Where can Analog Consultants help?
by Joerg Schulze-Clewing
Difficult Analog Designs
It isn't possible to furnish all expertise in house at all times. Sometimes it doesn't even make sense to have a full time analog engineer on staff. Whatever the scenario, there often comes a point when a design becomes difficult. Going it alone at that point might lead to huge problems later, be that in production, in the field, or simply a large cost overrun. I have seen boards that unnecessarily contained numerous precision chips totaling over $1000 just in parts and they still would not work reliably. This will often require a redesign from scratch, meaning that all the engineering hours that had gone into it were basically wasted.
Other designs are just partially difficult. This could be an RF section, something that needs to provide incredible precision or maybe a disposable portion that must be realized at rock bottom cost. In that case the consultant can handle the "nasty" parts while the other engineers develop the rest. That has worked very well in the past. It also provides an opportunity for the client's engineers to learn how to handle this type of challenge next time.
Analog consultants are not FPGA designers, so what should they be doing here? A lot. As soon as clock frequencies exceed 100 MHz or rise and fall times reach picosecond levels things become analog rather quickly. A minor oversight on the clock distribution, phase jitter, cross talk, unwanted transmission line behavior, all this can have effects that can drive the digital designer up the wall. The simulations show that everything should work fine but the design behaves outright weird. That can cost many work hours and throw the time schedule a real curve.
Analog consultants use tools such as oscilloscopes, analyzers and meters in different ways. They look for those subtle yet revealing signs of trouble. Bus contentions, transient pulses, runt pulses, brief bits of data that weren't supposed to be there. Very few digital designers master this art. When the going gets tough analog consultants pull out techniques that others would regard as a dowsing rod approach. Many times I have used a communications receiver, headphones and a suitcase of near field probes to find problems on digital boards. You would be surprised what kinds of issues are being unearthed this way. Weeks of diagnostic attempts by digital designers and software engineers can be spared.
System Level Problems
Mostly our tasks here are related to pre-compliance improvements, to achieve a better electromagnetic compatibility (EMC). The typical problems that I find are grounding issues, suboptimal cabling paths, incompatible mounting hardware or power supplies that do not hold what the manufacturer had promised. The ideal scenario would be to have an analog consultant on board during the initial design phase.
The most amazing results are the ones nobody expects. Often when I finish up my EMC pre-compliance work the system is tested to see whether everything still works as before. Suddenly the crew realizes that many problems that were believed to be software or firmware related just vanished. Days or weeks had been spent searching for a bug that simply wasn't there but had in reality been caused by noise coupling within the system.
Engineers often dread this topic and see it as a necessary evil at best. Analog consultants have dealt with it many times and know the tricks of the trade. Legal ones, that is. It often surprises me when, for example, engineers struggle hard to get certain clock harmonics under the limits to pass EMC class B. When I suggest frequency hopping they look at me in disbelief but when they see that the spectrum spread has knocked 5 or 10 decibels off their toughest harmonics they start to believe in it.
A similar situation is electrical safety. All engineers worth their salt know how to design a product that is safe. However, when requirements become very strict this may pose a tremendous challenge. An example that I frequently encounter in medical designs is when the patient interface must be defibrillator proof. That means it basically has to withstand a 5000 Volt surge into the terminals without any kind of dielectric breakdown and with next to nothing in current. In most cases the system must continue to function to specifications within a reasonable time after such a jolt. It can be done but it is not something that is taught at universities.
An analog consultant who is familiar with high volume production can take a holistic look at things. Being a very good listener is key in that role. I have found on occasion that the communication between Engineering and Production is, well, not always stellar. For some reason an independent party is often able to detect much more than could be found in house. I vividly remember a case where a very large SMT board would fail almost 100% after reflow soldering. We quickly determined that warping was the cause and a short trip to a lumberyard for some nice two by four solved that for good. The billed time for that assignment was under three hours, much less than the cost of a single circuit board.
One reason why a consultant may be a bit more successful here is the exposure to so many other situations. While never saying it, a consultant often instantly thinks "been there before" and can come up with a solution that has worked well in the past in a totally different scenario. Other times the consultant can suggest solutions that nobody ever used in the client's business. Often I heard comments such as "you really think that anyone makes such a machine?". Yes, they do, they just were never intended to be used in the manufacturing of electronics.
Analog consultants who have a long history of designs under the belt have learned how to be frugal. They have designed to very stringent cost targets, turning around every little transistor to see if it could be spared. Being able to design down to transistor level is absolutely paramount. Unfortunately that is not the case anymore for every analog engineer. Only when our mind can imagine every possible solution from transistor level all the way to fancy chips can we achieve an optimum in cost versus performance.
Cost thinking goes far beyond the design itself. A decade or more of experience is needed to be familiar enough with what is possible in production and how much it will cost. There has to be a Dollar calculator running in a corner of the brain at all times. Just as an exercise, but one that has to be done without offense, have you asked a group of engineers during a design review what their solution will cost in production? It doesn't have to be an exact amount but engineers should know a good ballpark figure and what the "heavy hitters" are in their designs. Seasoned analog consultants always do.
When contemplating an acquisition it is important to know what's really in the bag. This holds true whether a whole corporation is targeted or just the rights to a certain product design. The own engineers might be a bit biased, either overly optimistic or in the other extreme they could react with a slight hostility towards "being merged" with another group. That is where an independent party is of very high value. Does this or that project really have a chance? Are there any compliance matters that might turn into show stoppers?
I have dealt with due diligence, twice on the side of the acquiring party and once on the other side. Independent parties were always involved and they all provided very honest opinions. After all, we bestow plenty of money and other resources on legal counsel. Shouldn't the technical side be covered equally well?
If you think that analog consultants might be the answer to a pressing problem it may be best to discuss it: