Why I Hate Power Chip CompaniesJuly 23, 2011
I have given this talk many times, direct to various manufacturers, reps, and distributors. So I finally decided to write it in blog form. Maybe this will streamline my sales guy denial process…
As some of you know, I’m an electrical engineer. I design circuit boards that use computer chips. Almost every board I design has some sort of power chip on it. By “power chip”, I am talking about some sort of linear or switching voltage regulator. A recent board I did has 9 (!) voltage regulators. They generate the these power rails: +50v, +48v, +15v, +6v, +3.3v, +2.5v, +1.8v, +1.2v, and -15v.
There are many companies that make these types of chips. Here’s a list of some, but not all, of them: Linear Tech, Texas Instruments, National Semiconductor (recently bought by T.I.), Analog Devices, Emerson, Fairchild, Infineon, Maxim, Microsemi, Microchip, On Semi, Seiko, Sharp, ST Micro, Vishay, Alpha & Omega Semi, International Rectifier, Lineage Power, Micrel, Nuvoton, Power One, Silicon Labs, and many more.
To make matters worse, each of these companies make many different power chips. Texas Instruments alone makes over 1,600 different power regulator chips, not including their MOSFET’s, voltage references, and power control chips.
These companies make what amounts to many versions of the same thing: they convert one voltage to another. There are small differences, of course. This one might have 3% more efficiency. That one might have a soft-start function. This one needs less external parts. That one costs less. But at the high level view of them, the parts are basically identical.
So, how can someone decide on the part to use? You simulate it. You design the circuit and run a simulation on that design. If the simulation looks good, you can either hack up an evaluation circuit or just put it into your prototype PCB. This might sound simple, but simulating the circuit is non-trivial. Each manufacturer has a different way to simulate the circuit, and some are better than others. One manufacture allows you to simulate arbitrary circuits with their chips. But most manufacturers will only do “fixed topology simulations”– and by that I mean they provide the circuit and allow you to swap resistor values and such but they don’t allow you to add or remove components and require connections.
So, doing a simulation of a power circuit can take hours to a day or two. For sake of argument, let’s say it takes 4 hours– half a day. And let’s say that each manufacturer’s rep narrows my choices down to 2 chips per manufacturer. And I look at chips from each of the 22 manufacturers listed before. For each of the 9 regulators on my board. That works out to be 39.6 weeks of my time evaluating power chips. Oh, did I mention that I have 2 weeks in my schedule to do this?
Clearly I cannot look at all 22 different manufactures offerings. So here is what I do. First, I try to reuse designs as much as possible (but within reason). Next, I choose the company with the best simulator. That would be Linear Technologies with their LTSpice. Then I choose the top 2 companies in this field. That would be Texas Instruments and National Semiconductor. (Since National just got bought by TI I’ll have to rethink who’s the top 2.) That gives me three manufacturers to look at.
Even then I would need 5+ weeks to properly evaluate things for my 9 power rails. So at this point I have to rely on the Field Applications Engineers (FAE’s) for support. In some cases, they do the first part of the simulations for me. In others, they further whittle down my options so I don’t have to evaluate quite so many chips. I also need a reasonable amount of attention from the FAE’s to get the simulations right, to supply some eval boards, and to debug the circuits once the prototypes are built.
In the end, somehow, what would have been a 39+ week endeavor gets done in 2 weeks.
This brings me to the part I really hate: Just because I’ve limited my choices to 3 manufactures, that doesn’t mean that the other 19+ have stopped calling on me. I have to tell these folks, nicely, that they just don’t cut it. So if you’re one of those 19+ companies, here is the gauntlet that you have to pass through to get my business:
1. Do you have a simulator that works with the part that you’re suggesting to me? Does it work with the topology that I want to use?
2. Do you have an FAE that is good with power supplies and is local to my office? And by “good”, I mean someone who is dedicated to doing only power chips and knows which end of the soldering iron to hold.
3. What can you do to limit the amount of time I have to spend evaluating this part of yours?
4. What advantages does this part and your company have over the three manufacturers that I’m already considering? How can I verify that without spending too much time on the issue?
Sadly, even T.I. and National cannot negotiate this gauntlet unscathed– but they do better than most. Linear Tech almost always flies through. The other companies can almost never get through.
In the end, I have to turn away what might be a good solution. The realities of my work and the semiconductor business demand that I do. Oh well.