Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.
Status
Not open for further replies.
I have been collecting various bits. I have about 2000 0.25W resistors but only a few caps mainly ceramics. I have just about enough to make a couple of little LM317 regulators with variable pots.
I sent away for and have :- heat shrink (a lot different sizes) , 5 prototyping board (just circular copper pads), a 4 digit V & A meter, an assortment of LMXXX and not much more. I saved a lot of the coloured wire from the PC power supplies.
I am going to send away - my second order to Banggood - for 10 pairs of leads with crocodile clips (I only have 2), 10 pairs of banana plugs and sockets, a selection of fuses, fuse holders. 2 cheap V & A meters for the 4A 20V thing, 10 knobs with 1 - 10 on them. Banggood also have a rotary switch that looks OK.
Sounds like you are going to have a pretty comprehensive electronics/mechanical workshop. I am hoping to do some physical when our other house is refurbished and we move in. The present house has no room for a decent workshop.

I was going to make this thing roughly modular. The transformer, rect. for the 36V and 12V (fans and meter supply) are in one case. This is up and running. Attached to this is the second case for the regulator heat sinks. I think I will have to take it apart and rebuild with a third one on the end. No 'walls' between them. It's held together on 15mm aluminium channel underneath. It looks better than it sounds.
Modular is good, and there is no reason why the transformer, rectifier, and res capacitor cannot be in a separate case, providing there are thick wires between the rectifier case and the voltage regulator case. But, I am afraid that it would be unwise to separate the functions further. The reason is that long wires can induce frequency instability and compromise the PSU performance. In fact the layout, of the power transistors and the controlling electronics is critical- sorry

Should I add some photo's and maybe we should start a new thread?
Photos would be good- no need to start a new thread. Photos of your PSU will be in the context of this thread.:)

I have just seen your latest circuit.
I wanted variable Amps but didn't like to ask for it.
I was fine until I saw the words 'negative regulator'.
The circuit of post #58 is very simple to understand and is normally well behaved (famous last words). There have been thousands of similar circuits built.

A negative three terminal regulator (TTR) is exactly the same as a positive TTR that you are familiar with, except the polarity of all voltages are reversed. The LM337 TTR is the negative version of the LM317 TTR.:)

I'll concentrate my thoughts on the hardware...
Good move. The hardware is the critical area, in my opinion.

When you say 'massive conductors' do you mean the sort of wire in a PC PSU - 20A stuff or are we talking copper busbars.
Conductors are other items that can never be too big, from an electronic point of view. Ordinary heavy gauge wire, around 14 SWG to 18SWG will be fine for the heavy current paths. The way that the wires connect is also important, especially to ensure good current sharing between the power transistors.

If you could possibly add a few 'for example' names to the components it might help me look that type up and understand a bit more. I understand everything is up in the air - just a rough idea might help.
Will do.:)

spec
 
Last edited:
QuickCopperExport.jpeg


I seen this variation in an electronics mag years ago.

EDIT: This was meant to be a reply to post #20, I didn't realise there were so many more posts!
 
Last edited:
9910136100_1414877720.jpg




This is a circuit that I've actually built and tested. I used a bd140 for the PNP and a tip3055 for the NPN, so should be fine with a 2n3055. I believe the circuit originated from one of the old National Semiconductor data sheets or app notes for the LM317.
 
View attachment 102646



This is a circuit that I've actually built and tested. I used a bd140 for the PNP and a tip3055 for the NPN, so should be fine with a 2n3055. I believe the circuit originated from one of the old National Semiconductor data sheets or app notes for the LM317.

Thanks GB. This approach effectively converts the NPN power transistor into a PNP transistor. One minor problem is that the loop gain is pretty high and you have to be aware of frequency stability. But once you get that sorted, the performance would be pretty good.

spec
 
Last edited:
Sounds like you are going to have a pretty comprehensive electronics/mechanical workshop. I am hoping to do some physical when our other house is refurbished and we move in. The present house has no room for a decent workshop.


