Underestimating Complexity of Power Supply Design

Problem

=20

Power supplies are sometimes considered trivial = designs=20 problems, often relegated as first projects for engineers joining = industry,=20 usually with disastrous results.

These disastrous results include schedule slips, = cost=20 overruns, and excessive field failures, with associated warranty costs. = The=20 root of the problem is usually underestimating the complexity of power = supply=20 design, especially switching-mode power supplies.

The perception that power supply design is trivial = dates=20 from the era when power supplies were usually a transformer-rectifier = set and a=20 dissipative series or shunt regulator. The perception was probably not = true=20 then and is certainly not true now -- now that most power supplies are = of the=20 switching-mode variety.

Because this perception led to inexperienced = designers being=20 assigned to the power supply, power supply related problems became = severe in=20 the 1970-80 decade. The Navy claimed that if power supplies just met = their=20 specifications there would be a 20% improvement in fleet readiness. = They=20 dictated that the procurement agency had to investigate if less than = 10% of a=20 project development budget was devoted to the power supply development = -- an=20 unprecedented action. Projects on an 18 month schedule sat on the = shipping dock=20 for another 18 months waiting for a working power supply.

Switching-mode power supplies are:

non-linear circuits =20
difficult = to =20 stabilize, often having real and complex poles in the right-half = plane =20
subject to chaos =20
magnetically complex -- requiring consideration of proximity and = eddy =20 current effects and often requiring planar construction =20
analog, digital, and power circuits, requiring knowledge of each = discipline =20
sources of electromagnetic interference =20
etc.

They are not a trivial design task.

Relevance

=20

The problem and consequences of underestimating the=20 complexity of power supply design and application is relevant to almost = all=20 electronic systems, since virtually every electronic system requires a = power=20 supply.

Solvability

=20

Awareness is the first step in solving any problem. = At first=20 the problem was buried as being too embarrassing to talk about (after = all, how=20 could something as simple as a power supply cause so many problems), = but a=20 Navy/Industry team brought the problems solidly into the open with the = 1982=20 publication of the first edition of Navy Power Supply Reliability - = Design=20 and Manufacturing Guidelines [1].=20 With over 18,000 hard copies distributed, most of which were further = copied,=20 along with substantial press coverage -- both supportive and critical = -- the=20 lack-of-awareness issue was resolved and industry could get on with = solving the=20 problem. Although somewhat dated, reading this landmark publication = should be=20 mandatory for any system or power supply designer and their management. = The=20 Best Manufacturing Practices Center of Excellence website has the = second=20 edition [2] (the = most=20 current) available on = the=20 web.

In essence the problem was and is solved by taking power = supplies=20 seriously, getting someone on the program from the beginning = who=20 understands their applications and pitfalls, getting manufacturing or = the=20 vendor on board early, allocating sufficient resources -- including = schedule,=20 knowledgeable personnel, and software and hardware tools, and by = serious use of=20 design reviews and manufacturing screens.

Solution

=20

The solution to underestimating the complexity of power = supplies is=20 awareness. A short history lesson provides an = introduction.

Throughout the 1960's and 1970's, many program = managers=20 learned the hard way (though experience) that power supplies were the=20 critical-path item on their program, causing embarrassing schedule = slips and=20 cost overruns on an otherwise successful program. Unfortunately, people = talk=20 about their successes and not their failures, especially when it is = something=20 as embarrassingly simple as a power supply. Hence, the trouble with = power=20 supplies was one of the best kept secrets of these decades.

But the Navy was keeping score and by 1979 had=20 determined

"Failure data suggests that fleet readiness could be improved as = much as =20 20% if electronic power supplies simply met their specified mean times = between=20 failures in service use." P4855-1, page 1

To put it in stronger terms, if only power supplies = met=20 their allocated field reliability, it would be equivalent to adding a = couple of=20 aircraft carriers to the fleet -- complete with a full compliment of = aircraft,=20 munitions, and crew -- at no extra cost to the taxpayer.

The Navy then went on to contact 170 Navy power supply design = activities, convened a panel of experts to work on solutions, = and=20 published the solutions in Navy Power Supply Reliability - Design = and=20 Manufacturing Guidelines [1], arguably the = widest read=20 document ever written on power supplies.

Once the problems were aired, contractors admitted = they had=20 the same problems on Air Force, Army, NASA, and commercial programs as = they had=20 on Navy programs. The problems with power supplies were widely = discussed and=20 solutions found and implemented. By 1989, the Navy could claim that = power=20 supplies were no longer the number one failure item in the fleet. But = before=20 this occurred, some remarkable things happened.

