ALMA Memo May 2003 MEASUREMENT OF GAIN COMPRESSION IN SIS MIXER RECEIVERS

Transcription

1 Presented at the 003 International Symposium on Space THz Teccnology, Tucson AZ, April ALMA Memo May 003 MEASUREMENT OF GAIN COMPRESSION IN SIS MIXER RECEIVERS A. R. Kerr 1, J. Effland 1, S.-K. Pan 1, G. Lauria 1, A. W. Lichtenberger and R. Groves 1 1 National Radio Astronomy Observatory Charlottesville, VA 903 University of Virgia Charlottesville, VA 904 ABSTRACT When an SIS mixer is partially saturated by broadband noise, it contues to exhibit a lear response to a small CW test signal, with the small-signal ga dependg on the level of the saturatg noise. This allows the CW test signal to be used as an dicator of the receiver ga the presence of high-level noise. If not taken to account, ga compression can be a significant factor limitg the accuracy of high precision radio astronomy struments. Keywords: Superconductor-Insulator-Superconductor mixers, saturation, ga compression, dynamic range INTRODUCTION The Atacama Large Millimeter Array (ALMA) is strivg to achieve an absolute flux measurement accuracy of 1%. Plambeck [1] has poted out that ga compression (saturation) SIS mixer receivers is likely to be a significant factor limitg the measurement accuracy. In this paper, we report measurements of the saturation of an SIS receiver the 30 GHz band and show that this band the thermal noise from a room temperature black body source is sufficient to produce ~ 1% ga compression a mixer with four junctions series. At the 00 Space Terahertz Technology Symposium we described a method for calculatg the ga compression an SIS mixer with a broadband noise put []. It was noted that, even when partially saturated by a high-level noise signal, the response of an SIS mixer to a small CW test signal is lear, the small-signal ga dependg on the level of the saturatg noise signal. This allows the CW test signal to be used as an dicator of the receiver ga the presence of high-level signals.

2 MEASUREMENT PROCEDURE The ga compression measurements described here were made on a mixer-preamplifier for ALMA Band 6 (11-75 GHz) [3] with an termediate frequency of 4-1 GHz. The measurement setup is shown Fig. 1. A small CW test signal is troduced through the LO waveguide while the receiver put is switched between liquid nitrogen and room temperature sources usg the chopper wheel. (The 0 db cold pad between the mixerpreamp and the IF switch allows similar signal levels to be mataed throughout the IF system when measurg the mixer-preamplifier and when measurg the ga and noise temperature of the IF system usg the hot and cold IF loads the Dewar.) The RF test signal is adjusted to give an IF power output power P HS 10-0 db above the output noise power P H with only the hot load at the receiver put. With the cold load at the put, the IF power levels with the test signal on and off are P CS and P C. The output power due to the test signal alone is (P HS - P H ) with the hot load at the put and (P CS - P C ) with the cold load. If it is assumed that the receiver is not significantly saturated when connected to the cold load, it follows that the percentage ga compression due to the hot load is given by PHS P H C percent. (1) P P CS The mixer-preamp may not be the only contributor to saturation; fact, it is likely that the IF amplifier cha will have some degree of ga compression. This can be particularly important when the full IF bandwidth is large compared with the bandwidth of the fal filter before the power meter. If the percentage ga compression of the mixer-preamp is C 1 and that of the rest of the IF cha is C, then the percentage ga compression for the whole C Fig. 1. Setup for measurg ga compression. --

