Bryan Bonnici's Portfolio

A new year, a new site — Welcome!

There’s no such thing as a million dollar idea.

Only a million dollar execution.

A twelve year old could come up with twitter. The idea of publicly SMSing people isn’t really ‘out there’. What a twelve year old couldn’t do is make it into a global phenomenon. I’ve met Jack Dorsey, its founder. Shook his hand. Incredible guy. He walked into a business pitch, tattooed and pierced, and charged one of the investors’ credit cards as a proof of concept for his payment processor Square.  Enviable balls. That’s the sort of person who takes an idea into a global phenomenon. One day I hope to be like that.

A year ago I pitched an idea for some software and got first place, a wad of cash and an iPad. Actually turning it into something viable is something I’m still faking-till-I-make-it. They don’t teach you anything about business in an Engineering undergrad. And only the most superficial glance at law. Running a business solo requires you to be the engineer, accountant, lawyer, purchasing manager, marketing, and a gamut of other positions. It’s overwhelming really. You need to have the acumen of at least 5 or 6 different professions. I’m versatile, but damn. You also need to be personable and deft at convincing others to hand you money. They also don’t teach you that.

Not too long ago I had a chat with someone who wanted to work in the Video Games industry. He had no proficiency in programming, art, or any sort of technical skill that goes into the production of a game beyond QA. He wanted to be the ‘ideas guy’. He’s not the first person like this I’ve run into, but he was by far the most stubborn about it. People like this seem to think that the producers et al are completely incapable of ideas themselves and need to be told what to do. It’s stupid because if anyone knows how feasable an idea is it’s the person with advanced knowledge of that skill. There’s no tangible value to just an idea and he will never be hired because every company is looking for people solve problems (re: how to implement an idea), not people who come up with them.

Never mind that the games industry is crazy and the reason all those offices are stocked with snacks and cots is because they expect you to live there. Those bean bag chairs don’t look so cool when they’re the closest thing to family you’ve seen in a week. I’ve pulled 16 hour coding days before and they’re soul crushing.

I’ve seen great ideas fail and terrible ideas succeed. The successful ones usually have a wad of cash and great marketing behind them. The only thing consistent is that in either, nobody was ever hired as an ‘ideas guy’. In fact, I’d say marketing that idea is the most important aspect, because it’s marketing that triggers that desire to buy. If you can’t encourage (lots of) people to want your product then you’re DOA. Which is what’s happening with Nintendo. 70% sales forecast reduction in their flagship product. Holey moley! I’ve never seen that before. They went from distant first to distant last. I hope that guy isn’t trying to get a job there. Or maybe he’s their only hope.

I hope Ford Motor Company likes my idea for flying laser cars

High performance cameras are available to consumers at very accessible costs nowadays. So it’s not uncommon I run into someone with a DSLR that has no idea how to use it. The telltale sign is firing the flash in completely inappropriate situations. I see people do it on their cellphones too, but it really pangs on $1000+ cameras. It happened the other day. It peeves me so much I’m blogging about it.

There’s two situations that I can think of where the builtin flash makes sense:

1) You’re photographing a very small object very close. Let’s use an analogy that I’m thinking up as I go: Think of light like wind. You want as much wind to hit your subject as possible. So it doesn’t make sense to try and blow air through a straw at something halfway across a huge room. You need a huge blower, which would be equivalent to a hot-shoe flash or softbox/umbrella. The key thing to nail down is, the farther away your subject is, the less light hits it. In fact, the falloff squares in proportion to the distance. So if you double your separation to the subject, your subject gets only a quarter of the light it did before. If you instead move 3 times as far, your subject only gets one -ninth the light (a loss of almost 90%!). Four times, and your subject gets one-sixteenth the light as it did before. Engineers take note, I just applied Field Theory to something artsy. This is why it doesn’t make sense to fire the puny flash to illuminate a lecturer from the back of an auditorium. Please stop.