Modular is good, and there is no reason why the transformer, rectifier, and res capacitor cannot be in a separate case, providing there are thick wires between the rectifier case and the voltage regulator case. But, I am afraid that it would be unwise to separate the functions further. The reason is that long wires can induce frequency instability and compromise the PSU performance. In fact the layout, of the power transistors and the controlling electronics is critical- sorry


Photos would be good- no need to start a new thread. Photos of your PSU will be in the context of this thread.:)


The circuit of post #58 is very simple to understand and is normally well behaved (famous last words). There have been thousands of similar circuits built.

A negative three terminal regulator (TTR) is exactly the same as a positive TTR that you are familiar with, except the polarity of all voltages are reversed. The LM337 TTR is the negative version of the LM317 TTR.:)

Good move. The hardware is the critical area, in my opinion.


Conductors are other items that can never be too big, from an electronic point of view. Ordinary heavy gauge wire, around 14 SWG to 18SWG will be fine for the heavy current paths. The way that the wires connect is also important, especially to ensure good current sharing between the power transistors.

Will do.:)

spec
Hi spec. I should warn you I have a rather strange sense of humour (not humor - perish the thought)
Like here.
I found that XP welder, it's spot on 40.00VAC. That means we can make a 56VDC x 80A linear converter spec. Super. I think it looks like maybe a 30A continuous output.
I weighed it at just over 7Kg. It has a huge thermal cut out that tripped every few minutes. The wiring though is very strange. I know it sounds mad but it looks like there is a connection from the outside of the primary windings down into the secondary. It's a simple 2 coil one above the other transformer. Just for fun I will remove it from it's aluminium case and after getting the dust of ages off it take a few photo's and scribble a wiring diagram. It's got to be about 30 years old. Can you imagine the comments here if I suggested a 1200Watt (3200W peak) linear power supply. It's enough to give a chap an organism. What's 128 x 2N3055's between friends. I expect you already have a circuit.

In regards to modular.
The 52V DC power is in 1 case with a fan on the front. There is also a 12V AC - 17 V DC 1.6A supply I'll convert to 12V/14V DC (variable - LM338 at 5W) for the fans (11W ?) and meter(s). The next case is for the heat sinks with a fan on the top or bottom or both and the last case is will be your circuit with another fan if necessary. All 3 cases with the adjoining walls of the tops removed except for a small 10mm section to give stiffness to the roof. Depending on where the circuit for the regulator is placed I would estimate that the furthest power transistor could be a 200mm wire to the regulator board in the third case. Is this too long?

That will effectively makes a case 450mm L x 140mm D x 86mm H. I had better submit a few photo's.

I'll not start until you have perused this - at the moment it's only 2 cases, you don't think I'll need 4 do you ?

'very simple to understand' - Oh boy have you got your work cut out. I started to dribble just looking at it.
Correct.
:happy: I did! (see post #49)
Not correct. A true low impedance 0V output for a bench power supply is absolutely essential.:wideyed:

I have heard of 'wiggly amps' but never 'wiggly bits'.:joyful: You are not ex Royal Navy are you?

spec
As a matter of reference would 2 diodes drop 1.25V to 0.0V ?
No - not Royal Navy - but my great grandfather nearly had it up the Khyber Pass - not the one your thinking of. Those Fuzzy Wuzzies don't you know. Something like that. I am actually being serious.
Anyway I never did find that 'striped paint' and ' G rivet'.
In regard to politicians Sir Winston Churchill once said "What's the point, half of them think it's impossible and the other half are at it".
An MP had wanted to submit a bill to the House of Commons about - I'm sure you can guess.

Please let me know if 200mm is too long a wire between power transistor and regulator board. Bit of a problem if it is. Means degrading the cooling a bit.
Just had a thought. I might get it down a bit and make the heat sinks easier to mount without loosing much cooling
 
As a matter of reference would 2 diodes drop 1.25V to 0.0V ?

Perhaps I led you astray. The forward voltage drop of a normal silicon junction diode is around 0.6V at a low current, taking into account the current rating of the diode, that is. This would be at 25 Deg C. The forward voltage drops by 2mV per degrees C. A Schottky diode is the same except its forward voltage is around 0.4V, and even as low as 0.3V for some high conductivity types. That is why Schottky diodes make more efficient rectifiers. Schottky diodes are also fast so they are always used in switch mode power supplies, which operate at a relatively high frequency (10KHz to 4MHz).

