Pergamon Adv. Space Res. Vol. 14, No. 10, pp. (10)695--(10)699, 1994

Copyright @ 1994 CO,SPAR Printed in Great Britain. All rights reserved.

0273-1177/94 $7.00 + 0.00

A COMPARISON OF TWO DIFFERENT TYPES OF GEOSYNCHRONOUS SATELLITE MEASUREMENTS DURING THE 1989 SOLAR PROTON EVENTS

E. Normand

Boeing Defense and Space Group, Seattle, WA 98124-2499, U.S.A.

ABSTRACT

The proton telescope aboard the GOES-7 satellite continuously records the proton flux at geosynchronous orbit, and therefore provides a direct measurement of the energetic protons arriving during solar energetic particle (SEP) events. Microelectronic devices are susceptible to single event upset (SEU) caused by both energetic protons and galactic cosmic ray (GCR) ions. Some devices are so sensitive that their upsets can be used as a dosimetric indicator of a high fluence of particles. The 93L422 1K SRAM is one such device. Eight of them are on the TDRS-1 satellite in geosynchronous orbit, and collectively they had been experiencing 1-2 upset/day due to the GCR background. During the large SEP events of 1989 the upset rate increased dramatically, up to about 250 for the week of 19 Oct, due to the arrival of the SEP protons. Using the GOES proton spectra, the proton-induced SEU cross section curve for the 93L422 and the shielding distribution around the 93L422, the calculated upsets based on the GOES satellite data compared well against the log of measured upsets on TDRS-1.

INTRODUCTION

To monitor and record the environment of energetic protons (>10 MeV) in space, a number of satellites are equipped with instrumentation specifically dedicated for this purpose. These are mainly proton telescopes and examples of satellites that contain such instruments are GOES and IMP-8 (the CPME - Charged Particle Measurements Experiment and GME - Goddard Medium Energy experiment instruments).

One of the important uses of proton telescopes is to monitor the protons emitted during a solar energetic particle (SEP) event. During an SEP event the flux of protons emitted from the sun increases many orders of magnitude above its value during quiet times. Most other satellites do not contain instrumentation to directly measure protons. Nevertheless they do contain microelectronics, some of which are susceptible to single event upset (SEU) caused by both the SEP protons and the galactic cosmic rays (GCR). SEU is the change of a logic state in a microelectronic device, (e.g. a memory) caused by the deposition of energy by a high energy particle, e.g. a heavy ion from the GCR or recoils caused by interaction of energetic protons with atoms in the microchip.

Some devices are so sensitive to SEU that their upsets can be used as a dosimetric indicator of a high fluence of particles. Sometimes this happens by design, i.e., an extremely sensitive device is purposely put on a satellite to serve as a response rate meter. An example is the Intel 2164A 64K dynamic random access memory (DRAM), ten of which were flown as part of the Rate meter experiment on the CRRES (Combined Release and Radiation Effects Satellite) spacecraft/1/. The experiment was designed to provide upset data which would characterize the various regions of space (inner radiation belt and deep space, since the orbit had an apogee of 36,000 km and a perigee of 360 kin) according to their SEU risk.

In other cases devices may have been incorporated into the design of a satellite without a full appreciation of their sensitivity to SEU. An example of one such device is the 93L422 static random access memory SRAM, eight of which were on the TDRS-1 (Telemetry and Data Retrieval Satellite) Satellite in geosynchronous orbit. Many other devices that have experienced single event upset onboard satellites have been reported in the literature/2,3/.

(10)695

(10)696 E. Normand

1989 SOLAR-PROTON EVENTS

The large SEP events of 1989 provide a unique example that allows the data from two different types of geosynchronous satellite measurements to be compared. In the first case, hourly proton data from the GOES-7 satellite operated by the Space Environment Laboratory of NOAA, /4/ is available for the three main SEP events of that year, 12 Aug, 30 Sept and 19 Oct. In addition, we have available the logs of the RAM upsets experienced by the TDRS-1 satellite during these three events,/5-7/. In particular, we have daily RAM upset data for the Aug, Sept, and Oct SEP events which allows the upset time history to be compared against the GOES-7 hourly proton fluxes/8/. To accomplish this, we calculate the number of upsets by using the measured GOES-7 proton fluences for each day and integrate them with the proton upset cross section mode l /9 / fo r the 93IA22 SRAM. Additional details of the calculation such as accounting for the shielding afforded by the satellite structure and attributing the occurrence of upsets to the relevant portion of the SRAM are elaborated in the following sections:

CALCULATION OF NUMBER OF UPSETS

The number of upsets induced by the differential proton fluence dJ/dE, is given as/10/

Upset = I ~(Ep) dJ/dE (Ep) dEp (1)

where o = Proton upset cross section (cm2/device or cm2/bi0

d J/dE = differential proton fluence (p/cm 2 MeV)

Originally a single parameter model was developed for the cross-section, the so-called Bendel Model /10/. It was improved upon by introducing a second parameter,/9,11/where one parameter relates to the energy threshold at which the cross section goes to zero, and the second relates to the asymptotic value of the cross section. The cross section is given, in terms of the proton energy E and in units of upset cm2/bit, as:

o = (B/A) 14 [1-exp(-0.18 * y1/214 x 10 -12 (2)

where Y = (18/A) .5 * (E-A) (3)

For the 93L422, A=9.39 and (B/A) = 1.3578 and Figure 1 shows the cross section based on equation (2) as well as the original measured values. There is noticeable scatter in the data so that the fit is off an average of about 40%/9/, but the fit is optimized in a least squares sense.

