原创 Design engineers vs component police

Sometimes a designer himself could decide on the particular component needed without too much opposition from the Component Engineering group—often because there was no such group. But, any corporation that includes a department of component police makes a designer's job that much more difficult. Here's why.

These people have the justifiable responsibility of making sure the corporate inventory doesn't get overwhelmed by several variations of the same part, when in reality, only one type is needed. They also want to ensure that the designer's choices will be available for the life of the product, are not over-priced, and have multiple sources. Sometimes, though, they go a bit overboard in their zeal and may not always understand the components or the reasons for a choice.

Take a simple resistor—one company's NPI (New Product Introduction) group insisted that there were really only a handful of necessary resistor values. Its component engineer (a purely digital type) was looking at those resistors as logic pull-up or pull-down. I had to explain that you can't always make a voltage divider out of 1kΩ and 10kΩ resistors, and yes, there was a good reason for the required 1 per cent tolerance. Same for capacitors: Sometimes the company standard Y5P dielectric 0.1µF logic decoupling cap is not suitable for analogue circuits.

I remember an incident where a colleague had designed in an octal latch—I think it was a 74LS373—and on going over the timing analysis, I realised that the 20nsec data hold time would be a problem. The component cop continued to allow usage of the LS family because it was so ubiquitous and unlikely to go obsolete, but balked when I told him I wanted to use the ALS family in this particular function because of its much shorter hold time.

He was of the opinion that ALS was a passing fad, so I then told him I was going to use the F373 instead. He wanted to know why I wanted to use the F family, I told him simply "Because you won't let me use ALS." He finally agreed that maybe the LS device was not suitable and agreed to the F version. One of my engineering colleagues actually tried to specify IC sockets on his BOM (bill of materials) to prevent the cops from squelching his design choices, figuring that once the PCB was a done deal, there wasn't much that they could do. Of course, he wasn't successful in this attempt to put one over on them.

Preferred vendors can be a problem. The purchasing people have their favourites, and the engineer may not always get the desired part. I had requested some optical patch cords from a certain vendor; purchasing went with a different vendor who had supplied copper cables in the past. One of the fibre cables still had the unpolished end fibre sticking out of the connector—so much for quality control.

A similar quality problem occurred when I specified a certain brand of twisted-pair cable for a production test fixture. My own test fixture worked nicely with the brand I chose, but production was swayed by purchasing to go with its preferred brand. There was a huge variation in attenuation between pairs in the same cable. Let the losses begin.

Sometimes, you can't get parts with multiple sources, or even a second source. Trying to explain to the components group that a certain brand of optical transmitter and receiver was the only one available at a reasonable price did not stop them from whining—until they did some of their own research and determined that there really was no choice.

I did see their point, especially when that same vendor made a process change that resulted in a horrendous optical overshoot in the transmitter. I had to come up with a fix that required a large value trimmer capacitor, and because the value I needed was a year lead-time (after I bought about 400 of them myself), purchasing wouldn't allow it. We ended up using a varactor diode and a trimpot instead.

The most difficult component hassle was with a simple 32.768kHz crystal and the associated CD4060B CMOS oscillator/timer IC. No way were the component cops going to allow something as obsolete as the CMOS 4000 series in a new design! It took a lot of explaining that, old or not, it was not about to go obsolete and that the chip had recently been made available in a surface mount package. I was certainly not about to design in a separate crystal oscillator and counter chain.

The crystal itself was another problem. What's so hard about procuring a 32.768kHz tuning-fork crystal that is made in the billions? None, except for the fact that the corporate-preferred vendor had none in stock and there would be a twelve month lead time. I sent a "nastygram" to the entire department explaining that I could not sit around for a year and that the component police needed to find and approve a crystal now. They did.

Have you ever been a component engineer? Let's hear your side of the story.

Glen Chenier
Engineer