New 10amp bench power supply output transistors question. pdf of project inc

[multime]

I am building a new bench power supply, See pdf for full instructions. I originally built this in the 90,s not from the Vellman kit but from scratch and foolishly sold it about 10 years later. I have a question regarding the five output transistors Tip3055, are these still the best option for this build or are there upgraded versions or new types.

On the original build i opted for the metal can versions 2n3055 but they take up a lot of space so i would like to use similar size devices to the Tips. I will also be using a DPDT relay to join the two 15vac windings in parallel up to around 13/13.5v when the relay will switch over and and join the windings in series for the up to 30v output. So at the lower voltages around 16amp is available.

I didn't use the Vellman case on the first build, i would have liked to but couldn't afford the Maplin kit at the time.

[pqass]

Digikey has a version of both TIP3055 and 2N3055 that is current and not marked obsolete.
The difference is $1.27 and  $9.06, respectively. 
Of course you can't get any of the former (zero stock) and there's plenty of the latter.
I can't find a cheaper non-marketplace NPN power transistor capable of 15A and 90W.
So, it doesn't appear to need changing.
There's plenty here: https://canada.newark.com/search?st=tip3055

[David Hess]

There are all kinds of better replacements for the poorly specified 2N3055.  The 2N3771G and 2N3772G are suitable for the oldest designs with an 0.2 MHz Ft.  The 2N3055AG and MJ15015 are suitable for 0.8 MHz Ft designs.  The MJ15001, MJ15003, MJ802, and maybe MJ15016 and MJ3055 are suitable for 2 MHz designs.  These are all simply better specified parts.

For a more modern design, I would also consider the TO-247 packaged TIP3055G which is 2.5 MHz.

[Doctorandus_P]

2N3055 was apparently introduced in the early '60-ies.
https://en.wikipedia.org/wiki/2N3055

It is a really very old design.
Your circuit also has 5 of them in parallel, so they only drive 2A each.

It can be replaced with almost any power transistor that can handle the voltages and currents in your application.

The most important thing to probably look out for is the speed of more modern transistors. Putting in faster transistors may upset your control loop and lead to instabilities, which usually can be tamed with a well placed capacitor.

[multime]

Thanks for the input and suggestions all,

 It was the obsolete bit that got me thinking there must be an upgraded version hence my question. As i said i would prefer to use the Tip or slightly larger physical size T0247 as the 2n,s take up to much room in the case which i might not have. I came across this Tip35c which looks to be a decent upgrade, what are you views on it.

And taking into account what Doctorandus_P contributed to the present set-up using 5 tips so only conducting two amp each (they still got bloody hot though until the fan kicked in) and with the increased current at low voltages to 16 amp the  five now passing just over 3 amp each, if the Tip35c look good could i get away with only using two. It would certainly simplify the heatsinking requirements, bearing in mind there will be a fan running as well as the heat sink contribution. As for the well placed capacitor, were would it need to be added.

 

[pqass]

The TIP35C is almost 4x expensive.  https://canada.newark.com/search/prl/results?st=tip35c&sort=P_PRICE|1
If the PS needs to output 1V @ 16A, the transistors will need to dissipate 224W (15Vrms-1V=14V*16A).
You're going to need to do at least a ball-park thermal calculation to make sure you have enough transistors to dissipate the max wattage.  It's going to be a lot less than the Total Power Dissipation spec'd in the datasheet.



Roughly,
junction-to-case+case-to-heatsink+heatsink-to-ambient
 =1°C/W(TIP36C)+1.2°C/W(mica)+1°C/W(large heatsink-guess)=3.2°C/W
That means for every Watt of power needing to be dissipated, 3.2°C rise in temp. will result.
So, starting with an ambient of 35°C and junction temp max is 150°C, that gives us 115°C headroom for each device.
However, each device-to-ambient has a thermal resistance of 3.2°C/W, which means we can only pump 36W (115/3.2) per device!
How many TIP36C devices will we need?  224W/36 = 6

Alternatively,
junction-to-case+case-to-heatsink+heatsink-to-ambient
 = 1.39°C/W(TIP3055)+1.2°C/W(mica)+1°C/W(large heatsink-guess)=3.6°C/W
...bla.bla..bla...
How many TIP3055 devices will we need?  224W/32 = 7

I'm sure someone will correct me.

[David Hess]

2N3055 was apparently introduced in the early '60-ies.
https://en.wikipedia.org/wiki/2N3055

It is a really very old design.

More than that, the 2N3055 encompassed many different manufacturing technologies as time went on because it was so poorly specified, so Ft can vary over more than a 10:1 range and safe operating area and current gain varies considerably.

The most important thing to probably look out for is the speed of more modern transistors. Putting in faster transistors may upset your control loop and lead to instabilities, which usually can be tamed with a well placed capacitor.

