, a power resistor in the feed to the L78S05, calculated to drop enough voltage to just get the troughs down to 8.5V at 2A load current will help considerably.  The imput decoupling cap specced in the datasheet goes after the resistor.

Also look at the LM350 3A adjustable regulator.  That will let you trim the output to get it closer to 5.0V.   Its got a lower thermal resistance so can run its junction cooler than a 78S05 (by about a factor of two) with the same input and output voltages, load current and heatsink.  Otherwise all comments above about the L78S05 also apply to the LM350.  N.B. it has a worst-case *MINIMUM* load current of 10mA, so either make the top resistor in the voltage adjustment divider no more than 120R (the datasheet shows 240R, and still needs an extra 5mA load to guarantee it always regulates), or add a power LED off its output with a series resistor chosen to make up the difference with a little margin.

There are more complicated circuits that pre-regulate the voltage to a L78S05 or LM350, or use a LM337 with an external pass transistor to split the dissipation between the regulator and another power semiconductor to make heatsinking easier, or one can go for a full-up LM723 based PSU design, with multiple external pass transistors if you have ridiculous load current requirements and want to be able to preset a current limit (or even more complex circuits to add an adjustable current limit) but that's getting into serious project territory and is over-kill for just running TTL logic.

[bitman]

Thanks Ian - do you know if there are any projects out there where the 350 is used? I currently have two 7805s and I guess I may need a 3rd. I also have a few 34063AP lying around that may work better here.

Btw. I guess I really goofed during testing. I'm feeling about 12-15v from the bench supply. Totally ignored the heat that generates and it may explain quite a lot.

[iainwhite]

The 78xx datasheet shows a sample circuit which combines a power transistor with the 7805.  Maybe give that a try?

Look at fig 13 on page 20:
https://www.fairchildsemi.com/datasheets/LM/LM7805.pdf

[Ian.M]

LM350 is effectively just a LM317 on steroids!    Take any LM317 project,  fit a big-ass heatsink (well over double the size as you need half the thermal resistance), double the transformer VA rating, double any secondary fuses, and halve the value of any resistors in series with the input or load, and you've got a LM350 project. 

[bitman]

The 78xx datasheet shows a sample circuit which combines a power transistor with the 7805.  Maybe give that a try?

Look at fig 13 on page 20:
https://www.fairchildsemi.com/datasheets/LM/LM7805.pdf

Well, I don't have any bd536s Checking the specs to see what I may have close to it.

[Ian.M]

It doesn't even need to be close as long as it can handle the load current and input voltage and has a high enough power rating.  Check the S.O.A. (safe operating area) plot  (Ic vs Vce) in the transistor's datasheet, specifically the DC line, not any of the pulsed time lines.for what it can handle.  Anywhere under the line with enough margin is good, after derating for temperature: The permissible current for all lines drops proportionately by multiplying by the derating factor (Tj-25°C)/(Tjmax-25°C). The factor is always less than one, and drops to zero at Tjmax.  Below Tj=25°C. you dont derate.  If there isn't a DC S.O.A line, in the most comprehensive datasheet you can find, you need to change transistors to one with a DC S.O.A ine.

You can even use a NPN power transistor - but you need a lower current PNP to drive it to make it work like a PNP. See https://en.wikipedia.org/wiki/Sziklai_pair.

[bitman]

Ian.M - I'll pretend I understood the bulk of what you wrote there and look for a way to use my TIP31G (NPN) that I do have in a circuit like this. As I stated, I have a lot to learn which is why I was looking for a complete project that would cover the basics. Thanks though - I'll go back to this as a reference once I get "wiser".

[mariush]

LM1085 can do 3A and it's super simple to use.

You can get the fixed 5v version for 2$, here's a link : https://www.digikey.com/product-detail/en/texas-instruments/LM1085IT-5.0-NOPB/LM1085IT-5.0-NOPB-ND/363564
Or you can get the adjustable version for which you can set the output voltage with 2 resistors (costs the same, 2$) :  https://www.digikey.com/product-detail/en/texas-instruments/LM1085IT-ADJ-NOPB/LM1085IT-ADJ-NOPB-ND/363565

You'll need a heatsink for it, because the regulator will dissipate the difference between input voltage and output voltage as heat on the regulator.
So from the heatsinks section, these will do : https://www.digikey.com/short/q3rmb1

0.4$ : https://www.digikey.com/product-detail/en/assmann-wsw-components/V2006B/A10761-ND/3476155
0.85$: https://www.digikey.com/product-detail/en/advanced-thermal-solutions-inc/ATS-PCB1049/ATS2091-ND/5030469
1.5$ : https://www.digikey.com/product-detail/en/aavid-thermalloy/529802B02500G/HS350-ND/1216357

All kinds of them in lots of sizes. Don't buy the smallest, look at the technical specs and buy something with a bit of surface.