The problems with putting two diodes in series with the output of an LM317, in order to get zero volts are:

(1) The forward drop of the two diodes would vary with temperature, so the output voltage of the PSU would vary with temperature.
(2) The forward drop of the two diodes would vary with current, so the output voltage of the PSU would vary with current drain from the PSU.
(3) The forward drop would vary from diode to diode, so you would not exactly cancel the 1.25V nominal reference voltage of the LM317, especially as the LM317 reference voltage varies from one LM317 to another anyway.
(4) There are more complex reasons why the two diode approach would not be satisfactory, but it would take too many words to explain.

Having said all that, ordinary diodes (1N4148, 1N400x) are used as voltage references in circuits.

Please let me know if 200mm is too long a wire between power transistor and regulator board. Bit of a problem if it is. Means degrading the cooling a bit.
Just had a thought. I might get it down a bit and make the heat sinks easier to mount without loosing much cooling
200mm is a bit long, but don't compromise the power transistor cooling- that is paramount.

spec
 
Hi spec, somehow I clicked the wrong thing and lost you. Finger trouble, all's well.
Thanks for clearing that diode business up. One does read some strange ideas on the Internet.

I was going to place 4 finned heat sinks in a square with the fins facing inwards. There is a dead spot directly in front of the motor. The other reason for this configuration is that all 4 have the same airflow. All the connections to the power transistors are on the outside of the square. I was going to mount this arrangement as a single insulated unit. Air flow out of a fan is very complex. Tip clearance is critical and it is surprising how many so called high quality fans have a large gap.
If I place two heat sinks with fins facing each other then a gap and two more facing each other providing the fan blades cover all 4 the air flow will be pretty constant over them. Connecting the 4 heat sinks together is simplified as is an insulated mounting to the case. One could use crude baffles to force the air down the fins of the heat sinks, it might even be better that my original idea. It would roughly reduce the length of the wires to 75mm to case 3 and then whatever distance to the control connection, say a total of 125mm.

I think I am getting a bit carried away here spec. Some people say I should be.

Will it still be OK to mount the power transistors in electrical contact with the heat sinks using just thermal paste. It is by far the best cooling solution?

I'll start adding case 3 and take some photo's.

We had a power failure this morning. I have just replaced the 2 batteries in my UPS. Blasted warning fires off even if the PC is switched off. What a way to wake up.

JP
 
Will it still be OK to mount the power transistors in electrical contact with the heat sinks using just thermal paste. It is by far the best cooling solution?
It is quite true that for best thermal conductivity the power transistor insulating washers should omitted, but this would mean that the heatsinks would all be at roughly 52V, which should be avoided for practical reasons. You best bet wold be to use a TO3 aluminum oxide insulating washer or, next best, mica. You will also need a TO3 mounting kit, comprising:
(1) two insulating stepped washers,
(2) two metal bolts,
(3) two metal nuts
(4) one metal connection tab (for soldering a wire to the collector, which is connected to the case)
(5) three metal washers

By the way, a fifth power transistor will also requires a heatsink. (UPDATE of 2016_12_10) The fifth power transistor can also be an MJ2955 in a TO3 case, if you like. Otherwise it could be a TIP42C, which is in a smaller TO220 case. The fifth power transistor will only dissipate around 10W maximum though. Each of the four 2N3055 will dissipate 52W maximum, which will be OK, due to your Rolls Royce heatsinks and fan cooling.:cool:

spec
 
Last edited:
Hi Spec I agree entirely with a screwdriver between those heat sinks and the chassis but using insulating washers will take the C/W up about 30%.
RS, CPC, Maplin and everywhere else I have tried no longer stock TO-3 mounting kits. CPC do stock a rather inferior sheet material that one can make an insulator out of.
I have everything except the insulating material. Any ideas where I might obtain some decent insulating material otherwise it's going to be 'frying tonight'.