That is a real problem with the 2N3055 because its Ft varied so significantly over time, which is why I recommend the slower parts in my list as replacements in old 2N3055 designs.  This would not stop me from using a faster part, but I have a lot of experience tracking and and fixing stability issues.  Even so, it is sometimes easier to just use a slower part which is less likely to have problems.

It was the obsolete bit that got me thinking there must be an upgraded version hence my question. As i said i would prefer to use the Tip or slightly larger physical size T0247 as the 2n,s take up to much room in the case which i might not have. I came across this Tip35c which looks to be a decent upgrade, what are you views on it.

There is nothing wrong with the TIP series of transistors.  They are a little faster than newer 2N3055s and a lot faster than old ones, but that will not prevent them from working.  The On Semiconductor TIP3055G is a little slower than the other TIP parts but still faster than a common 2N3055.

[multime]

pqass, great video, thermal design shown simply. 

 David Hess, how exactly does the speed of the transistor make a difference. I never gave it a thought, put simply the way i saw it was turn dial select 10volt , i am not seeing where any switching is involved are we talking about keeping the supply stable at 10v, i thought that would be the job of the Lm723. I know the pass transistors need to follow it but i just didnt equate this design with speed more like a lumbering giant.

I have decided to go down the tip style route, Tip3055, Tip35c or Tip3055g. Just need to see if i can find a deal on them anywhere that doesn't involve over the top postage

 

[bobbydazzler]

I don't know what type of pcb you plan building this on (if using some type of perfboard at least use fr4) but I'd beef up the traces much more than the description suggests - 20awg solid core wire should do the job and being solid core it will help equalize the temperature across the pcb.

[David Hess]

David Hess, how exactly does the speed of the transistor make a difference. I never gave it a thought, put simply the way i saw it was turn dial select 10volt , i am not seeing where any switching is involved are we talking about keeping the supply stable at 10v, i thought that would be the job of the Lm723. I know the pass transistors need to follow it but i just didnt equate this design with speed more like a lumbering giant.

There are two different problems associated with changing the transistor's speed.

One is that it directly affects the frequency compensation of the feedback loop.  The bandwidth mostly depends on the output transistor's speed because it will be the slowest part inside the feedback loop, so depending on exactly how the frequency compensation is configured, changing it may cause oscillation or instability within the feedback loop.

The other one is layout related.  A circuit layout tested with slower devices may produce local oscillation if a faster transistor is substituted.  This is more of a problem when the power transistors are mounted remotely on a heat sink with long leads.  Sometimes a ferrite bead, small capacitor, or RC snubber is mounted close to the transistor pins to suppress oscillation.

Another factor not directly related to speed is that the SoA (safe operating area) of the transistor varies with process technology.  So the oldest (and slowest) 2N3055s which were built on a MESA process are very slow, but have a square SoA making them very rugged.  Later parts built on faster processes have a much more limited SoA.

[Doctorandus_P]

I sense a frowned brow about why the speed of the output transistor should be a factor when you "just want a steady 10Vdc output.

Loop stability is a serious concern with any control loop.
And loop is the magic word here.
Loop implies feedback.

To stay with your example: A steady 10Vdc output.
In your schematic one of the two uA723 voltage regulators measures the output voltage with a voltage divider, and compares it with it's internal voltage reference.
If it sees a too low output voltage, then it opens up the output transistors more, if it sees a too high voltage, then it closes the output transistors a bit.

There always is noise in the circuit, and the uA723, together with the whole output stage adds a lot of amplification (T2, T3 & T4 through T8).
So assume the noise makes a peak, and the uA723 measures 10.005V instead of 10.000V.
In that case it tries to lower the output voltage to get back to the set voltage, but it takes a bit of time (micro seconds, or some tens or hundredths of nano secons) to do so.

 If the control loop is maladjusted, then at the time that the output voltage reaches 10.000V, the uA723 has actually over corrected because of the the time delay of it's output signal through the output transistors, and even though it "stops lowering it's output voltage", the output voltage may shoot through a bit and go as low as 9.992V. So then a bit later, the uA723 notices the output voltage is now too low, (8mV, while it first was 5mV too high), and it will increase the drive level to the output transistors. And in this maladjusted example, it then will drive the output to 10.012V, en then to 9.985V, and this escalates until the whole power supply is a big oscillator.

The speed with wich this occurs is partly defined by the semiconductors, (uA723 itself, output transistors) and partly by (Usually) capacitors added to the circuit, which form RC constants.
Your power supply has the output capacitors (C4 & C9) as most power supplies do, and this lowers the speed with which the output can change voltage.

