[In this episode of The Chinesium Mine we summon the ghost of Jim Williams to help reverse engineer some very bad frequency meters]
Back when I had the Lister diesel generator I spent some time looking for a frequency meter for it. I eventually settled on the “D69-Hz” digital panel meter from a questionable eBay store.
I bought 2 of these meters and was rather disappointed, both examples seemed very sensitive to temperature. And I wanted to use them to measure small changes in frequency in a location that sees big changes in temperature. Bother. The proverbial chocolate teapot (also useless because of excessive temperature sensitivity)
Out of curiosity I decided to dive a little deeper to figure out just why they were so bad.
The obvious way to make a frequency meter is to use some kind of timer/counter algorithm to count cycles of the unknown frequency against a stable reference. But that was too obvious for our Chinese designers. This meter uses a frequency-to-voltage converter bolted onto the front of a digital voltmeter.
The F-V converter is a monostable made from a 555 timer chip, that generates a narrow pulse for each zero crossing of the input. These pulses are then averaged by the digital voltmeter chip. Everything runs off a single 3.3V rail, and the 555 timer used is of the CMOS variety, a LMC555.
As the legendary, late Jim Williams explained in his writeup of the “zoo circuit” (see figures 23-16 to 23-18) the output stage of a CMOS digital IC is made of MOSFETs that switch cleanly from one supply rail to the other. There are no Vbe voltage drops like with bipolar transistors.
Williams was talking about a CMOS logic inverter chip, the 74C14, that he used to squeeze some more performance out of his “zoo circuit” V-F converter. But a CMOS 555 timer chip has the same output stage as a 74C14, and the designers of our Chinese F-V converter used CMOS here for the same reason: to eliminate the error due to the temperature dependency of those Vbe’s.
I scoped the output of my so-called “LMC555” and discovered that it was only putting out about 1.8V peak-peak with its 3.3V supply, and the output increased when the chip was warmed up.
Hah! a fake! A common or garden bipolar 555 timer remarked as a more expensive CMOS version. By far the most likely explanation for the missing 1.5V of output is two bipolar transistor junction voltages at 0.7V each. The 555 timer internal schematic on Wikipedia shows a Darlington pair for the high side of the output stage, so that would be our two Vbes right there. The low side is a common emitter, which would only drop a few hundred millivolts.
I yeeted the counterfeit chip out of there and substituted a TLC555 (which happened to be the cheapest CMOS 555 timer in Farnell’s online store at £0.66)
C3 is the monostable timing capacitor, I had to change it from 5.6nF to 3.2nF to reduce the pulse width in compensation for the increased output voltage of the new 555 chip.
Having done this, R12 was tweaked to read 50.0Hz with a mains frequency input at room temperature. Most Chinese meters I’ve seen so far have a “fortune trimmer” like R12 that can be twiddled to give almost any answer you want. I don’t think I even want to know the tolerance or tempco of this part.
My modified D69-Hz seemed much more stable, so I put it in the Hayburn Labs environmental chamber alongside an unmodified one and a thermometer.
The environmental chamber is not a precision instrument (actually it’s a freezer from a charity shop) but it’s still incredibly useful for freezing and boiling prototype circuits to discover any temperature-related quirks before the customers do.
The mains frequency varied by more than I expected during this experiment, due to those meddling wind turbines. I had to go and look up the actual National Grid frequencies corresponding to the photo timestamps.
At 23C the meters read 49.7 and 49.9. National Grid official value was 50.111Hz so our modified meter is reading 0.211Hz low.
After refrigerating to 0C and allowing some time for temperature to equalise, the stock meter gave a reading of 47.9 and the modified one 49.6. Actual mains frequency was 50.041Hz so the modified meter has an error of -0.441Hz.
Stewing at 40C for 20 minutes produced a result of 51.6 for the fake 555 and 49.7 for the TLC555. Actual mains frequency was 49.881Hz so the modified meter error is -0.181Hz.
So over the 0-40C temperature range, we have an error of 3.7Hz (7.5%) for the unmodified meter and 0.26Hz (0.5%) for my modified one.
Conclusion: When I originally tested the Lister diesel, it gave a frequency of 51.25Hz unloaded and 50.00 on full load. And the temperature in Container Labs varied from below zero, to over 30C. The error with temperature of the unmodified meter would be more than twice the generator’s actual change of frequency, so it would have been completely useless.
By replacing the fake 555 chip, we reduced the temperature sensitivity by a factor of 15. But the error still works out as 20% of the frequency range covered by the Lister diesel.
As a check I got out the same Fluke meter that I originally used to test the generator. It agrees with the National Grid website to within 0.01Hz.
(National Grid official value was 49.854)
(National Grid official value 50.034)
With this level of accuracy, I guess Fluke’s budget must have stretched to an actual quartz crystal in there somewhere, to give an accurate reference for a frequency counting algorithm. Our poor D69-Hz meters probably have a RC oscillator for the clock.
You can still buy the D69-Hz on Amazon, if you’d like a more laborious way of wasting money than simply setting fire to a £10 note. Believe it or not, I don’t earn a commission if you buy through this link. 😀
The 5.6nF capacitor and phoney “LMC555CM” are visible in the product photo, so I’m pretty sure it’s the same Shenzhen Chocolate Teapot Industries product.
Rating: 1 Zhong out of 5
🀄️Depleted Chinesium