August 1, 1995

DISTORTION

It has been customary to test modem transformer distortion at 600 Hz, -10 dBm. This was reasonable before V.32. While this was maintained for V.32, it was only a happy coincidence that the units worked. It is utterly useless for V.34.

All transformers are power/frequency rated devices. That is, reducing a transformer's operating frequency in half, will cut its power handling capacity four times. Since modems have to supply a 300 Hz -10 dBm signal, this is equivalent to -4 dBm at 600Hz. Additionally the 300 Hz signal is not a sine wave, adding at least another 6 dBm. Indeed, 0 dBm, 300Hz, is not an unreasonable level to measure distortion at.

Distortion of a transformer, changes only slightly at low levels, but will increase asymptotically as levels are high enough to cause the transformer to approach saturation. Thus a transformer may have -90 dB distortion with a 600 Hz, -10 dBm signal, but only -20 dB, at 300 Hz, O dBm. Reducing the 600 Hz distortion by 6 dB may also reduce the 300 Hz distortion by 6 dB, but the 70 dB difference between 600Hz -10 dBm and 300Hz/0dBm has not been solved.

Distortion is NO PROBLEM at low levels and high frequencies. It is only a problem when transmitting a high power low frequency signal. It is this signal that adds errors to a received signal under duplex operation. A lower transmitted signal would not only have less distortion, but from a lower signal level. That is, a 10 dB lower transmitted signal, with 20 dB lower distortion, would feed back a 30 dBm lower distortion caused error signal. Accordingly a transformer that has -100 dB distortion at 600Hz -10dBm will perform no better than one with only -75dB distortion. Specmenship abounds on this point.

Solution: Rate the unit at actual lowest operating frequency and highest operating level. BG supplies differing units rated from -40 to -70 dB distortion at 300 Hz, 0dBm. Unfortunately, present published distortion measurements have been known to contain a bit of poetic license. (It is also true that at 600 Hz, -10dBm, few can measure distortion with accuracy.)

Whether it is because of differences between differing modem manufacturer's parts, or lack of sufficient engineering rapport between modem engineering and transformer engineering, absolute distortion figures are, at present, insufficiently defined. Empirical data from actual modems seems to be the rule.

Howard Buschman, President

Home <www.bglabs.com>