|Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting
Departments | Features | ADs | Equipment | Music/Recordings | History
SEVERAL READERS have asked me what I thought of the recent Consumer Report's tests on loudspeakers using a computer to evaluate the results. Briefly, what CU did was to feed the speakers with a pink noise signal and then take power response readings at 10 degree intervals in two perpendicular planes. At each angle, 30 readings were made, representing the rms value of 1,000 measurements in each of 30 one-third octave bands. The computer converted the data from decibels to sones and then it was transposed into simple percentages. Frequencies below 110 Hz were excluded because of room variations and differences due to positioning.
Now, I must admit that I was horrified when I'd read this far. A speaker with several 2 dB peaks could be lumped together with one having an enormous 15 or 20 dB peak somewhere in the spectrum. There would be no distinction between a peak at 150 Hz and one at 12 kHz or 7 kHz. And how do you grade coloration? As all speaker engineers know, tiny 1 dB peaks--or even smaller at certain frequencies-can cause quite severe colorations. The density of the enclosure material plays a large part in these variations which can hardly be measured much less evaluated by a computer! However, on reading further I discovered that a listening panel was also used, and it was claimed that they confirmed the computer analysis "to a reasonable degree." Looking at the list of speakers tested, I can well believe it. None were really bad, and I can think of several other systems which would score quite high on the percentage tests but would sound abominable. Incidentally, the CU listening tests involved comparisons with a reference laboratory speaker using tapes originally made with that loudspeaker in an anechoic chamber-a method of evaluation which allows a high degree of accuracy.
So all-in-all, I would say that I wouldn't disagree with the CU conclusions, although I have strong reservations about those computer-derived percentages. The highest score was 89 percent; my guess is that a speaker with a score of 100 per cent CU accuracy would still sound like a loudspeaker, that is it would still have some distortion and some coloration.
Measuring Wow & Flutter
The article by Robert Berglas aroused a great deal of interest, and most people saw the snags I mentioned in the footnote. Gary Flynn, of Atlanta, Georgia, writes: ". . . an instrument of this type (Heath-Schlumberger counter) measures the average frequency during the sampling period, which in the Heath instrument is 1 second. The flutter modulation of the 1 kHz carrier, however, since it is an a.c. signal like any other, has an average value of 0, and therefore has no effect on the average frequency of the carrier as read on the frequency counter."
D. E. Peter, of Hollywood, Calif., says: "To determine a peak-to-peak wow component at 6 kHz, the counter must count for about 1/4 cycle of 6 Hz or 1/24th second. Assuming a standard flutter tape is used, the counter will count 3150 Hz x 1/24 second or 131.25 Hz. The 0.25 Hz will cause the output sequence to be 131 ... 131.. . 131 ... 132 ... 131.... The uncertainty in the last digit caused by the non-coherence of the counter clock and the counted frequency will cause a measurement uncertainty of 0.760 percent. In other words, the wow measurement will be obscured by noise if the wow is less than about 0.5 per cent, rendering the technique useless." As Mr. Peter says, a computing counter, such as the Hewlett-Packard 5360 series, could measure wow and flutter combined but this seems like a complicated way of doing things.
A. R. Collins, of the Acoustical Company in England, comes up with some different figures. He writes: "Firstly, the plus-or-minus one count ambiguity in the counter dictates that, with a 1 kHz reference signal, the accuracy of the reading will be subject to an error of plus or minus 0.1 percent, assuming the 1 second gate time which is common on counters measuring frequencies in the region of 1 kHz.... To resolve wow components up to 6 Hz, the sample period shall be 1/12th second or less, in which time the counter will indicate 83 cycles. The plus or minus one count uncertainty will degrade the accuracy to worse than plus or minus I percent. The use of a higher test frequency to permit the use of short gate intervals will improve the accuracy." Mr. Collins goes on to say that a phase-locked loop method is inherently a better technique for wow and flutter measurements than a digital system and one such application was described in Wireless World for December, 1971. The author is R. Ockleshaw, and some of his remarks on manufacturer's specifications would agree with those made by Robert Berglas. In brief, the PLL system produces an output voltage which is instantaneously proportional to the difference between the incoming frequency and a reference frequency. Any phase/frequency error is transformed into a changing d.c. level. The complete circuit uses three ICs plus a single transistor and should cost no more than $25.
This is a handy gadget made by HR Manufacturing, of 1917 57th St., Sarasota, in sunny Florida. This company formerly made the Rabco turntables and arms, now made by Harman-Kardon. What does Control I do? It switches off the amplifier, receiver, tape recorder, or whatever in the absence of a signal. Two wires are connected to the speaker sockets and the equipment to be switched is plugged into a socket at the rear of Control I, which in turn is plugged into the power outlet. That's all. I found that signals as low as 50 milliwatts would keep the unit switched on so you could use Control I for very soft background music. (Some very interesting applications here!) Switch-off time depended on the signal level and it varied between a few seconds up to just over 9 minutes after the cessation of the signals. The circuit uses 3 transistors and a heavy-duty relay and there is an override switch as well as a manual on-off switch. Price is $29.95.
(Audio magazine, Oct. 1973)
= = = =
Prev. | Next