Power supply: using discrete components vs regulator IC

[NewBeginner]

Hi everybody,

In my quest to build my own linear lab power supply I saw that many lab supplies are built using discrete components as opposed to readily available regulator ICs.
Without taking into account the lower power handling capabilities of the IC regulators (and maybe aside cost), what are the advantages/disadvantages of using discrete components and not regulators?
Building a lab supply using regulators is not a good practice? Why? I mean from a strictly technical/electronic point of view.

For example if I need a lab supply capable to deliver 3A, I could parallel 3 LT3080-1s. Why would this kind of design considered a bad design (in case it would be considered this way )?
I would like to find out what are the advantages/disadvantages of one over the other (from an electrical/electronic point of view and without taking into account the costs).


Thank you

[Bored@Work]

You will find out that you have to throw so many extra parts into it  to make it a decent lab bench that a regulator IC or a bunch of them hardly pays off.

[tom66]

The LT3080 is pretty pricey (around $3 each) and also only comes in a TO220 five-lead package. Compared to a power transistor in a TO-3 it can't compete on thermal resistance. The only possible advantage of an off-the-shelf regulator has to be the integral thermal limiting - but a well designed power supply should never get anywhere near the limit. You'd need to make your own control loop, so you're saving very little there.

[NewBeginner]

@Bored@Work Your answer was not helpful . I'm joking. I saw your signature . Could you give me a little more details about the extra parts?

@tom66 Could you elaborate a little more? What do you mean by "You'd need to make your own control loop"? I am new to electronics and still learning .

As I said in my first post I would like to know the advantages/disadvantages of discrete components over the regulators beside the costs or power dissipation capability. I understand that regulators can handle less power
than say a power transistor.
But other than this, from an electrical/electronic point of view what other advantages/disadvantages exists? I am not referring only to the LT3080 but to regulators in general.

I searched for more details on the net but all I could find was things like: regulators suck, they are pricey, they can't handle much power.
But other than that I couldn't find other reasons.
From what I saw the circuitry inside a regulator is more or less similar to the ones built with discrete components.

Let's assume that I don't care about costs and that my power requirements can be satisfied by paralleling multiple regulators or maybe even using a single regulator.
In this case what other advantages would I have in using discrete components over a regulator?

Thank you

[tom66]

@tom66 Could you elaborate a little more? What do you mean by "You'd need to make your own control loop"? I am new to electronics and still learning .

As I said in my first post I would like to know the advantages/disadvantages of discrete components over the regulators beside the costs or power dissipation capability. I understand that regulators can handle less power
than say a power transistor.
But other than this, from an electrical/electronic point of view what other advantages/disadvantages exists? I am not referring only to the LT3080 but to regulators in general.

I searched for more details on the net but all I could find was things like: regulators suck, they are pricey, they can't handle much power.
But other than that I couldn't find other reasons.
From what I saw the circuitry inside a regulator is more or less similar to the ones built with discrete components.

Let's assume that I don't care about costs and that my power requirements can be satisfied by paralleling multiple regulators or maybe even using a single regulator.
In this case what other advantages would I have in using discrete components over a regulator?

Thank you

If all you want is an adjustable regulator, you can use the control loop built into the IC. However, if you want any current limiting or voltage-controlled voltage-output, you need a separate control loop.

Other advantages of discrete components are future repairability (think repairing it in ~20yrs when the LT3080 isn't available any more but any generic high power transistor is, etc.) and it's more of a learning experience to design a power supply with just op-amps and discrete parts.

[saturation]

Mostly to get better performing lab grade PSUs.   Many industrial linear bench supplies were designed decades ago and makers are unlikely to change a design if the existing model still sells.   However, newer analog ICs are giving the old standby's competition.

Compare an HP/Agilent 3620 vs the LT3080, or others in both families of PSU or ICs.  3600 series has been around for 15+ years and is proven reliable, and still in catalog.  LT3080 and family ICs are ~5-10 years old.