There also is C1. Which is connected to the raw output of the bridge rectifier on one side, and the parallel combination of R7 and R22 on the other side.
C1 = 150pF
R7 = 4k7
R22 = 22k

So these two resistors are 3k87 in parallel, and with the 150pF capacitor it forms a time constant of
3k87 *150pF = 580ns which resembles a frequency of around 270kHz.
So that would be the maximum speed at which the the output reacts to the voltage differences dictated by the uA723.

I may have some gross mistake here, It's been years since I last did this.
The whole thing is much more complex, because it involves the speeds of each and every component used, and even wire length and PCB routing can throw it off.

A quick search:
https://html.duckduckgo.com/html?q=power+supply+loop+stability
... shows up much more in depth information.

Here a method to measure the actual delays in a power supply circuit:
https://www.omicron-lab.com/fileadmin/assets/Training_and_Events/Webinar/2014-11_Webinar_LoopGain.pdf

And this search:
https://html.duckduckgo.com/html?q=home+built+injection+transformer
brings up:
https://adilmalikn.wordpress.com/2019/07/07/homemade-inject-transformer-for-psu-loop-analysis/

[multime]

bobbydazzle, thanks for your input, in answer to your pcb question, i make my own pcb,s, i may still have a photo copy of the last board i made for this design if i come across it i will upload it here. .

David Hess, thanks for the explanation i understand a little better now. Reading your text it looks like i may be in for some fun, 

[Kleinstein]

With the transitor speed it is not so easy to say a slower transitor is less likely to oscillate. There can be problems in both cases. A faster transistor is more likely to show local oscillation from long cables. A slower transistor can make the main regulator loop unstable under some conditions.  It is usually still OK with a well behaved load, but can cause trouble with capacitive load.

The robust and slow mesa type is quite rare now - mainly as 2N3773 and possibly fake.

[David Hess]

David Hess, thanks for the explanation i understand a little better now. Reading your text it looks like i may be in for some fun, 

I rather enjoy the empirical part of feedback loop design because I have regularly gotten results which confirm the theory, with the result that unknowns like the actual performance of the output transistor can be calculated accurately.  When I do my own designs, I include space for the compensation networks but almost never need all of the parts included.

It is possible to design the feedback loop to handle a wide range of output transistor characteristics but this is rarely needed or done.

[floobydust]

Just to mention, there is a nested feedback loop just for the pass transistors+driver stage, R6 and R17 setting a voltage gain of around 6 there. I find this arrangement needs a compensation capacitor across R6 to behave itself. But Darlington T3 BD646 is a slow part and might look after that.  OP if you run into drama, look there.

[perieanuo]

hi,
i used 2sd2688 (2 pcs) in 30v/10 output scenario, with 3v3/10a load test for like 5 minutes (the transistors have active cooler controlled by a microcontroller, i pwm the ventilator after reaching 40deg C and it reaches 100% after 60degC)
my old schematic tells me TIPL790A was also tested, but i don't remember, it was years ago)
the ps is used on regular base (i did used a 2xtl082 schematic circulating on the net with the mods done by the community and some of them done by myself, like current/voltage/temperature monitor with some arduino-like incorporated on bottom of pcb, used finally with ina219 module for I/V measurement and 100K NTC for fan temp control, oled as display)
not sure but i remember i put the schematic on eevblog somewhere...anyway i'll atach the pdf with the last revision, it have to be re-verified cause i redraw some of the schematic that works in my power supply (you know, with parts you actually can buy from farnell so the bom may be up-to-date, not with '80 components...). maybe you can cross it with your study
[edit] be careful, the 5V for atmega328(which is obsolete, should be replaced by newer micro) is either analog either dcdc (REG_5V_smps is AP1509-50SG-13), i did the layout for AP1509, it works great, but that tiny diode sod993 is killing you when you're doing the soldering...well, it's power efficient
otherwise, the L78M05ACDT variant is not power efficient but simplier to solder as hobbyst.

[perieanuo]

aaa, you choose the 723 approach. i don't recommend that one, i buildt 2 or 3 variants, spikes kill them pretty quick, but i don't want to develop in 20 posts why the 723 is a great circuit, but that schematic dissapointed me, after trying to add all sort of shoke on outputs, quicker/bigger reverse diodes etcaetera...my observation regarding the final stage was the main point

[CJay]

aaa, you choose the 723 approach. i don't recommend that one, i buildt 2 or 3 variants, spikes kill them pretty quick, but i don't want to develop in 20 posts why the 723 is a great circuit, but that schematic dissapointed me, after trying to add all sort of shoke on outputs, quicker/bigger reverse diodes etcaetera...my observation regarding the final stage was the main point

Would quite like to know what spikes you were seeing and what schematic to be able to kill it, also wondering what design you came up with to solve the problem?