And even when I’m close to my subject, I don’t use it. it’s a very harsh, sharp light. There’s a reason why pros use indirect and diffuse light – because firing a flash right at a person will result in parts that are totally blown out while others are very harshly shadowed. I’ve had some really unflattering photos where I look like a cadaver because someone fired their flash inappropriately.

2) Photographing Snow. So you take your camera out to capture that “It’s snowing!” moment and you walk away disappointed, wondering why it didn’t come out so well. Snow is small and flakes facing the camera are typically dark. Firing the flash will typically provide enough supplemental light to make it pop. The snow will help you out in two ways – the flakes will be fairly close to the flash and are very reflective. And in the daytime, there will be enough ambient light to disguise the fact that you fired the flash. Here’s an example:

A Streetcar's Last Christmas

…but please make sure to protect your camera from the elements! Plastic shrouds are much cheaper than repairing water damaged DSLRs!

3) Documentation. Okay so there’s a third. This is for when you don’t care about the artistic merit of the photo. Something happened somewhere dark that you need to take a picture of.

Take the twenty minutes to learn how to use the non-auto modes of your camera. You spent a lot of money on it so don’t cheat yourself. They’re not hard at all to learn. I always shoot in Aperture (A or Av) or Manual (M) modes. In Aperture, you control the size of the aperture (which, for the novices, controls how much the surrounding background or foreground objects are in focus) and the camera picks an exposure time to properly expose the photo. When I disagree with the camera’s choice of exposure time, I switch to Manual and set it myself. Keep bumping the ISO up until you can take pictures faster than about 1/120 of a second and you don’t have to worry about the flash. It’s important not to shoot slower than this or you get motion blur of your subject; shooting slower than your focal length also adds your camera shake into the mix and r̛̩̮̩e̳̪͔̼̺ͅs̹u͓̫l̨̦̠̭̺͇ts̗̥͈̰̜̙̮ ̥̭i͏̮͚͙̹͎͎͍n̬̤ͅ ̩a̸̗ r͇̻ͅu͓͇̪̕i̵̜̪̖̠n̠͎̖e̘̲d̬͇̳͓̠̳͠ ͎͚̮̭͡p̘͇̳̙̲͕h̹͎̯͙o͚̪͙t̠̭̩͢o̧͍̟.

Image: Emscan

Life is not easy for the PCB designer. Everyone wants more, faster, better out of their electronics. Heck, so do I. It took me over 10 minutes to dump all my photo RAWs off my SD card so I could edit them in Photoshop. 10 minutes of just sitting there and watching a file dialog.

Admittedly I’m using USB 2.0 and Class 10 SD cards which are old news. Engineers have gotten together to develop USB 3.0 and Ultra High Speed SD. These buses can transfer data at over 1Gb/s. And there are interfaces even faster still. There are a lot of challenges when developing things that work that fast. The idea of “ideal” goes out the window and the effect of parasitics, harmonics, propagation delays, and a whole bunch of other fun stuff is unavoidable. The novice board designer can easily bang a USB 1.1 port on their board, 2.0 with a little extra consideration, but 3.0? They’re crazy.

So what’s an engineer to do when they want to push materials to the limit? Or they want to design something that follows Electromagnetic Compliance (EMC) regulations? I attended a talk by Elya B Joffe, an EMC board designer. He started off with an inspirational quote: “No two EMC Engineers agree unless they’re drunk.” The intrepid board designer must pay special scrutiny. They may be bombarded with information that’s not necessarily supported by facts. So not only are there serious design challenges, there’s not necessarily a clear cut answer.

In his case, the 3000th harmonic of a power supply still generated enough interference to jam a sensitive detector meant for aviation. The power supply even complied with MILSPEC guidelines. He had to move it elsewhere in the plane. This is all to point out it’s possible to enter ultra high frequencies even at low clock rates because of very high transient speeds; you might run into this stuff where you don’t even expect it.

