Electronic Design

Bob's Mailbox

I would like to ask about designing a sinewave amplitude attenuator with programmable attenuation. Preferably, it will just contain basic components (op amps, transistor, resistors, caps, etc.). Input: 1-V p-p sine wave (1 kHz frequency), symmetric at 0-V level. Desired output: still at 1 kHz, but the amplitude varies from 0 to 1000 mV.


For the last 30+ years, the multiplying digital-to-analog converter (MDAC) has been a good way to do this. A 10-bit MDAC will attenuate a 1024-mV p-p sine wave to any amplitude such as 1023, 1022... 1001, 1000, 999, 998... 637, 636, 635... down to 1 and 0 mV p-p. These days, nobody builds their own MDACs because the ones you can buy are so good. You can also get 8- and 12-bit MDACs. Can I recommend a part number? How about the DAC101S101 (bus-compatible)? Look it up at National’s Web site at www.national.com.

Now if you have decided to take on a project, to reinvent the wheel and make your own DAC, to use hundreds of parts to generate the same function as two chips (the DAC and the op amp), be my guest. But nobody has been publishing circuits on how to do this for 30 years, because the item you buy for $2 is so much better.

You could put a string of 1000 22-O 1% resistors in series. You could put in 999 switches. You could decode the digital code. You could do it in binary or in BCD. You could make it very accurate. You could solder day and night. But that’s just to show that if you had a time machine, you could go back to 1950 and make your own MDAC. Bulky, slow, expensive. Have fun. I did design a 10-bit MDAC module around 1972. But it had to have low phase shift at 1/2 MHz, very fast settling, and low glitches. And to this day, that is not easy to do on one chip. It was 2 by 4 by 0.39 in. high, and it was beastly to make. We finally figured out that the customer had made a mistake when he wrote up the specs, so we got out of the contract and sent him something else that made him happier.


I recently purchased some germanium TO-3 power transistors just for fun to see if I am smart enough to use them. The newer IGBTs, etc., are wonderful. You can switch lightning voltages (almost), and they are certainly the way to go. But I have an old CD ignition system that I have had on three old cars over the course of 23 years, and I picked it up off a junkyard car when it was 20 years old. So 43 years of use with germanium power transistors under the hood says something for a design.

I have noticed the old TO-3 transistors have a very nice lightweight aluminum low-profile case. It must be good for heat transfer, corrosion, and weight savings. Why were steel cases ever used for TO-3 transistors? Steel is heavy, it offers poor heat transfer, and it rusts. Could cost be that high for aluminum?


Aluminum TO-3s are cheaper and used when the customer wants the lowest price. The thermal impedance is about the same as steel. (Aluminum is only better as a heatsink on a per-ounce basis.) But the weakness of aluminum TO-3s is for thermal cycling. Even with the best die attach, after about 5000 full-range (150°C to cold) temp cycles, the die attach gets flaky and turns into a cold-soldered joint, and the thermal impedance goes way up. If you’re running a high-power application, as you might if the temp cycling is extreme, the die can overheat. In the steel package, the life is 40 times longer, or more. If you don’t run the die to 150°C, on an LM317, the degradation is much less. If you only go to 85°C, which most germanium can stand, the degradation is much lower. In a “transistor ignition,” the germanium transistor’s die will rarely get above 45°C, so it will last a very long time. At National, we haven’t made aluminum TO-3s for more than 20 years.


Your suggestion in your Feb. 28 column to connect a solid copper ground plane to the neutral conductor of the power line is dangerous. Although the neutral conductor is nominally at earth potential, load currents through the impedance of this line will raise the neutral conductor and ground plane above earth potential. This will become worse if the neutral connection becomes flaky or if there is a fault and the lines are improperly fused, which is more likely on an experimenter’s workbench. Instead, the ground plane should be permanently bonded to green-wire ground. A groundfault circuit interrupter on the mains should also be used.


Thank you for the correction.


Comments invited! [email protected] —or:
Mail Stop D2597A, National Semiconductor
P.O. Box 58090, Santa Clara, CA 95052-8090

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.