The Analog Monster; G5 chirping & Arcade Monitors

John Siracusa wrote an OK article about the problems with Analog circuitry in the digital world.

Only OK? Yeah — it isn’t exactly accurate.

The transformer, in particular, was quite a noise-maker in my 150-in-one kit. Perhaps not surprisingly, one of the many things a transformer can do is convert AC to DC and vice versa. Hmm.

Wrong. Transformers are used for converting AC from one voltage to another or for isolating one piece of circuitry from another (as long as the isolation point is AC). Transformers have a handful of other uses, but converting AC to DC is never one of them. For that, you would probably start with diodes, including the rather common configuration of 4 diodes in a rectifier.

John has since corrected his article. I figured he would; I’m sure he received about 18.2 billion email messages with a correction.

---

Now that that is out of the way, Siracusa’s triggered a few random thoughts on the subject.

John is absolutely right in that analog is really really hard. But he is dead wrong if he thinks that Apple is somehow shipping crappy analog electronics inside the machines. A more accurate assessment is that Apple — and just about every other brand name computer manufacturer — ships with the best quality components that can be had at a reasonable price and yield reasonable performance. And by performance, I’m referring to power consumption, heat generation, noise floor, efficiency, and all those other wonderful metrics used to quantify analog performance.

Analog implies highly complex behavior. And with that behavior, you get lots of different variables that can be tweaked to achieve any given goal.

That Apple hasn’t solved this problem in three years is indicative of how hard analog is. From John’s reply to this post:

I simply find it unacceptable for a PSU to make noise from its “non-moving” parts (i.e., not the fans). I find it hard to believe that Apple cannot find a single vendor that can meet the power/size/heat/durability/price requirements of the Power Mac G5 PSU without chirping. But if this really is the case, then Apple simply needs to increase the price it’s willing to pay for this part until it can satisfy that goal…

Therein lies the problem. Every coil within the power supply is a moving part, even if those movements are microscopically tiny. First, there is heat expansion. As the power supply heats up and cools down, the metals (and other materials) will expand and contract. This causes stress and generally “loosens” the various componentry. To compound the problem, different materials will expand/contract at different rates.

Secondly and the most likely cause of the chirp, the various coils used throughout the power supply generate varying magnetic fields as the frequency and voltage passing through changes. The end result is lots of microscopic motion through the circuits.

Frankly, I would think that three years of dealing with this problem would have yielded a solution by now, too. But I’m also not surprised that it hasn’t given the voodoo I have dealt with in maintaining power supplies over the years (mostly in arcade games).

There may also be other hard limits that the G5 is bumping up against. The Quad G5 ships with a different standard power connector (I think it is standard — hrm… can’t find the reference) than the previous generation. This is because it can draw enough current — upwards of 10amps — that the old style cords might fry. Dangerous. So, the new connector is specifically present to prevent the user from burning down their house by using some ancient, low-current, standard power cable.

Now, if the power supply were any less efficient, it would cause the power consumption of the machine to go up and — at 10 amps — much of an increase would potentially bump it out of the realm of being usable in your average house. While the previous generation of dual proc machines used the standard connector, I would bet that there was a huge impetus to stay with that connector given the increase in production costs of moving to the new connector alone.

Honestly, I don’t know where the lines or drawn or what Apple has investigated in regards to fixing this particular problem. I would be very interested to learn more.

In the audio industry, some high quality audio amplifiers maximize performance by sacrificing power efficiency in the name of signal quality. Specifically, class A amplifiers consume vast quantities of power to achieve relatively little amplification. Yet, many class A amplifiers are sold because of the high linearity of the amplification. Higher electric bills, but better sound (usually — as with anything analog, it is quite easy to screw up even the most basic of circuits in an effort to achieve market differentiation). You decide. At the “high” end (as in, you’d have to be “high” to buy this stuff ’cause it really isn’t that much better), there are even companies making amplifiers that are so inefficient that they need a special circuit or two wired directly into your breaker box to be able to safely consume the power they need. Of benefit, at least you don’t have to heat your home theatre in the winter.

