•iSO THE BELL SYSTEM TECHNICAL JOURNAL, AL^RCH 1957 



from the cavity. This also provides a convenient way to measure the Qo 

 of the cavity. " The frequency is measured roughly b}^ means of a cavity 

 frequency meter and more precisely by means of a transfer oscillator and 

 high speed counter. The magnetic field is measured by means of a nuclear 

 magnetic resonance set-up, its frequency being measured on the same 

 counter as the microwave frequency. The nuclear resonance signal is 

 recorded on the same trace as the electron resonance signal. Thus if the 

 magnetic field is homogeneous enough, the nuclear sample will see the 

 same field as the electronic sample and g-values can be conveniently 

 determined to the accuracy of the nuclear moment (this also assumes 

 that the signal is large enough, so that no additional error is introduced 

 in determining the exact location of the resonance.) The field modula- 

 tion coils are mounted on the pole faces and are energized by a oO-watt 

 power amplifier. A field of 50 oersteds p.t.p. is available at 1,000 cps and 

 a slightly higher field at 100 cps. 



The magnet is a Verian 12" modified so that it can rotate around an 

 axis perpendicular to Ho . This was done mainly in order to make 

 anistropy measurements more convenient. This enables one to make 

 quick saturation measurements in isotropic materials without having to 

 change the incident RF power. This is accomplished by rotating the 

 magnetic field and measuring the signal strength versus angle. Since onlj'- 

 the RF field perpendicular to the do field causes transitions, the signal 

 in an unsaturated isotropic sample should go as cos^ d; where 9 is the 

 angle between Hi and Ho . From the deviation from tliis dependence, the 

 saturation parameter can be found. This could also be done by rotating 

 the cavity, but at microwaves is not as easy as rotating the field. 



VII. A NOTE ON THE EFFECTIVE BANDWIDTH 



There seems to be some confusion as to how narrow one should make 

 an audio amplifier preceding a phase sensitive detector (lock-in) or why 

 the band width of the IF amplifier doesn't enter in a superhet scheme. 

 Those and similar questions have to do with the effective band width 

 of the system Ap which appears in (26). Since similar questions have 

 been rigorouslj^ analyzed by other authors,^^' ^^ the present discussion 

 will try to stress some of the phj^sical ideas underlying the different 

 detection schemes. 



We consider first the simple scheme illustrated in Fig. 11. It consists 

 of an amplifier with band width Aj^i centered around vi followed by a 

 phase sensitive detector with a reference voltage at Vi . The output of 

 the phase sensitive detector has an RC filter of band width Ai'2 . One 

 can see that in such a system the only noise components centered around 



