1146 THE BELL SYSTEM TECHNICAL JOURNAL, OCTOBER 1951 



The ring-type record-reproduce head shown in Fig. 1 was lapped sHghtly 

 to obtain a reasonably good fit with the surface of the disc. 



A single-frequency recording was made with the head in contact with 

 the disc using a-c. bias in the usual way. Then the open circuit reproduced 

 signal level was measured, first with the head in contact, and then after 

 introducing paper shims of various thickness between the reproducing head 

 and the medium. Thus the effect of spacing was measured at a particular 

 frequency and recording speed. The signal was then erased and the process 

 was repeated for other recorded frequencies and for several record-repro- 

 duce speeds. Measurements were also made for cases in which the recording 

 and reproducing speeds were different. Considerable care was required to 

 keep the disc and head sufficiently clean so that reproducible results could 

 be obtained. 



HIGHLY POLISHED PLANE 

 SURFACE PLATED WITH 0.0003" 



COBALT NICKEL ALLOY \ RECORDED TRACK 



^ /^ 0.1 18" WIDE 



PIVOT 



/ 

 RECORD- REPRODUCE 

 HEAD (brush BK9n) 



-BRASS DISC 



Fig. 1— Mechanical arrangement of recording set up. The one head served for re- 

 cording, playback, and erase. 



Figure 2 shows typical response curves measured at 21 in./sec. with the 

 reproducing head in contact and with 0.004 inch spacing. The difference 

 between these two curves will be called the spacing loss corresponding to 

 this spacing and speed. From these data and more of the same sort it is 

 found that, within experimental error, spacing loss can be very simply re- 

 lated to spacing and the recorded wavelength, X, by the empirical equation, 



Spacing loss = 55{d/\) decibels (1) 



where spacing loss is the number of decibels by which the reproduced level 

 is decreased when a spacing of d inches is introduced between the repro- 

 ducing head and a magnetic medium on which a signal of wavelength X 

 inches has been recorded. 



The fact that this expression fits the experimental data reasonably well 

 is indicated in Fig. 3 where spacing loss data measured at a number of dif- 

 ferent speeds, frequencies, and spacings are plotted against d/\. 



