With Neither Paper Nor Napkins
I have a story pertaining to "Ode To The Paper Napkin—The Untold Story," \[Aug. 7, p. 54\]. Back in 1955, I was fresh out of college and working for a small transmitter company, ITA, that's no longer in existence. One time, I was driving with the owner of the company from Philadelphia to Chicago to exhibit our equipment at the National Association of Broadcasters show. When we reached the Pennsylvania-Ohio border, we ran into a severe snowstorm. We left the turnpike and checked into the nearest motel (one star). To kill some time, we decided to design a remote control system for our transmitters that would use standard telephone lines.
Needless to say, we had neither paper nor napkins. We welcomed the good, strong, cheap, untextured toilet paper of the second-rate motels. One and a half rolls later, we had a design that worked the first time and was in production for several years. As the only engineer, aside from the boss, I was given the job of putting it on real paper as soon as we returned from the trip. It was reduced from 152-ft. by 5-in. toilet paper to two "D"-sized sheets
Stanley S. Friedman
DRS Communications Company, LLC
Getting Help To Build A Test Board
I read the article "Pay Attention To The Clock And Output Bus To Improve High-Speed ADC Designs" \[June 26, p. 137\]. It was very interesting for our group. We develop high-speed ADCs (40-MHz, 10-bit Nyquist) and are trying to build up a test board for characterizing. Perhaps you can give us advice for two basic questions: Do you know any sufficient (analog) input signal source? Are there any sufficient bandpass filters available on the market?
Thomas van den Boom
Author replies: The basic requirements for a signal source for testing high-speed ADCs are that it has low phase noise (jitter) and produces a signal of low distortion. The low phase noise isn't a problem for many signal generators, but the low distortion requirement is a real problem, even for the best generators available. For this reason, I suggest a bandpass filter between the generator and the ADC input.
We have found the TTE "LE1161" series of filters to perform well. They have a 3-dB bandwidth of 15% of the center frequency. We found that the TTE filters produce a good, clean sine wave with a stable square-wave input too. We use the HP 8662 signal generator, but inexpensive generators also perform well when used with the TTE filters, as long as the generator doesn't have significant phase noise. You can tell if there's a lot of phase noise as "spreading" will be seen around the base of the carrier in an FFT plot. This spreading looks kind of like a pine tree and is often called a "Christmas tree."—Nicholas Gray
The Industry Still Has Ethics
Though I have no need at present to blow a whistle, I really enjoyed the implied ethics in your column of the Aug. 7 issue \["Whistle Blowing: Should I? And How To Do It Safely," p. 150\]. It's good to hear of people with deep experience in the industry that really care about ethics!
Try as we will here (LANL), the real academics of ethics needs to be thrust upon organizations, such as the DOE, the FBI, and Congress, if this country is to maintain any kind of motivated and talented crew at the National Labs. That's a subject too involved to get into here, but thanks for your article.
First, Find Out What's Going On
I just thought I would drop you a note to point out that Don is full of it \["Garbage In Does Not Always Mean Garbage Out," July 10, p. 66\]. His statistical math is physical nonsense. It's the electronic equivalent of saying that you can predict exactly what an unknown signal will look like by turning down the gain on the scope until it flatlines. That's true, but utterly useless. It makes no sense in electronics, physics, planning, or scheduling. The best approach is always to figure out what's going on so that you can take action to resolve the problems. It is, of course, much easier for planners to remove any such data from their high-level considerations and then just be surprised and disappointed when programming takes twice the time allotted, as usual.
Response: The concept of vector addition of uncorrelated variances is pretty solidly accepted even within Boeing where it's used every day to do statistical analysis of mechanical stack-ups. It also lurks behind signal processing techniques, like boxcar integration, where we improve signal to noise on a repetitive signal because the signal is correlated and noise is not. It has been used in project planning for at least 40 years. Nevertheless, from the tone of your message, I suspect that I'm not the person to convince you of this.
Happily, we both agree on the value of "figuring out what's going on so that you can take action to resolve problems." I'm sorry that my column wasn't more convincing.—Don Reinertsen
Don't Let Marketers' Lingo Fool You
Just to keep things technically straight and not propagate misleading information (being the exacting engineers that we are), "brightness" is perceived and can't be measured \["Both Larger And Smaller Sizes Are Fueling The Innovation Of Flat-Panel Displays," Aug. 7, p. 83\]. "Luminance" is measured and, therefore, has associated units, such as nits, foot-lamberts, etc. Contrast ratio is then a ratio of luminance values. For some reason engineers have picked up on the marketers' cop-out and continue to promote this incorrect terminology.