[multime]

Thanks for all the contributions all, some great info.

I sense a frowned brow about why the speed of the output transistor should be a factor when you "just want a steady 10Vdc output.

How perceptive of you  , great explanation by the way.


I think the way to go is to build the thing, check work and switch on and proceed to test and see if any problems arise, I know Vellmans design worked, so any faults are down to my work or the new component types, the former is easy checked your then left with the one problem. I could switch on and have a totally fault free build  .

Talking of the build, the thing that is holding things up at the moment is sourcing a metal case. Dimensions around  L 10" x  W 10" x  H 6"  obviously at a reasonable cost.

 

[CJay]

Talking of the build, the thing that is holding things up at the moment is sourcing a metal case. Dimensions around  L 10" x  W 10" x  H 6"  obviously at a reasonable cost.

If you don't mind eBay, second hand and some work, you can pick up really useful enclosures dirt cheap plus you get the fun/utility of stripping out and maybe recycling bits from weird old industrial, scientific, oddball electronics.

F'rinstance the last one I bought for the enclosure was a Checkpoint Meto anti-theft loop controller which had easily double the purchase price worth of reusable components inside it as well as all the mains input/filtering and cooling I need too, just needed a 'nice' front (lasercut and etched acrylic) panel and slightly less metalworking than I'd have had to do if I'd bought something brand new.

[multime]

No don't mind used, as long as it hasnt been battered to death, the works no problem either and i am used to working with metal. I was looking on ebay earlier to day put something in my watch list but i think its a little on the wide side. What category are you searching under.

[CJay]

I tend to just browse surplus electronics, industrial electronics and look for 'interesting stuff' but that particular one came from a seller called CDL or Computer Disposals

I also use a handy website to get measurements from images

https://eleif.net/photo_measure.html

It can be pretty accurate, the Meto box was less than a cm different in any dimension when it arrived.

[multime]

Cheers, going to have a look now

[wizard69]

On the original build i opted for the metal can versions 2n3055 but they take up a lot of space so i would like to use similar size devices to the Tips. 2n3055

I'm not sue this is a valid concern.   Those alternative TIP cases will not go close together on most heat sinks as you will be dissipating do much heat in a tight area.    In fact the spacing required might not be that much different than using the T-03 cased transistors.   Lots of variable to consider here but a robust design may not be bigger, but the heat you need to remove from each transistor and the area required on a given heat sink to do that will likely dictate a rather large heat sink no matter what transistor you use.   You would need to walk through the equations to find out what sort of transistor heat sinking is required.    As a rough guess you would likely need to budget +50 watts per transistor, that is a lot of thermal power per transistor.   

T-03, at least considering them gives you other design options that may work to your advantage.   For example there are some really nice T-03 sockets out there.   for example: https://www.te.com/usa-en/product-2-1437504-9.html?mkwid=mobaVbvD%7Cpcrid%7C386964346943%7Cpkw%7C%7Cpmt%7C%7Cpdv%7Cc%7Cslid%7C%7Cproductid%7C8080-1G1%7Cpgrid%7C78782457763%7Cptaid%7Cpla-1016296564002%7C&utm_content=mobaVbvD%7Cpcrid%7C386964346943%7Cpkw%7C%7Cpmt%7C%7Cpdv%7Cc%7Cslid%7C%7Cproductid%7C8080-1G1%7Cpgrid%7C78782457763%7Cptaid%7Cpla-1016296564002&gclid=EAIaIQobChMI9eucpNT08QIVBgmGCh3PzA8CEAQYAyABEgKYkfD_BwE

[David Hess]

On the original build i opted for the metal can versions 2n3055 but they take up a lot of space so i would like to use similar size devices to the Tips. 2n3055

I'm not sue this is a valid concern.   Those alternative TIP cases will not go close together on most heat sinks as you will be dissipating do much heat in a tight area.    In fact the spacing required might not be that much different than using the T-03 cased transistors.   Lots of variable to consider here but a robust design may not be bigger, but the heat you need to remove from each transistor and the area required on a given heat sink to do that will likely dictate a rather large heat sink no matter what transistor you use.

That is exactly the case.  Once the minimum heat sink area is computed, there will be plenty of area for large transistor cases to be spaced out.  Smaller transistor cases actually have to be spaced out further because of their higher thermal resistance.  Sometimes it is advantageous to use two smaller transistors in parallel to space the heat out on the heat sink.

One nice thing about the TO-3 package is that it is easier to get the mounting pressure right because it uses two flange mounted bolts placed symmetrically.  The asymmetrical mounting of the TO-220 package for instance can result in the package area with the die getting bent up and way from the heat sink.  Larger tab mounted packages like the TO-247 are probably a little better about this.

Update: Mounting Considerations for Power Semiconductors