You did mention more modern power transistors than the 2N3055 ?
I do have the 60V LM317AHVT - not enough power for the 5th power transistor I assume?
No problem.

The 4A transformer is the one I added a few more turns to to get 20V DC. I think 3A is more realistic. I thought a LM338 should be OK but I can fit 2 LM338's driven by a LM317 if I need to. ('Fun' project). Rather nice little heat sink.

The first 2 photo's are the battery charger. The lamps are so shorting out is not a blown fuse.


Charger1.jpg Charger2.jpg Supply2.jpgSupply4.jpg

Not quite as I expected but never mind at least they are there.
 

Attachments

  • Supply1.jpg
    Supply1.jpg
    51.4 KB · Views: 246
Last edited:
Hi
RS, CPC, Maplin and everywhere else I have tried no longer stock TO-3 mounting kits. CPC do stock a rather inferior sheet material that one can make an insulator out of.
I have everything except the insulating material. Any ideas where I might obtain some decent insulating material otherwise it's going to be 'frying tonight'.

Hi John,

Here are some TO3 mounting components:
https://uk.rs-online.com/web/p/heatsink-mounting-accessories/0263239/
**broken link removed**
https://uk.farnell.com/webapp/wcs/s...tegoryId=700000005562&langId=44&storeId=10151
**broken link removed**
**broken link removed**

spec
 
You did mention more modern power transistors than the 2N3055 ?
I do have the 60V LM317AHVT - not enough power for the 5th power transistor I assume?
John, the LM317H is a three terminal voltage regulator, rather than a transistor.:)

spec
 
Hi

Hi John,

Here are some TO3 mounting components:
https://uk.rs-online.com/web/p/heatsink-mounting-accessories/0263239/
**broken link removed**
https://uk.farnell.com/webapp/wcs/s...tegoryId=700000005562&langId=44&storeId=10151
**broken link removed**
**broken link removed**

spec
Hello ARC Angel - ne spec.
I phoned RS, CPC and Maplin. Maplin said they no longer did TO-3 insulated mountings. I was very surprised and asked them to check. When they phoned back I pointed out that the insulated mounting was for the TO-3 style and not just the 2N3055. I said that they were only selling half the product and without the mounting any TO-3 was useless. The chap went away and returned and said they were running this type of transistor down as it was old fashioned.

For CPC you have to book an appointment with their technical assistance branch. The chap at CPC even went so far as to find the sheet material.

What doesn't help is unless you know the correct description their searches are a waste of time. If technical assistance at CPC can't find an item I stand no chance. I tried TO-3, TO-3 mounting, TO-3 insulation nothing. It was suggested I use some form of double sided self adhesive heat transfer material for CPU's or the heat conductive pads they do in various thicknesses up to 5mm. It was at this point I saw the video on how poor these insulators are at transferring heat, even the best compared to just paste. Ah - and light fell upon the face of the earth. Your url is not CPC-Farnell but Farnell. The sites look almost identical.

Something very strange going on. Both Farnell and CPC Farnell are both 'under maintenance' with identical web pages. The plot thickens.

My order with DigiKey was 3 x LM317AHVT's and 4 x 2N3055's. That came to about £13 - with £17 odd for delivery. I assumed with an 0800 UK number I was dealing with a UK stockist. I had been told the LM317HV was the 60V version of the LM317. The only place I could find it was DigiKey. DigiKey new that a LM317HV was the LM317AHVT but CPC didn't, it doesn't exist. - they do however stock the LM317AHVT. £0.65P each. One is not amused.

Sorry about calling the LM a transistor, good job I didn't call it a tranny.

I wonder why they miss the thermal conductivity off the description of those TO-3 mountings.

If you can't sleep read this :-
**broken link removed**

The IBM 4381 was the largest air cooled main frame IBM did back in the 80's. I had the pleasure of using one for just over 10 years. This cooling problem we have pales into insignificance when I remember that beast. As an engineer quite often what looks right is right. The plenum design for cooling the processors in the 4381 looked absolutely gorgeous. We had it upgraded and that included a great pile of bits. I only wanted one, never got it. The solid copper bus bars that thing had were straight out of Flash Gordon. Never failed once in 10 years. IBM don't even comprehend what a 'blue screen of death' is.

Any thoughts on - You did mention more modern power transistors than the 2N3055 ? - or am I dreaming again. I am happy to order a TIP41C (or 2).
I am yours to command as to what components we need and how we use them.

When I was looking at the figures that chap wrote down during his experiments on mounting the TO-3 I noticed that the temperature difference between the heat sink directly behind the TO-3 and the heat sink edge about 50mm away was no more than 4C during all the tests. However the TO-3 case to the heat sink a few mm away was very much higher at 10C even under the best test conditions. At a guess I would say there is rather a limited surface area to conduct the heat. I think I just stated the obvious.

Sometimes spec lady luck not only stands near you but takes her home with you. These cases are held together underneath by 2 x 15mm aluminium channel rails. I had the choice of extending them or taking the whole thing apart and making new ones. I found a piece of aluminium bar that is a perfect push fit into the channel. I would say about a 0.01 mm interference fit. A drill and tap and I can add cases to my hearts content.

Indecently the 52V transformer is wired for 26V with small caps in the photo.

Thanks spec.
 
Something very strange going on. Both Farnell and CPC Farnell are both 'under maintenance' with identical web pages. The plot thickens.
CPC and Farnell (Element 14) are all part of the same group. CPC caters for repair shops and consumer, while Farnell are much much bigger and cater for industrial.

They have suspended their web sites for maintenance (so neither website will work properly for about six months.:cool:)

spec
 
I wonder why they miss the thermal conductivity off the description of those TO-3 mountings.
The thermal resistance will be in the product datasheet. The thermal resistance of a TO3 mica washer is between 0.75 degrees centigrade per Watt (DCW) and 1 DCW, providing the mating surface is flat, the correct tension is used for the mounting screws/nuts, and heat sink compound is applied to both sides of the insulating washer.

If you can't sleep read this :-
**broken link removed**
Wow- you must be a mechanical expert- my head hurts.:arghh:

spec
 
You did mention more modern power transistors than the 2N3055?
There are hundreds of power bipolar junction transistors (BJTs) that outperform the 2N3055/MJ2955.

The 2N3055/MJ2955 suffer from the following shortcomings:
(1) Current gain falls off badly after about 4A
(2) High thermal resistance, junction/case, of 1.5 DCW
(4) High saturation voltage, collector/emitter
(5) High base/emitter voltage
(6) Low, frequency response of 300KHz and 2.5MHz, depending on which version of the 2N3055 you get.
(7) Relatively low (for a power transistor) collector/emitter voltage of 70V
(8) Relatively limited SOA

The TIP3055/TIP2955 are plastic case versions of the 2N3055/MJ2955, but there are some differences in the specification and in my opinion the TIP35C/TIP36C would be a better substitute.

The ON Semiconductor, MJ15003G/MJ15004G, are generally considered to be the the modern replacements for the 2N3055/MJ2955.

A really nice pair of transistors are the, MJL3281A/MJL1302A (just look at the flatness of the current gain graph).:)

And if you want to handle some serious current and power, the MAG6332/MG9412 are the boys.:D

spec

PS: Afraid to say that, apart from audiophile power amplifiers, power MOSFETs have completely replaced power BJTs

DATASHEETS
(1) https://www.onsemi.com/pub_link/Collateral/2N3055-D.PDF
(2) https://www.onsemi.com/pub_link/Collateral/TIP3055-D.PDF
(3) https://www.onsemi.com/pub_link/Collateral/TIP35A-D.PDF
(4) https://www.onsemi.com/pub_link/Collateral/MJ15003-D.PDF
(5) https://www.onsemi.com/pub_link/Collateral/MJL3281A-D.PDF
(6) https://www.rlocman.ru/i/File/dat/Semelab/Transistors_Bipolar_BJT_Single/MG9413.pdf
(7) https://www.rlocman.ru/i/File/dat/Semelab/Transistors_Bipolar_BJT_Single/MG9412.pdf
 
Last edited:
I phoned RS, CPC and Maplin. Maplin said they no longer did TO-3 insulated mountings. I was very surprised and asked them to check. When they phoned back I pointed out that the insulated mounting was for the TO-3 style and not just the 2N3055. I said that they were only selling half the product and without the mounting any TO-3 was useless. The chap went away and returned and said they were running this type of transistor down as it was old fashioned.