A 40 + year old transistorized based design provides notable capability.  The Power Designs Precision PSU are IMHO, better exponents of all discrete design with high performance compared to today's newer analog ICs.  Its also fairly easy to build if one wanted too, given how simple the schematics are. 

See this thread:

https://www.eevblog.com/forum/product-reviews-photos-and-discussion/looking-for-goodinexpensive-bench-power-supply/


Note this are PSU for < 3A DC applications.


 

Hi everybody,

In my quest to build my own linear lab power supply I saw that many lab supplies are built using discrete components as opposed to readily available regulator ICs.

Thank you

[T4P]

In fact even modern analog power regulator IC's still cannot beat the tracking speed of the classic PSU's (the analog loop)

[NewBeginner]

@saturation, tom66 Thank you ,

Any other advantages/disadvantages?

If I understand this correctly the discrete components are preferred because they can offer
better performance (lower noise/ripple, better transient response etc ...) and easier maintainability (easier repair).
For example if you want to design a very low noise power supply (in the uV range) then you would probably go for discrete components.
But then again if you don't select your discrete components very carefully or if your design is not carefully thought-out, you could end up with an even worse performance than that
offered by a modern regulator IC.

On the other hand, I think, there is nothing wrong in designing a power supply around a regulator IC as long as you are aware of its performance limitations (noise, power, etc ...)
and you are satisfied with this limitations.

It's like in the case of cars: you can't expect from a family car to have the same performance, in terms of acceleration or speed, as a sport car .

In the past days I've tried to find more objective info on this subject, but without much luck.
Reading all those "regulator ICs suck" opinions (not backed up by arguments), almost made me believe that there is something really wrong with regulator ICs .
But I don't think that is the case.

Please do correct me if I am wrong. I am a beginner and I want to learn. But I don't want to end up saying that regulators "suck" just because most of the people on the net say that.
I want to be able to back up that claim (if it is indeed the case) with real arguments.

Maybe this anti-regulator wave was started by those copy-paste power supply builders who later on realized that the regulator based supply they have copied was not
able to "handle" their requirements. Next, they copied a well-proven, discrete component based, supply which met their requirements.
So they concluded that regulator ICs suck , without thinking that maybe the regulator based supply was designed with lower requirements in mind.

I don't know. Maybe I am just talking nonsense


Thank you

[T4P]

You do not even have to be careful just any opamp and a big beefy NPN pass transistor
Even better than scrolling through tons of pages of modern IC regulators to see their limits


IC Regulator are meant to be tweaked once and left forever in a equipment, in fact many LM317's and LM337's are used to power the opamps in classic power supplies!

[NewBeginner]

@T4P

Even better than scrolling through tons of pages of modern IC regulators to see their limits

I don't think that the fact that you need to search for the right IC, will disqualify a class of ICs from being used in a design.
If I understand correctly, when you design a product a lot of effort goes into selecting the right components with the right properties to make sure that you get the
desired performance for the final product.
I believe this applies not only to the field of electronics but to other fields as well.

Thank you

[T4P]

Ah well you don't understand what i mean, i was talking about lab PSU's

[NewBeginner]

Ah well you don't understand what i mean, i was talking about lab PSUs

I am not trying to be rude, but maybe you are not explaining things very well. I too was taking about lab/bench top PSUs.
Maybe not a lab-grade PSU (I am not sure what would take for a PSU to be rated as a "lab-grade" PSU) but a PSU used for everyday work in an electronics lab.
Anyway thank you for your feedback

Maybe other forum users could shed more light on the subject (from my previous message) with more details and objective arguments.

Thank you

[david77]

As you say there's nothing wrong using a regulator IC to design a PSU if you can live with it's lilitations and drawbacks. Even a very basic homebrew LM317 circuit can be sufficient for many people starting out in electronics. I've used one for some time and I still have it.
But for most people who're seriously into electronics the time will come that something better is needed. You'll want CV/CC operation, dual power, more power, current limiting and so on. All these things are hard to get in one single IC, they're all some form of compromise.
I'm not up to speed what's happening in the linear regulator market these days, but with a few jelly bean components you can make up the PSU that suits all your needs without too much trouble.