At UHF, everything wears a mask. Inductors become capacitors, because the tightly packed coils have a parallel plate effect. To confuse things further, Capacitors become inductors, because parasitic inductance is governed by package size of the component. Same with resistors.  To minimize parasitic inductances, package sizes like 0402 (4 thousanths by 2 thousandths of an inch) or even 01005 (1 thousandth by half a thousandth) have been developed. 0402 is chicken feed. 01005 is sand.

That DC power supply? It’s not DC. The high frequency switching of each IC attached to that bus causes large inrush currents which is noise on your line. Remember, everything takes current – never think of it as a voltage supply! In fact, supply line noise is a massive problem for engineers as IC technologies keep pushing frequency up and line voltage down. If a chip has a 1 volt Vcc and draws 150mA, and the power supply has to stay within 5% of Vcc, then the maximum impedance can be only 330µΩ. The best short circuit possible on copper is about 40µΩ!

Worse off, if your traces are long, you have to consider your traces as transmission lines. There are various formulas, but a safe bet is if 6*(Propagation Time)*(Line Inductance) is greater than the rise time, handle it as a transmission line. This means things like reflection and ringing when combined with gate capacitance. Expect ringing on the output, or worse, voltage spikes bad enough to destroy the transistor. Start saving for that Agilent Momentum license because this isn’t something calculated on pen and paper.

Properly terminating transmission lines is therefore necessary in good UHF board design. This can be as simple as a terminating resistor (equal to the line impedance Z0) to AC GND, pulling whichever direction the output is weaker. This drains battery so you can also stick a capacitor in there, but this can’t be used on clock lines and you must pick an RC at least twice the delay time or three times the rise time.

At UHF, signals propagate more on the dielectric than on the trace, so your decoupling/bypass capacitors need not even be connected physically to the device, so long as they are connected to the planes nearby. In fact, it’s better to connect it between the planes since the planes have lower resistance than any trace. As for values, there’s two main sides: those who spread (1/10/100nF) and those who same (100/100/100nF). I lean towards making all caps the same, because if they’re the same packages, you get 3 different resonances; same caps results in only one resonance and an overall lower inductance. This also has the side effect of reducing the number of parts you have to use to populate your board. Note, that the dielectric constant for FR4 (4-4.7) is only true at several MHz, not GHz. You can probably forget it as a board material at high speeds.

When designing UHF boards, use these principles:

1)      Keep GND as low impedance as possible
2)      Keep VCC as low impedance as possible
3)      Keep a low AC impedance between power and return conductors
4)      Provide each logic device its own transient current source
5)      Never ever ever ever route a signal across a plane gap. That’s a slot antenna and an automatic 20dB more EM radiation. Move the signal to a different layer!
6)      Keep board layers symmetrical

1 & 2 mean dedicated layers; 3 and 4 mean capacitors.

In addition:

Never leave a copper pour floating.
No copper pours on signal layers.
Route your most sensitive signals first.
Keep current loops small to minimize inductances.

I really enjoy taking photos of construction and renovation. Today’s was a little solemn, however, because I visited the mall where I learned to walk – Erin Mills Town Centre. The iconic postmodern clock at that mall is being torn down and replaced with a giant blue ball. I’m not kidding. I’m not getting any overall artistic vision here.

No Mo' PoMo 4

I posted some last photos of the clock, which has already stopped ticking, on my Flickr. Demolition begins early February.

I picked up an old Futaba VFD from Active Surplus. VFDs still manage to out-cool their LCD counterparts, and for $2 I couldn’t say no. I’ve never dealt with VFD before, but I know that they’re functionally similar to vacuum tubes.

It looks like a custom part from an alarm clock. Custom parts usually cause some trepidation because that means lots of figuring out pinouts and data protocols.This thing has no driver chip on it so this means just figuring out pinout. Luckily Futaba was smart and made it obvious which pins do what. Looking at the front, it’s immediately obvious which pins are filament (the outer two on both sides), grid  (the ones that touch the border each rectangularish segment) and data (the rest). And then on the back, they are kindenough to provide a diagram for each data pin!