Prior to the release, and in anticipation of = Navy Power=20 Supply Reliability - Design and Manufacturing Guidelines [1],=20 the Navy released NAVMAT NOTICE 3080 [3],=20 which, among other things, required a Navy Program Manager's procuring = activity=20 to investigate and justify any program that did not meet certain = criteria.=20 These criteria included:

"The development schedule shall allot at least 18 months from power = supply =20 specification release to production start, including completion of =20 qualification and reliability growth screening."
"At least 10% of the total electronic system development cost, = excluding =20 software, shall be budgeted to the power supply = function."

These are remarkable requirements. To the best of my = knowledge no program manager had ever been constrained before or since = to=20 allocate a specific percentage of their resources to a specific = function.

There were probably few program managers who were = allocating=20 this much to power supplies or who wanted to do so -- they thought they = had=20 better use for their money.

To further make program managers aware, if the = schedule was=20 less than 12 months or the funding less than 5%, special approval had = to be=20 obtained from the Deputy Chief of Navy Material, Reliability, = Maintainability,=20 and Quality Assurance (then Mr. W. J. Willoughby, Jr.). And this was no = rubber=20 stamp approval. The program manager had to explain to people = knowledgeable of=20 the history of past problems why their program was different. If he = could do=20 that, no problem, but he had to do his homework.

The bottom line is that this established awareness at the = highest=20 program management level.

The above illustrates the steps the Navy took in = making=20 program managers and subordinates aware of serious power supply = problems. After=20 making them aware, a plan was provided, Navy Power Supply = Reliability -=20 Design and Manufacturing Guidelines [1],=20 that was designed to reduce the risk of the problem to acceptable = levels (power=20 supplies meeting their allocated reliability in the field).

History often repeats itself when the members of the = current=20 generation do not have knowledge of the problems of past generations. = The=20 1980's decade saw the recognition and solution of a serious reliability = problem=20 with power supplies. Power supplies are still a serious design = and=20 application problem. Not being aware of the potential problems = and the=20 ways to reduce risk leave any present program manager susceptible to = their=20 worst nightmares - schedule slips, cost overruns, and poor field=20 reliability.

To keep history from repeating itself on your watch = there=20 may still be no better way than getting a supply of Navy=20 Power Supply Reliability - Design and Manufacturing Guidelines and = making=20 sure everyone associated with the project from program manager to = manufacturing=20 foreman has a copy on their desk. They may not read it -- it was = written by a=20 committee and is dry reading -- but try at least get them to invest the = six to=20 ten minutes it takes to read the executive summary and skim the = document. Make=20 sure you read it also.

Disclosure: I was on the committee that wrote the = first=20 edition -- but that does not mean I agree with everything. = Never-the-less, it=20 is one of the better committee efforts concerning power supplies. When = you see=20 something in the document that has been overtaken by events or is wrong = from=20 your viewpoint, don't reject the whole document. Most of it is still=20 relevant.

Personal Anecdote

=20

There are lessons to be learned from two programs in = two=20 separate large corporations that got into problems by underestimating = the=20 complexity of power supplies. I learned the details about the first by = talking=20 to two engineers who were involved. I had personal experience with the=20 second.

The first resulted in an 18 month schedule slip on a = system=20 schedule that was 18 months from go-ahead to completion of = qualification=20 testing. At the beginning of the program a buy decision was made and = the power=20 supply development was awarded to the lowest bidder. When it became = apparent=20 the design activity could not deliver, the contract was pulled and = awarded to=20 the next lowest bidder. When it became apparent that they also could = not=20 deliver, the power supply was pulled inside. However, instead of = drawing on a=20 group of designers who had strong experience in switching-mode power = supply=20 design, the company gave it to some excellent designers with strong = backgrounds=20 in communications and analog design (one of the engineers I talked = with). When=20 they ran into trouble, the corporation finally tapped their power = supply design=20 experts (the other engineer I talked with) and the job was completed. = The rest=20 of the system was completed on the original schedule up to the start of = qualification testing and waited on the dock for qualification testing = for 18=20 months before the power supplies were available. The final = result was=20 an 18 month schedule slip for underestimating the complexity of power = supply=20 design.