I'm all for people building their own PSU's but I would urge them to go discrete, be creative, salvage as many parts as you can from dead equipment. That way you can get something decent for very little money. A good PSU is as important as a soldering iron or multimeter in my opinion.
I'm not totally convinced by most of the cheapie chinese supplies we see today. They may work ok but I find them all very ... boring and samey.
Due to creative thinking my current main PSU cost me maybe 50 Euros or so and it gives me 2x0-35V/0-5A. And I learned something building it, how to prevent opamps from oscillating wildly depending on the load .

[M. András]

what about higher current opamps with built in adjustable current limit and supply up to 60 volts? opa549 8 amps continuosly. but you would need multiple tabs or some pre regulator preferrably some switcher. i know this isnt for a beginner. but the multi tap transformer+that opamp couldbe made easly shematic is in the datasheet

[merlinb]

what about higher current opamps with built in adjustable current limit and supply up to 60 volts?

They're great, but they're also very expensive! Discrete power transistors are cheap as chips. Money is a powerful motivator!

[Rerouter]

hmm, what about something like this for a semi discrete regulator,

http://tinyurl.com/c5z4xke

as it gets away with a much lower voltage and thus cheaper and faster op amps,

[T4P]

Well a "semi-discrete" is what i know as a discrete PSU!

[Rerouter]

just that by my logic, fully discrete would involve discrete components making up the op amp :/

either way i am slightly curious if that design would work out, as it would vastly simplify current limiting for me (have a transistor pull down the base of the regulating npn)

the main issue this layout leaves me with is the 2 transistor Vbe drops, i have heard of transistor biasing, but not found sources to lower the effective Vbe junction voltage (you can buy pre-biased transistors with 250mV Vbe and am left wondering how they are different)

[saturation]

You're welcome. 

There are EEs today whose entire careers are spent designing PSUs, linear and SMPS.  This holds whether its a module for a product [ like the one's in desktop PCs] or bench supplies for labs.   

However, like all tricks of analog, you should design your own PSU for learning purposes; make it easily disassembled once tested and working, so you can recycle parts for another project.  A simply linear and a basic switching supply would cover all the bases for your lifetime.  This would not only teach you core concepts of how they are made, but how to evaluate them too.  You'll then appreciate all the work done in commercial supplies, and the regulations they need to meet.

If you built a supply from new parts with all the bell's and whistles of a purchased lab supply, it will likely cost more than a top rated supply, and that's already with the 50%+ profit margin of the seller.  For at least for 2 reasons, not counting the cost of your design and debug time: manufacturers buy parts at bulk rates while prototypers buys parts at single unit cost; manufacturing may be partly by robot versus manual, further reducing costs.

Lastly, there is a large second hand market of good lab PSU, most don't fail if not abused.  Just getting a good one, even damaged, and repairing it will be worth your time as it will last you another lifetime.

If you look at the image of Jim William's desk, most of those supplies are from the late 1960s, early 1970s.

http://www.eetimes.com/electronics-blogs/social-mania-blog/4217103/How-messy-is-your-desk-

@saturation, tom66 Thank you ,

Any other advantages/disadvantages?

If I understand this correctly the discrete components are preferred because they can offer
better performance ... But then again if you don't select your discrete components very carefully or if your design is not carefully thought-out, you could end up with an even worse performance than that
offered by a modern regulator IC.

there is nothing wrong in designing a power supply around a regulator IC as long as you are aware of its performance limitations (noise, power, etc ...) It's like in the case of cars: you can't expect from a family car to have the same performance, in terms of acceleration or speed, as a sport car .

In the past days I've tried to find more objective info on this subject, but without much luck.
Reading all those "regulator ICs suck" opinions (not backed up by arguments), almost made me believe that there is something really wrong with regulator ICs .
..,,
Thank you

[T4P]

mmm, 1970's My one was also made in the 1970's EEVBlog Member Digsys has the same model but more power