The G5 chirping issue has been covered in great detail around the net. I have a chirping G5 on my office desk. Doesn’t bother me. Yet. But with all things analog, things change over time.

Reminds me of a bunch of other “analog monsters” that I have encountered over the years:

---

Both my titanium powerbook and the MacBook I’m typing on now “hiss” when I’m blasting tons of bits out the ethernet port. Seriously. I can hear the damned packets as the hit the wire or hit the port. Not so coincidentally, most network ports have an inductor — a coil of wire similar to a transformer — to act as an isolator between the port and the rest of the computer’s guts. I have no idea if that is the source of the hiss, but it wouldn’t surprise me.

---

But random noises from modern Macs (and PCs) is hardly the first time that transformer like devices have been the bane of an industry. The classic arcade game Tempest used a color vector monitor that was plagued with problems (last item). This guide gives a wonderfully detailed description of all the fun things that can go wrong in such a monitor.

When I first ran across this problem, it was originally quantified as image instability caused by a buzzing in the flyback transformer. At the time, the recommendation was to disassemble the transformer and place a drop of oil on the transformer’s core where the two sides of the core met. This would reduce vibration, thus stabilizing the picture.

Of course, that was only a temporary fix. The modern wisdom is to actually go in and replace the components with better quality or new components.

Analog components often do have a lifespan and the failure mode does not tend to be the digital working / non-working hard transition. Analog components often fail gradually, taking out other components with the primary failure points.

---

Which brings us to the Great Capacitor Caper of the Late 1990s. One of the most common capacitors used in power supplies — computer and otherwise — is the electrolytic capacitor. They have relatively large capacitive values per volume, generally have a polarity (if you hook ’em up backwards, they tend to explode) and can handle the midline voltages commonly found in power supplies (5v to 150v — sometimes more, sometimes less).

Now, electrolytic capacitors are relatively expensive to manufacturer. In particular, the electrolytic compound is a critical component of a capacitor’s performance characteristics and, as can be imagined, the formula for a particular manufacturer’s electrolytic compound is a closely guarded secret that could easily have cost tons of money to develop.

In the late 1990s, it seems that a Taiwanese company decided to liberate the formula from a competitor and started manufacturing electrolytic capacitors for well below the then market price.

Unfortunately, they either stole only some of the formula or they failed to correctly duplicate the processes involved. The end result is that millions and millions of faulty capacitors were manufactured and sold throughout the industry.

To make a bad situation much worse, the faulty capacitors failed only after many months of otherwise working service. The failure mode tends to be rather catastrophic; at the least, the top of the capacitor would swell and the capacitor would stop working. At the worst, the capacitor would explode.

Many many manufacturers used the faulty capacitors. Including early Airport base station units.

The problem was so widespread that an entire industry sprung up around repairing these bad capacitors.

Sometimes, even when you do properly test the components and design good circuits, you could still be nailed by really odd failure modes due to shoddy componentry. Fortunately, said capacitors are no longer being manufactured. I hope.

---

Sometimes, it pays to put in a few extras….

Before I joined Apple, I had a chance to chat with Jon Rubinstein at WWDC. I had a very specific question in mind.

I had one of the very first Titanium Powerbooks manufactured as I purchased the first display unit that showed up at a nearby CompUSA. I used that machine hard.

One day, I was processing around 5GB of raw log data through a very complex series of regular expressions to feed to a statistical analysis package. This was while riding a Metro North commuter train.

While waiting for the job to finish, I sort of dazed off. Pretty soon, I was being poked gently by another passenger. “Is your computer supposed to be smoking?” Sure enough, when I looked down, there was a thin stream of bluish smoke shooting out the back of the powerbook. Not a lot of smoke, but I have generally found that if an electronic component leaks any of the magic blue smoke, it immediately stops working.

But the computer kept working just fine. No sign that anything had gone wrong.

Which was exactly what I wanted to ask of one of the hardware engineers at Apple — that I happened to be introduced to Mr. Rubinstein and could ask him the question was a happy coincidence.

His response?

“Oh, it was probably just a capacitor. We put a lot of extras in, just in case…”

Heh.

This entry was posted on Sunday, March 5th, 2006 at 10:28 pm and is filed under Technology. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.