You'll need a bridge rectifier to convert your AC voltage from the transformer to DC voltage. You could use 4 diodes to create one, but a bridge rectifier is sexy. You want 2A max, but the price difference between models is so cheap you may as well go with something bigger :

0.78$ 600v 6A : https://www.digikey.com/product-detail/en/smc-diode-solutions/GBU606TB/1655-1852-ND/7244880
0.93$ 600v 10A : https://www.digikey.com/product-detail/en/smc-diode-solutions/GBU1006TB/1655-1843-ND/7244871


So the linear regulator needs around 1v above 5v to output up to 2A and around 1.5v above 5v to output 3A, so your linear regulator needs to "see" at least 6.5v DC

You'll convert the ac voltage from a transformer using a bridge rectifier which will give you a peak dc voltage of 1.414 x the AC voltage from the transformer, minus 2 times the voltage drop on one diode from your bridge rectifier - that's gonna be around 1v.

You're probably going to go for a 9v AC or 10v AC transformer, since these are super common and cheap, and that means your peak DC voltage will be 1.414 x 10v  - 2 x 1v =  ~ 12v DC

How big the transformer should be ... well, you can sort of estimate with the formula   DC Current = 0.62 x AC Current

so for example, if you have a 25VA transformer, that means it can do 25VA / 10V ac = 2.5A ... and after rectification with a full bridge rectifier you'll have around 0.62 x 2.5A = 1.55A  ... so you need a slightly bigger transformer.

The cheapest on Digikey is 13$ for  a 10v AC 30VA transformer : https://www.digikey.com/product-detail/en/triad-magnetics/F6-10/237-1699-ND/4878628

This means your peak DC voltage will be around 12v DC ( 1.414 x 10v - 1.5v.. 2v )

So now you only need a big capacitor after the transformer, so that the minimum voltage the linear regulator sees is always above 6.5v .. but just be safe and go with 7v
You can estimate the capacitance with the formula :

C = Current / [ 2 x mains frequency x (peak dc voltage - minimum desired voltage) ]

So your current is 2A, you're in US so you're 60Hz mains frequency, your peak dc voltage is 12v and minimum desired voltage is 7v :   C = 2 / [ 2 x 60 x (12-7) ] = 2/600 = 1/300 = 0.003333 Farads or 3333uF
So you could go with a 3300uF 25v capacitor but 4700uF would also work.

It doesn't have to be super expensive, doesn't have to be thousands of hours at 105c rated, here's some  (3300uF to 4700uF 25-50v rated, sorted by price) :  https://www.digikey.com/short/q3rm2d
 
You may want to add a 1uF 50v or higher voltage rating ceramic capacitor on the input of the linear regulator (as close as possible to the pins) as well as a regular electrolytic capacitor on the output of the linear regulator, let's say 47uF or 100uF 10v or higher voltage rating on the output (they're so small and barely any price difference you could go with 25v or 35v rated capacitors on the output)


Job done.

When you're done, you'll notice a lot of energy is wasted as heat on the linear regulator .. you have up to 12v in, only 5v out .. so your efficiency may be below 50%. Then you can go on your next project, replacing your linear regulator with a switching regulator. There's loads of them out there.

[Ian.M]

Ian.M - I'll pretend I understood the bulk of what you wrote there and look for a way to use my TIP31G (NPN) that I do have in a circuit like this. As I stated, I have a lot to learn which is why I was looking for a complete project that would cover the basics. Thanks though - I'll go back to this as a reference once I get "wiser".

A TIP31G has a Tjmax of 150°C.  Its S.O.A. chart indicates its good for its full 3A Ic, (before derating) up to about 13V Vce.  Assuming you can keep its junction under 87°C it will be good for 1.5A.  You can make up the rest in the regulator, or parallel two TIP31Gs with 0.22R current sharing resistors in series with each individual emitter.   Its R_{th_JC} is 3.125 °C/W, so, assuming a big-ass heatsink that you can keep under 55°C, it can handle 10.4W, which at 1.5A is fractionally under 7V drop between the average input voltage and the output voltage.  That should be just enough for regulator headroom and ripple, though using two TIP31Gs would be a better choice, as the power is split between them and the current reduced so the maximum permissible voltage drop will be much higher and you can use a smaller heatsink.  Note that all the boosted regulator circuits need a minimum of 0.7V extra headroom (minimum voltage between in and out) over the bare regulator, because of the drop across the resistor driving the external pass transistor.  Also note, if they are going on the same heatsink, you'll need an insulated mounting kit for the regulator, because its tab isn't at the same potential as the TIP31G tab which is connected to the collector, so the whole heatsink will be sitting at Vunreg.  Putting thermal pads under the transistor to have an isolated heatsink (e.g. because you want it exposed out of the case back) would increase the thermal resistance too much - you'd need a copper heat spreader to use a larger area thermal pad to make it work and still need a monster heatsink, and at that point it would be cheaper to use either two TIP31Gs or a beefier transistor.

Its minimum H_FE is 10 @Ic=3A, and 25 @Ic=1A.  We need a reasonable extrapolation for Ic=1.5A.  Figure 8. DC Current Gain in its datasheet shows the typical  H_FE dropping from 65 @Ic=1A to 45 @Ic=1.5A and 25 @Ic=3A.  65/25 is 2.6 and 25/10 is 2.5, near enough the same ratio.   45/2.6 is 17.3, so we'll design to a minimum H_FE of 17 @Ic=1.5A

Therefore, to use it in a Sziklai pair, for 1.5A load current you need a PNP transistor to drive its base, that can handle at least 1.5A/17 which is just under 90mA (+ handle the dissipation for the full voltage drop).  If you parallel two TIP31Gs at Ic=1A each, you need 2*1A/25, which is 80mA so the base drive requirement actually drops! Its actual gain is probably quite a bit higher (see Fig 8 again), so for a one-off, if you are prepared to select on test, you could get away with a wimpier PNP driver transistor than sound design principles would require.   

What small to medium PNP transistors do you have handy?

[David Hess]

Higher current and power fixed 5 volt linear regulators like the LM323 used to be common and inexpensive but now the 3 amp LM350, 5 amp LM338, or their improved replacements are a better option.

Multiple integrated regulators can be configured to share the current load which has power dissipation and stability advantages however it requires an operational amplifier to enforce current sharing.

I like adding an external PNP power pass transistor or an NPN power pass transistor with a negative regulator.  If you do it right as shown below from the National Semiconductor 1980 Voltage Regulator Handbook, the integrated regulator's built in current limit can protect the power pass transistor.

Update: Had the wrong scan attached.

[bitman]

What small to medium PNP transistors do you have handy?

None useful. I got a handful of 2N3905s and similar transistors. I have some TIP transistors on order of which some are PNPs - but it will be quite a while before they arrive (if I'm lucky before Christmas). I've spent most my time using a lab supply for 5v and 200mA tops for a very long time so I never got "stuff" for higher amps/voltages. A few old grab bags had some higher caps but no transistors that I can use. Well and the fact that I haven't identified but a bit more than 50 % of what's in those bags it may very well be I have some but in a different package. From my best efforts I don't think I have any. It really strikes me how many components have NO markings what so ever on them?

Thanks again - reading the replies here is making me understand some important relationships I weren't aware of. Such as the size of the caps required. Btw. I also don't have resistors less than 1ohm and having spent quite a bit of today looking at tons of diagrams it seems I need some 0.x resistors.   Note, that most of the stuff I have is "cheap" ie. highly inaccurate components (10-20% off). From what I read that will also be a problem when it comes to designing simple power supplies, so I probably need to get ALL components needed instead of just getting a few things I am/was missing.

It's interesting to see the "tricks" you can play to deal with heat and efficiency. 

Ohhh - and I stumbled across this [ur]https://www.digikey.com/en/product-highlight/t/texas-instruments/webench-design-center[/url] ... a bit too much information for a sort like me, but it sure does provide some food for thought.

[Ian.M]

2N3905 is a little wimpy, but will probably be fine driving the base of a TIP31G in a Sziklai pair configuration to pass up to 1.5A as above.


Q1 is the 2N3905, Q2 is the TIP31G, then you can use the combo as the power PNP in the regulator datasheets external pass transistor circuits.  N.B. the simple ones don't have a current limit so don't short them or pop goes the external pass transistor(s), and when you remove the short, pop goes your 5V load as the pass transistor will now be a dead short.

[bitman]

2N3905 is a little wimpy, but will probably be fine driving the base of a TIP31G in a Sziklai pair configuration to pass up to 1.5A as above.

When you say wimpy are you talking about the total current? I have most of the ones in this set remaining: https://www.amazon.com/gp/product/B01LXA3O1V.
How do you easiest compare components btw? It's quite long winded to look up every component's data sheet to compare properties? Are there easier ways where you can specify a list and see how they compare?

[Ian.M]

My objection was the total power dissipation.  The driver Q1 sees the total voltage drop - Q2 Vbe, and may, worst case have to pass 90mA if Q2's gain is right on the lower limit.  Its maximum power rating is 625mW.   625mW/90mA is fractionally under 7V, so if the input to output drop goes over about 7.5V it may burn out Q1.  If so, to be 100% safe, you'd need to look for one rated more than 625mW.  A good clue would be a PNP with a larger package than the usual TO-92 small plastic D shaped one.  However as the typical TIP31G gain is more than double its worst case gain, odds are the 2N3906 will have plenty of margin, so if you can keep your finger on it without burning at full load current, it will do for a one-off.

The Old-Skool way to quickly find transistor comparison data was to look it up in Towers International Transistor Selector - a dead-tree book that listed most current and recently obsolete transistors in alphanumeric order of part number, with key parameters in a tabular format, and when the part was fairly generic, a brief list of possible substitutes.   There's a scan of an old 1975 copy at archive.org - Google for it, but newer editions are much harder to find.   There were similar books from other countries - the German ECA vrt books are particularly comprehensive.

Nowadays you use the parametric search tools on most big international distributers websites or try https://alltransistors.com/.   

N.B. for many Japanese transistors in through hole packages, the part number printed on the package starting with a letter is only part of the full part number, and it will often need 2S added in front.  See: https://en.wikipedia.org/wiki/JIS_semiconductor_designation.

When you've looked one up, if you've got stock, write a brief summary of the type, key parameters (Vceo/Vce, Icmax, Power, H_FE), and typical application on your stock card or draw label.

N.B. always double check the real datasheet when you find a candidate transistor - the selector guides tend to have a few mistakes in them.

[David Hess]

In the past, a TO-39 part (the 2N3053 was popular) would have been used as the driver and there were a lot of good ones but that is no longer the case.  Today, a TO-126 or TO-225 packaged transistor could be used.  It might be more economical to use a pair of TO-92 packages transistors in parallel and I have seen that done even when TO-39 parts were available.

It is a good idea to implement current limiting of the driver transistor by adding a collector or emitter resistor and maybe shunt diodes to draw off the base current during overload.  In the past, sometimes a low value low power carbon composition resistor was used as a fuse.  Today they make "fusible" resistors which are not suppose to fail with a short.

[MK14]

https://archive.org/details/TowersInternationalTransistorSelector

If you scroll down a bit, it lets you download PDFs and stuff.

[bd139]

Art of electronics chapter 9, which happens to be the free sample: https://artofelectronics.net/wp-content/uploads/2016/02/AoE3_chapter9.pdf

This covers linear and switching power supply design including thermal side of things, foldback current limiting (handle shorts without blowing up), crowbars (stop overvoltage destroying the device you are powering if something fails) etc etc.

I would personally, if I was to build one, build a linear supply for bench use. Transformer + rectifier + filter cap + LM723 + MJE3055T pass transistor + SCR based crowbar. Or a single LT1084.

But I'd probably buy one. It's way more expensive to build versus buy here. If you're lucky you might get a canned Farnell or Lambda supply from ebay with the above characteristics.

[David Hess]

With a design where the error amplifier is external, it is easy to implement remote sense to the binding posts for much better load regulation.  But integrated designs are competitive when you consider the leads between the binding posts and circuit.

The integrated regulator plus a discrete power transistor is by far the simplest way to make a blow out proof design though.

You can learn a lot building power supplies like these but if you want to save time, refurbishing a used HP or similar could be better.  I would not bother buying anything new.

[bitman]

But I'd probably buy one. It's way more expensive to build versus buy here. If you're lucky you might get a canned Farnell or Lambda supply from ebay with the above characteristics.

No debate here - it's time consuming and expensive. However, the goal isn't to have a power supply (well, I need it but still) it's to learn the core concepts here like power/heat issues and options. I've read plenty about liniar vs. switching supplies but that doesn't help me understand how to size the components (or know which ones are easy replacing others etc).

As an example (and this isn't an ask for answers), when I read the datasheet for the lm78xx I see formulas referring to Iq Ireg, Bq and Vbeq - I have NO clue what those refers to (Well, I get "I and V" but not the designation under them). It quite frankly wasn't clear to me how big a difference even a small difference in a resistor can make in these circuits. Which to me is a good thing - I love the learning quest.

I've got plenty of books and beginner projects books, but most are simple arduino type stuff ie. external power supply and plug a module here, there and code. I know code - code isn't a challenge. I don't get the physical hardwired stuff and finding good books that guide you through that is well HARD. They're either nothing but theory or so simple you don't learn anything. At least when I watch Dave's posts I can pick up stuff every time - but finding a reason/place to apply it is well, HARD   It's beyond my capabilities right now but it's still learning.

[tecman]

LM323 is a cheap, simple solution

paul

[bitman]

LM323 is a cheap, simple solution
paul

At US$40 a pop?? 

[David Hess]

LM323 is a cheap, simple solution
paul

At US$40 a pop??

Exactly, LM323s were inexpensive once but no longer although you might find one surplus for a reasonable price.  That just leads back to using an LM350, LM338, or one of the various improved TO-220 packaged adjustable regulators of which there are many.

One reason I like the idea of using a 7805 or 317 with an external transistor is that the current limit can be tailored to the right level.