3 receiver is, to first order, C = (C 1 + C ). To measure the ga compression of the amplifier stages followg the mixer-preamp under test, a small IF test signal can be troduced through the directional coupler followg the IF switch Fig. 1. In measurg C, it is important not only to ensure that the IF put (noise) power is the same as when C is measured for the whole receiver, but also that the put noise spectrum has the same shape. MEASUREMENTS Figure shows the data from a typical saturation measurement. The SIS mixer has four junctions series and the LO frequency is 30 GHz. The percentage saturation is seen to vary almost susoidally with a period of ~8 mutes. When liquid nitrogen was added to the cold load, a change of phase was observed the ripple, dicatg that it is caused by the reflection of the test signal at the surface of the LN. LO power emergg from the receiver is also reflected at the LN surface, but the magnitude of the LO reflection, dicated by the modulation of the DC mixer current, is sufficient to contribute significantly to the measured ga variation. To check that the CW test signal does not itself cause significant saturation, its level was changed by ~4 db durg each measurement and no significant change was observed the results. Fig.. Ga compression (saturation) data as a function of time, measured on a four-junction SIS mixerpreamp with the LO at 30 GHz and the small test signal the upper sideband. The susoidal ripple is caused by reflection of the test signal at the surface of the LN the cold load as it boils away. The horizontal end segments at 0.5% dicate the degree of ga compression the IF stages followg the mixer-preamp. Saturation the IF amplifiers followg the mixer-preamp was measured with exactly the same setup used to measure the overall receiver saturation, except that the test signal is now jected at the IF through the 0 db coupler the Dewar, shown Fig. 1. The RF hot and cold loads are connected at the receiver put as before, thus ensurg that the IF amplifiers see the same noise power and spectral characteristics as the overall saturation measurement. The results of this measurement are shown the end segments of the data Fig. and dicate that ~0.5% saturation is due to the IF amplifiers followg the mixerpreamplifier. -3-

4 DISCUSSION Large-Signal Ga Compression and Incremental Ga Compression Fig. 3 shows the output power of a receiver with ga compression at higher put powers, as a function of the receiver put power. With zero put power, the output power from the receiver is G 0 P Rx, where G 0 is the receiver ga at low put power and P Rx is the equivalent put noise power of the receiver. P H represents the put power from a hot noise source sufficient to cause some degree of ga compression and produce an output power G LS (P H + P Rx ), which is lower than the output power G 0 (P H + P Rx ) which would be produced the absence of any ga compression. G LS (P ) is the large-signal ga of the receiver. The cremental ga G c (P ) is the slope dp out /dp of the ga curve. At low put powers, G LS = G c = G 0. The cremental and large-signal gas are related by G c dp dp out dg ( LS P) dp. () Fig. 3. Receiver ga curve, P out vs P, showg large signal ga G LS and the cremental ga G c = dp out /dp. The quantity P Rx is the equivalent put noise power of the receiver. Our earlier analysis [3] of ga compression was concerned with the large-signal ga compression, which is difficult to measure at low levels. In the present work, we have described a simple method for measurg the cremental ga compression. The largesignal ga compression can be related to the cremental ga compression usg equation (13) of [] which gives the large signal ga as a function of the normalized RMS noise put voltage S : -4-

5 G LS G 0 1 V CV C S arctan 3 exp S 3 3 dv, (3) where e S GP R, C 3 = 3.3 is a constant for all SIS mixers, P sig is the signal 0 sig L Nhf put noise power, and R L is the IF load resistance seen by the SIS mixer. From this, the cremental ga is determed usg (). Figure 4 shows the large-signal and cremental ga compression plotted together as functions of S. Usg Fig. 4, it is possible to deduce the large-signal ga compression from the (measured) cremental ga compression, and thereby to apply an appropriate correction to a receiver ga calibration made usg hot and cold loads. Referrg to Fig. 3, the output power measured with the hot load front of the receiver G LS (P H + P Rx ) can be corrected to give the output power G 0 (P H + P Rx ) which would be measured if the receiver had no ga compression. For the mixer-preamplifier ga compression measurement shown Fig., the large-signal ga compression is 1% (after correctg for the 0.5% compression the followg IF amplifiers). From Fig. 4, the correspondg value of S is , and the large-signal ga compression is 0.5%. Fig. 4. Incremental and large-signal ga compression as functions of the normalized signal put noise power S. These universal curves apply to all SIS mixers. Agreement with Earlier Work Figure 4 can also be used to deduce the cremental ga compression of a mixer whose large-signal ga compression has been estimated as described []. Fig. 5 (from []) shows the large-signal ga compression produced, a 30 GHz SIS mixer with N junctions series, by a room temperature source, as a function of the sgle-sideband mixer ga, -5-

