Design times for power supplies

[ndictu]

I was going through a book on power supplies (Power Supply Design Cookbook, 2001, M. Brown) and there is a table comparing linear and switching power supplies. There is the usual cost, number of components, efficiency etc.

But what surprised me is "Development time to production", which is 1 week for the linear supply, but 8 person-months for PWM switching regulator, and 10 person-months for "quasi-resonant switching regulator". This seems really high to me, I mean there are ICs that do most of the work for you, can't see what the 8 months are for.

Now, I'm a huge noob so maybe I'm missing something. I'm sure it it much harder to make SMPS but I think even I could do it in less time than that.

Edit: I know it's from 2001, no idea how things were back then. But all the reference designs in there use some kind of IC with some external components, pretty similar to what I've seen in modern designs.

[Alex]

Electronics is a very forgiving art. Almost anyone can put together something that 'works'. On the other side of the spectrum, sky is the limit. You can spend your life designing a perfect resistor, and in fact many people do.

SMPS are more complex than linear power supplies for the same specifications. Additional steps that are less important in linear supplies are involved, like design of magnetics, loop stability, efficiency optimisation, intereference suppression and many others. The times quoted are on the high side, but you should consider that a person like M Brown will have spent a great part of his career in companies offering custom and possibly state of the art power supplies to meet certain specifications. In this case one could say that these times are quite short. CAD tools, rules of thumb, previous designs and collaboration reduce this time, but the same tools can be used for linear supplies so the ratio of turnaround times for linear Vs SMPS does not change much over the years.

[saturation]

PSUs are just purchased today, and most are some form of SMPS.  You treat it much like buying a battery, specifying what you need, then choose a vendor.   There are many suppliers whose main products, if not only products, are PSUs of various capabilities so a designer can spend an entire career here.  Some advances over time have been in efficiency, size, weight, parts count, and cost.  Linear is increasingly a niche app these days.

[mikeselectricstuff]

But what surprised me is "Development time to production", which is 1 week for the linear supply, but 8 person-months for PWM switching regulator, and 10 person-months for "quasi-resonant switching regulator". This seems really high to me, I mean there are ICs that do most of the work for you, can't see what the 8 months are for.

Those are pretty ridiculous figures - maybe they include the  time to learn all the theory, build the whole thing from discrete components, hand-wind the inductors do lots of approvals.....
 

[Zero999]

I'm sure the figures quoted are for a mains powered custom design from scratch.

The book was also written over ten years ago when things were different. No virtually no one bothers designing an SMPS, they just buy one mass produced in China.

[ejeffrey]

There were a lot fewer options for integrated devices in 2001.  Also, the design of a line-operated SMPS is much more involved than for a battery powered buck or boost regulator.  They usually require custom magnetics, external MOSFETs, bootstrapping considerations, opto-isolated feedback, and so forth.  They have a lot more circuitry on the live side than a linear supply, so PCB layout for safety and isolation is a bigger concern.  Because a lot of the switching noise is conducted back onto the mains, they are much harder to satisfy EMC requirements.  Designing a supply that has good efficiency and ripple performance over the entire universal voltage range without grossly over-specing the components (especially the magnetics) is also non-trivial.  Designing a line-powered linear supply on the other hand, is nearly as simple as designing a low-voltage supply.

8 person-months is probably a bit long, at least with today's integrated controllers, but designing a good SMPS from scratch is still a lot more complicated than designing a linear supply.  So much that it is almost unheard of to design a power supply for a device.  If you are designing something, your first thought is going to be to use a wall-wart supply.  If a relatively low power single voltage supply is sufficient, even the most mass-market devices will use an off-the-shelf supply.  Option 2 is an OEM power supply to be integrated into your device.  These are available from devices like a computer power supply that have the power entry module, switch, and fuse built in, or devices that use screw terminals for power in and out.  You can get them in a range of powers and single or multiple output voltages.  If you are making lots of units, you might consider a semi-custom design where the PSU manufacture adapts a canned design for your needs, particularly if you need to fit in an odd shape.

What is quite common is to buy a single supply line operated PSU and generate additional voltages with on-board linear, switch mode, and/or switched capacitor supplies.  These are much simpler to design and test, especially if the auxiliary supplies are low power -- for instance negative rails for an analog front end.

[Neilm]

I'm sure the figures quoted are for a mains powered custom design from scratch.

The book was also written over ten years ago when things were different. No virtually no one bothers designing an SMPS, they just buy one mass produced in China.

There are still applications for designing an SPMS from scratch. For instance - I have recently designed one that has many kV isolation between the two circuits.

Neil

[qno]

I have spent several weeks optimizing and testing SMPS.

The problem is a dynamic load.
If your load is always the same it is relatively simple. If the load changes you need to test it in any situation.
Specially fast switching loads are a problem.
The control loop tend to swing in the standard configuration.
Also efficiency can be improved several % points by tweaking snubbers.


 

[EEVblog]

No virtually no one bothers designing an SMPS, they just buy one mass produced in China.

You are kidding right?
A SMPS module is the solution to everything? That's just crazy.
There are countless reasons why you'd want to roll your own SMPS circuitry into a product.

If you had said, "hardly anyone bothers to design a SMPS from scratch, they all use off-the-shelf controller chips and follow the app notes." then I would have agreed with you.

Dave.

[EEVblog]

The difference is that a linear supply it's dead easy to get "right" the first time, as all the parameters are easy to calculate, and the practical figures will match very closely every time. Rarely would you need to "re-spin" the design.
With a SMPS on the other hand you might be shooting for a specific minimum efficiency at a given load current across an input voltage variation, and that's going to depend a lot on the exact components you use (which might have cost/efficiency trade-off's) and possibly the exact layout etc. So it's very common to have to re-spin a SMPS design, even several times, to tweak it to the performance you require. Not so easy to calculate that stuff and get figures that match exactly, you really have to measure it.

Dave.

[EEVblog]

No virtually no one bothers designing an SMPS, they just buy one mass produced in China.

If you are talking about just external mains power adapters, then you'd be right.
But for anything that goes inside a product, there are countless reasons why you'd want to roll your own SMPS circuitry into a product.

A more correct statement would be "hardly anyone bothers to design a SMPS from scratch, they all use off-the-shelf controller chips and follow the app notes".

Dave.