The second started the same, but ended up on = schedule. Again=20 a buy decision was made and the power supply development was awarded to = the=20 lowest bidder. Again, when it became apparent the design activity could = not=20 deliver, the contract was pulled. But this time work was pulled inside = to give=20 more management control over results. Also, the company informed their=20 government customer of the problem and the government offered their = expertise.=20 That is how I got involved. The company found and gave the project to = the best=20 qualified power supply engineer in the corporation, told him the = schedule was=20 sacred and to do everything possible to keep the system on schedule = using the=20 full resources of the corporation. He and his management first selected = the=20 best corporate facility in the country to produce the power supply, = which was=20 on the west coast, and assembled a team to do the job, mostly from east = coast=20 engineers. They were placed in a hotel near the west coast plant and = worked=20 between 60 and 80 hours a week. Two of the engineers were control = experts and=20 had the best computer tools available in the corporation to work the = stability=20 issues. Another was an electromagnetic interference expert who made = sure=20 everything in this area was done right. Magnetic designers, layout and=20 packaging experts, and manufacturing experts rounded out the team. Most = of the=20 system tests were completed using laboratory power supplies, but when=20 qualification testing started, the final power supplies were there. = The=20 final result was a system delivered on time.

What are the lessons learned? The problems started = on both=20 of these programs by awarding the power supply contract to the lowest = bidder.=20 This invites buy-in by unqualified design activities. The Navy = recognized this=20 problem in Navy Power Supply Reliability - Design and Manufacturing = Guidelines and other directives which state:

"A fundamental criterion in power supply source selection should be = the =20 bidder's stated, perceived, and demonstrated ability to comply with = these =20 guidelines." [Navy Power Supply Reliability - Design and Manufacturing = =20 Guidelines]

Lesson #1. The lowest bidder is not the primary consideration = for=20 contract selection. The primary consideration is that the = bidder must=20 be perceived as being able to do the job. The bidder usually=20 establishes this by the experience of their designers and manufacturing = personnel in the design and manufacture of similar power supplies. Also = their=20 track record in accurate bidding of the cost and schedule on similar = power=20 supplies. I once asked a successful power supply buyer what the = three=20 most important things he would recommend to other buyers. His = list=20 was:

Vendor Selection =20
Vendor Selection =20
Vendor Selection

Lesson #2. When in trouble ask yourself how you can do things = better, rather than repeat past mistakes. In the first = example, they=20 just repeated their first mistake -- giving the contract to the lowest, = and=20 unqualified, bidder -- then when they brought it in-house, continued to = give it=20 to an unqualified bidder, an in-house design activity with no power = supply=20 experience.

Lesson #3. Ask for help when you first get into = trouble and=20 then apply a big enough hammer to get the job done. Large corporations = have=20 many shortcomings, but one advantage they have are formidable resources = they=20 can apply to a problem. In the first example, when they finally pulled = the=20 contract in house, they gave it to the best they had locally -- who had = little=20 to no power supply experience (after all, power supply design should be = no=20 problem for experienced analog and communication designers) instead of = tapping=20 the corporate power supply experience. Eventually they tapped their = in-house=20 expertise and finally got an excellent power supply -- at an extremely = high=20 cost in terms of schedule slip and the associated cost overruns. In the = second=20 example, when program management perceived a schedule problem, they not = only=20 asked that the full resources of the corporation be applied, but also = asked for=20 any resources their customer, the government, could offer. By asking = for help,=20 and recognizing what was needed to get the job done, it got done.

This failure to ask for help is getting worse = instead of=20 better in today's environment of product teams. Teams and their = management=20 pride themselves on their ability to get the job done. But often they = get in=20 over their head because a generalist team member thinks they can do the = job of=20 a specialist. When they finally realize they can't, the program = schedule is=20 usually at risk.

On the Web

=20

Navy = Power=20 Supply Reliability - Design and Manufacturing Guidelines

This is a digital "how-to book" on the Best=20 Manufacturing Practices Center of Excellence website. The website = states:=20 "The main goal of BMP is=20 to increase the quality, reliability, and maintainability of goods = produced by=20 American firms. The primary steps toward this goal are simple: identify = best=20 practices, document them, and then encourages industry, government, and = academia to share information about them." There is much more of = interest on=20 this website.

References

=20

[1] NAVMAT P-4855-1 Navy = Power=20 Supply Reliability - Design and Manufacturing Guidelines, December = 1982

First edition, replaced by the second edition = below.

[2] NAVMAT P-4855-1A (NAVSO = P-3641), =20 Navy Power Supply Reliability, Design and Manufacturing Guidelines, = Revised=20 Including High Voltage, January 1989 (Stock No. = 0518-LP-204-4800).

Second edition, available = on the World-Wide Web.

[3] NAVMAT NOTICE 3080, Power = Supply=20 Reliability, 16 April 1982.

The Navy document, followed by NAVMAT=20 INSTRUCTIONS, that put the teeth into NAVMAT P-4855-1 Navy = Power Supply=20 Reliability - Design and Manufacturing Guidelines, December = 1982

Legal=20 Statement

Do not use this information for design without=20 independent verification of the information.