6 Fig. 5. Ga compression produced by noise from a room temperature source an SIS mixer with N junctions series, under the assumptions listed the text. (From [].) under the followg assumptions: (i) the put noise bandwidth B 1 each sideband is equal to 0% of the LO frequency, (ii) the IF load impedance is 50 ohms over the extended IF band 0 < f IF < B 1, and (iii) the small-signal ga is constant over 0.8 f LO < f sig < 1. f LO. We were not able to measure the ga of the mixer used these measurements because it was tegrated with the preamplifier but, based on experience with similar mixers without tegrated preamplifiers, we estimate the (SSB) mixer ga to be the range -3 to -7 db. Under the above assumptions, Fig. 5 dicates a large-signal ga compression of 0.6% to 1.5% when the receiver is connected to a room temperature source. This is slightly higher than the 0.5% large-signal ga compression deduced usg Fig. 4 from the measured cremental ga compression a discrepancy not surprisg given the uncertaty of assumptions (i) - (iii) used the theoretical calculation. Post-Mixer Noise Contribution In the present discussion, it has been assumed that all the noise power at the receiver output origates or before the mixer, and thereby contributes to the saturation of the mixer. In fact, noise origatg the IF preamplifier and subsequent amplifiers does not appear at the output of the mixer and therefore does not contribute to saturation of the mixer. In most practical cases this will cause little error the ga compression analysis the noise contribution of the IF amplifier beg far less than that required to cause significant ga compression (e.g., a room temperature source). However, a few unusual cases, e.g., if the mixer saturates at a very low put power, or the IF preamplifier is very noisy, it could be necessary to separate the noise of the IF stages from that of the mixer and source analyzg the saturation. -6-

7 Square-Law Detector vs Power Meter Our itial measurements of saturation were made usg an IF power meter with the RF chopper wheel runng slowly (several seconds each position), but ga drift the room temperature IF amplifiers made consistent measurements difficult to obta. The power sensor was replaced by a tunnel diode detector and the chopper run at ~ 10 revolutions per second (~0 Hz choppg rate); then the drift was much less significant. Square-Law Detector vs Spectrum Analyzer If a spectrum analyzer is used stead of a square-law detector (or power meter), caution may be necessary estimatg the quantities (P HS - P H ) and (P CS - P C ). This is because most modern spectrum analyzers use an envelope detector as opposed to a square-law detector, and the dicated signal power the presence of noise is not simply the sum of the signal and noise powers. If envelope detection is used with the usual log display, the correction factor is 10.4 x (H(dB)) db, where H is the dicated signal-to-noise ratio [4]. This is shown the upper curve of Fig. 6. Fig. 6. Correction for system noise when measurg a CW test signal. The correction factor is plotted as a function of the dicated signal-to-noise ratio H. The upper curve applies to measurements usg a spectrum analyzer with an envelope detector (usual modern spectrum analyzers) and a log (db) display. The lower curve is for measurements usg a power meter with a square-law detector. From [5]. Source Mismatch A possible source of error measurg ga compression occurs if LO reflections from the hot and cold loads are sufficient to modulate the mixer ga at the chopper frequency. The mixer bias current is a good dication of the LO level at the mixer. Separate measurements can be made of the receiver ga and mixer current as functions of LO power, and the resultg ga vs mixer current curve allows the degree of ga modulation due to -7-

8 chopper modulation of the LO power to be estimated. This was not significant the present measurements. REFERENCES [1] R. L. Plambeck, "Receiver amplitude calibration for ALMA," ALMA Memo 31, August 7, 000. ( [] A. R. Kerr, "Saturation by noise and CW signals SIS mixers," Proc. 13 th Int. Symp. on Space Terahertz Tech., Harvard Univ., pp. 11-, 6-8 March 00. ( [3] E. F. Lauria, A. R. Kerr, M. W. Pospieszalski, S.-K. Pan, J. E. Effland, A. W. Lichtenberger, "A GHz SIS mixer-preamplifier with 8 GHz IF bandwidth," 001 IEEE International Microwave Symposium Digest, pp , May 001. ( [4] A. A. Moulthrop and M. S. Muha, "Accurate measurement of signals close to the noise floor on a spectrum analyzer," IEEE Trans. Microwave Theory Tech., vol. 39, no. 11, pp , November [5] A. R. Kerr, S.-K. Pan and J. E. Effland, "Sideband calibration of millimeter-wave receivers," ALMA Memo 357, 7 March 001. ( -8-