Art of Electronics, 3rd edition, corrections

Some people still do analog. It's not as popular as it used to be. But customers still need to connect to real-world signals. It's fun to interview recent EE grads who barely know how a 2-resistor voltage divider works, much less a transistor.

If you used the sot23 uP, how would you set the trip level?

We've used tens of thousands of the Bourns 3314G pots with no problems. They don't need to be programmed, don't need JTAG interfaces, none of that stuff. If you keep the required resolution at a half per cent or so, they are quick and easy to set. Around 0.1%, people can spend time tweaking.

I need to coarse-tune an LC oscillator, to get things into the pull range of a varicap. A small trim cap is ideal. I've used the Maxim digital capacitor in the past, but it's EOL now. It needed a couple of pages of code, too.

I'm still a huge analog fan. I've adopted a new mantra, the world doesn't run on 3.3 volts. But ... if you have a controller in your design, and if anything in its program needs to be touched during initial test & setup, then why not have calibration and tweaking functions under digital control?

Most of our products have an ARM, and a cal table in flash, and usually an FPGA or three. The issue here is a VCO that needs to have low phase noise and be very temperature stable. Varicaps have horrible tempcos that vary with the capacitance, so I want to keep the pull range small. So I want to tweak the gross resonant frequency to make up for the L and C tolerances. The options seem to be either a 32-step digital capacitor (like the EOL Maxim part, which I've used before) or a tiny ceramic trimmer cap.

Turning the cap with a screwdriver once, at the start of the production test, doesn't seem like a terrible burden. Some of the objections against trimpots and trim caps simply aren't rational.

We use one tiny trimpot in gain-setting applications that works fine past 1 GHz. It costs 11 cents.

The comparator was used to turn an RGB LED from blue to white to indicate w hen an OCXO reached operating temperature. Instead of tweaking a trimpot d uring production to set the threshold, it would probably have been reasonab le for a controller to watch for the current to fall to X% of its cold-star t value. I could have programmed it in-system as part of the mastering pro cess for the FPGA and USB firmware. For that matter, the FPGA itself could have done the job with a hacked single-slope ADC.

Ultimately even the trimpot was overkill because the OCXOs behaved consiste ntly enough to work with a fixed pair of resistors. But the original plan was to support a couple of OCXOs from different vendors, so the trimpot mad e more sense initially.

-- john, KE5FX

Do "both" in one step: Do the layout with a number of capacitors in parallel and laser trim(cut) links to remove appropriate capacitor, OR if amount of capacitance change is rather small, just trim away some of the copper pad of the capacitor. Digital trimming .. their way,and your way at the same time. No value drift.

Well, you'd have to have different BOMs for the two versions, so you could just put in the right resistors for each OCXO.

Cheers

Phil Hobbs

Yep. More and more of my "Analog" chips have built-in self-calibrating, controlled, most recently, by a uP via an SPI bus. So there's a significant digital content on the chip... mostly "quiet" except during a calibration cycle. ...Jim Thompson

You could just sense the current and turn on the LED below some value. There's probably a value that works with all your OCXOs.

I like an LED that lights up proportional to oven current, bright at turn-on and getting dim after warmup. More drama.

I guess a uP could have an algorithm that observes the current and decides when the oven has stabilized. But that has lots of corner cases. Seems like a lot of work for one LED.

It could easily cost us kilobucks to stock a small uP, get the development and programming tools, write and test the code, and put all that into production. That extends time-to-market, too. That might make sense if we planned to build tens of thousands of units. Otherwise, stuff one resistor alongside the selected OCXO.

The laser trimmer would cost 10s of kilobucks, and needs to be programmed, and the board would have to be powered up and communicated with while it's in the laser trimmer. And a laser can cut things but it can't un-cut them.

A small screwdriver seems like a better investment.

The Murata trim caps that we stock are 3x3 mm and cost 25 cents.

There were some variations within the same OCXO P/N, though. The (proporti onal) ovens were set at the factory to target each crystal's individual tur nover temperature. The spread in post-warmup current consumption at 25C wa sn't very large, but it was there, and it seemed like a good idea to accomm odate it with a trimmer. The ability to work with different OCXOs was a se condary consideration.

-- john, KE5FX

I thought about doing that but it seemed too vague to be useful. The idea was to be able to tell at a glance if the OCXO was finished with its initia l warmup. A simple 2-minute time delay would have been fine, but nooooo... .

There was also some overshoot when the oven controller reached its target t emperature. Already having gone full Steve Jobs with my OCXO monitor LED, I didn't want it to waffle between blue and white for several seconds while it rang down. Naturally the extent and duration of the ringing varied fro m one part to the next. At the same time I didn't want to use a lot of hys teresis on the comparator in case the oven came back on during operation, s o the trimmer setpoint was fairly critical. There were at least a half-doz en better ways to implement this "feature," but it just didn't seem to be w orth much thought at the time.

This particular board already had a 48 MHz 8051 and a 1.2M-gate equivalent FPGA, so if I had it to do over again, I'd read the current-sense resistor with a cheap ADC.

-- john, KE5FX

Do you really need an OCXO heater status LED?

Hard to say, always room for debate there. Nobody really needs anything but air, food, water, and shelter. But life's little luxuries are what make our existence bearable, and I count OCXO heater status LEDs among them.

More seriously, the LED serves to discourage users from trying to measure phase noise and Allan deviation with a box that they just powered up 12 seconds earlier. And it's a useful sanity check for one of the more important components in the instrument.

I really like RGB LEDs. You can communicate a surprising amount of information with just one of them, even allowing for color-blind users.

-- john, KE5FX

My liquid CPU cooler pump has a throbbing red LED. Not flashing, it's PWM'd or something to make it throb. I guess red is the new blue.

"Need" is a somewhat fluid concept in the developed world. Like the fluids (presumably mostly genunine Chinese tapwater plus antifungals) my gag gift LED-lit computer speakers squirt internally, in time with the sound.

--sp

What's the p/n.

MFR1 DIGIKEY 490-2008-1-ND MFR2 MURATA TZY2Z2R5A001R00

The low values are NP0 +-300 PPM, which is mediocre but still better than a varicap. The digital caps are better, ballpark 30 PPM.