For CPC you have to book an appointment with their technical assistance branch. The chap at CPC even went so far as to find the sheet material.

What doesn't help is unless you know the correct description their searches are a waste of time. If technical assistance at CPC can't find an item I stand no chance. I tried TO-3, TO-3 mounting, TO-3 insulation nothing. It was suggested I use some form of double sided self adhesive heat transfer material for CPU's or the heat conductive pads they do in various thicknesses up to 5mm.
Yes, some electrically insulating washers are also pretty good thermal insulators. The rundown on heatsink material in order of lowest thermal conductivity is:

(1) Beryllium oxide: lowest thermal resistance (apart from diamond), brittle, highly carcinogenic, highly expensive.
(2) Aluminum oxide: brittle, medium cost, best choice, if you can get it
(3) Anodized aluminum: danger of puncturing anodized surface, expensive
(4) Mica: excellent electrical and thermal characteristics, brittle, low cost. Unless you buy from top-line manufacturers, like Avid, the thickness of the mica varies greatly, and thus the thermal resistance also varies accordingly. Mica is also a good choice.

And, quite simply, the other materials are not worth bothering with, if you need low thermal resistance that is.

The common thru-hole cases are:
(1) TO3: Rolls-Royce, in all respects, low thermal resistance, even and consistent contact, high reliability, does not need a clamp, expensive, getting rare.
(2) TO220: compact, easy to mount, needs a clamp for consistent thermal contact and reliability, cheap, the most common thru-hole package
(3) TO218; bigger version of TO220, with correspondingly lower thermal resistance. Also needs a clamp for even pressure and consistent heat sinking. A rare package now. Will fit to a TO3 PCB footprint.
(3) TO264/TO247: even bigger bigger version of TO220, with correspondingly lower thermal resistance. Also needs a clamp for even pressure and consistent heat sinking. A nice package.:cool:

**broken link removed**
**broken link removed**

But thru hole cases are rapidly being phase out, and many of the latest and greatest power devices, especially MOSFETs, are only available in surface mount (SM).

There are a bewildering number of different SM cases and the industry has made an awful mess of case identity- inevitably on data sheets the same case will be referred to by different reference numbers in different places, so it is practically impossible to determine which case is which, and when you get a data sheet that covers multiple devices and multiple cases, you may as well give up.:arghh: Why this situation has arisen, I do not know.:wideyed:

**broken link removed**​
spec
 
Last edited:
The 2SC5200 (npn) & 2SC1943(PNP) are nice linear gain power transistors and as such are popular with audio amplifier builders. I expect they could be useful in power supply circuits too. I have read elsewhere that good amplifiers and good power supplies have a lot in common.

Spec: this thread is an absolute treasure trove on practical power supply building, have you considered writing a book? I think there would be a big demand to read this practical advice rather than pages of intimidating mathematics as in certain books! Iam thinking of the superb book, "The Art of Electronics" which without a great deal of mathematics manages to convey a very in depth discussion on most aspects of electronics. John Potter or anyone else reading who is still relatively new to electronics, I would heartily recommend the aforementioned book. Anyway spec I would read your book if it was full of this sort of practical and non mathematical advice on power supplies.
 
The thermal resistance will be in the product datasheet. The thermal resistance of a TO3 mica washer is between 0.75 degrees centigrade per Watt (DCW) and 1 DCW, providing the mating surface is flat, the correct tension is used for the mounting screws/nuts, and heat sink compound is applied to both sides of the insulating washer.


Wow- you must be a mechanical expert- my head hurts.:arghh:

spec
Just a bit of light reading before lights out. You must be joking, I gave that sort of thing up years ago - and I was never any good at it then.
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top