Correspondence 329 can be obtained. For n stages, the over-all voltage transfer function is D The parameters of the kth section may be found by solving from the expressions for center frequency 1 d C k L k Wk = = (4) and fractional 3-dB bandwidthThe fundamental principle underlying the design of constant-resistance networks similar to those discussed is the dimensional equationBy utilizing this relation to constrain certain element values in a network, it is also possible to design constant-resistance low-pass, high-pass, band-pass, and band-stop filters of II, T , bridged-T, and lattice configurations.A 16-dB gross improvement in the minimum detectable signal of an infrared detector using a Xenon gas laser preamplifier with 17-dB gain was recently reported by Bridges and Picus [l].We have independently investigated the use of a Helium-Neon laser preamplifier which had substantially higher gain. I t utilized the very high gain transition a t X = 3.39p, thereby permitting operation without regenerative reflectors and thus yielding amplification over the full Doppler linewidth. Measured improvements in minimum detectable signal of 45 dB (gross) and 32 dB (net) were obtained relative to a room temperature PbS detector.' S o amplifier noise was observed in our experiments and none should have been detectable as the following analysis indicates. The S S R for a system consisting of a laser amplifier followed by a square-law envelope detector, for a coherent input signal under matched conditions and with single-mode operation, can be represented by where P, = signal hvB/q. =amplifier input quantum noise power P I h = Planck's constant v=infrared signal frequency B =amplifier effective instantaneous G =amplifier gain q,, =amplifier population inversion Af=post-detection bandwidth d=quantum efficiency of envelope detector NEP=Soise Equivalent Power of detector (referenced to Af =lc/s) square bandwidth efficiency= 1 -(nr/gl)/(nu/gu) A general theoretical treatment has been presented by Steinberg [3]. The three terms in the denominator represent [4]added fluctuation noise, shot noise and detector NEP, respectively. Additional contributions due to 290°K background are neglected in the wavelength region considered here. Also, we take G -1 =G.For threshold signals, neglecting the relatively small shot-noise contribution and Af<>hvd2B/qa. Thus, for infrared wavelengths, e.g., X > lp, substantial improvements are obtainable, since as wavelength is increased, quantum noise decreases and detectors have progressively higher NEP's. For example, for X =3.39p, and taking B = 100 Mc/s, Af = 1 cps, q. =0.64 and quantum noise limited conditions (large G) a P,,,=l.2X10-'6 watt/cps1'2 is calculated. In our experiments and those of [l], detectors with NEP's in the range 10-10 watt/ cps"* were utilized. Using this value of NEP, and G = 3 2 dB, q.=O.64, and B=l00 Mc/s [5], the last term in (2)...

MARCHThe reason for the dif€erence is obvious as the diode impedance varies periodically from a low to a high value. In the first case, this impedance is in series with the resonant circuit and the change results in strong FM and AM signal components. In the second case, the loop-coupled diode appears as a shunt impedance and the main effect of its time-varying impedance is to produce small changes in the cavity Q, and hence slight amplitude modulation at the signal.Using the arrangement in Fig. l(b), reasonably pure waveforms could be obtained for harmonics from the eighth to the sixteenth. Measurements with a receiver for the eleventh harmonic, for instance, indicated that unwanted harmonics were over 33 dB below the level of the required signal.No s i d c a n t steps were taken to optimize the input matching, and the coupling loop was adjusted to be a reasonable match only for the twelfth h o n i c . Even so, the e5ciency was about 9 percent for the eighth harmonic, dropping to 5 percent for the fifteenth.Unless elaborate filter circuits are to be used, it is obvious from the above simple experiments that the diode circuit must be carefully arranged so as to minimize the effects of the diode impedance changes. (watt) (wan) 2.030 38.5 x 1.462 21.6 x Detector responsitivity 170 V,W with black polyethylene 6ltcr. -C E S [I I S. M. Krakauer, "Harmonic generation, rcftification, and lifetime evaluation with the [2] K. L. Kotzebue, "A circuit model of the S t e p m v c r y diode," Proc. IEEE (Come-step mzovery diode," RM. Abstract-Measmme& of the peak a d average ouQmt powof a 337damncyamidekser.rergorted.Tktestsservedtoiadicatetbe.sef.l-~a t~B i g B~P q . e a c y a * o v~w a v~~ . *l i ona-av-wer--,atiquidb codedIpSb--IUiX-ud ¶-8thlUtOr.This letter reports measurements of the peak and average output power of a recently developed submillimeter far-infrared cyanide laser' oscillating at 337 microns (890 GHz).The cyanide laser, l o c a t e d at the M.I.T. National Magnet Laboratory, Cambridge, Mass.' and manufactured by G . and E. Bradley and Co., is similar to that of Gebbie et al.' It uses a continuous flow of butyronitrile in a 2-meter-long discharge tube. Power is coupled out by a hole in one of the reflectors into l-cm-diarneter light pipe (oversize cylindrical waveguide). Pulsed output power at a 1-100 pps repetition rate is optimized using a Golay cell.Pulse characteristics were determined using a hetiumcooled indium antimonide detector-mixer having a time constant near lo-' second^.^ Pulse shape was found to vary from pulse to pulse, possibly due to multifrequency operation. A scope photograph of a representative pulse is shown in Fig. 1. The InSb detector response was not increased at this frequency by the application of a magnetic field to the detector. Average power was measured using a recently developed room-temperature thermal detector,' useful from 50 GHz to 10 microns. A black