SOLAR PROTON DIFFERENTIAL FLUXES FROM GOES-7 DATA

The tabulation of GOES-7 integral and differential proton f luxes/4/was utilized to analyze the three large SEP events, in Aug, Sept and Oct of 1989. Five channels were used, those with average proton energies of 10, 30, 50, 60 and 200 MeV. GOES data from the highest channel, 640-850 MeV, has also been recorded but with a different type of detector. Since there is some uncertainty as to how best merge the data, this higher energy proton data is not yet available/12/. Plots of the differential proton fluence for the entire 12 Aug, 29 Sept and 19 Oct, events are shown in Figure 2/4/ .

The hourly data i n / 4 / o n l y provides the integral fluence J(>E) rather than the differential fluence dJ/dE. Since the dJ/dE for the 1989 SEP events are so smooth, a single fit could be used over the entire spectrum. For the single fit we used that favored by McGuire/13/, which is based on the non- relativistic limit of a simple stochastic acceleration model for solar particles suggested by Forman, et al./14/. In this limit, a Bessel function expression for the differential flux simplifies to:

dJ/dE = KE3/8exp[-(E]a)l/4] (4)

where K and a are constants. We have integrated this expression to obtain the corresponding relationship for the integral flux:

J(>E)=4Kalt/8F[11/2, (E/a)l/4l (5)

where 1-'[11/2, (E/a) 1/4] is the incomplete gamma function/15/. For the 12 Aug, 29 Sept and 19 Oct, 1989 events, we used equation (5) to obtain the K and a constants used in the equation (4) relationship for dJ/dE. In calculating the K and a constants the J(>50) and J(>100) fluences were

Comparison of Geosynehronous Satellite Measurements (10)697

used. Figure 2 also shows the differential fluences for the three events as calculated using equation (4). Equation (4) provides a good fit out to about 150 MeV, after which it significantly under- predicts the differential fluence. As expected in the important region 50<E<100 the fit is very good and it is also good for E<50 MeV except for the 29 Sept SEP event for which it is overly conservative. Equations (4) and (5) are used to convert the dally integral fluences, J(>E), to dJ/dE for the three SEP events.

I .OOE-IO ,

i • I •

_o

J O

A=9.Og, B=12.751

I~(E-12 . . . . . . . . ,_ . . . . . . . i I

10 100 1000 Proton EnwW, MeV

Fig. 1. Proton-induced Upset Cross Section for the 93L422 SRAM as a Function of Proton Energy.

1.00E+10

1.00C-.~

1.00E,~

1.00Ed)7

1.00C-~

1.01~05

Fig. 2

• 12 Aug RamData

• 29 Sep Rare-Data

• 19 Oct Rare-Data

- - - -~ - - - 12Aug Rare-Pa

----41--- 29 Sept Rare-Rt

- . | . q l - l ¢ •

1 . . . . . : . . . . -" . . . . -" . . . . i ; . . . . : . . . . . . . . . I 0 50 100 150 200 250 300 350

Proton Energy, MeV

Differential Proton Spectra at Geosynchronous Orbit for Large 1989 Solar Flares

SEU UPSETS IN TDRS - 1 DURING THE LARGE 1989 FLARES

Because the 19 Oct, 1989 solar flare was the largest, it will be dealt with first. The TDRS-1 satellite recorded 249 known upsets in its RAM memory during the 19 Oct, 1989 SEP event. We believe that almost all of these were due to the solar protons because, during quiet times, TDRS-1 experiences about one upset every other day due to the GCR. The RAM memory consists of eight 93I.,422 microchips. The A and B parameters for the 2-parameter fit to the SEU cross section for the 93L422 have already been given. In addition to the total number of upsets, we were able to obtain logs of the upsets recorded as a function of time/7/. We also have a probability distribution function of the aluminum shielding surrounding the memory chips on TDRS-1 /16/, based on a ray- tracing analysis with 6000 rays. With these data we were able to compare, on a daily basis, the predicted number of upsets, as a function of shielding thickness, with the number of upsets actually recorded.

Unfortunately for purposes of analysis, the manner in which the TDRS-1 memory is organized significantly complicates the calculation of the recorded number of upsets. The memory is divided into four pages, each comprised of two microchips containing 256 8-bit words. The two chips comprising page 3 serve as spares and so always represent unused memory. The two chips that make up page 2 contain key parameters in 137 of the 256 words which are checksum protected. If these locations experience an upset, the checksum failure flag is raised. The contents of these locations are checked every 50 milliseconds, and the cumulative results of the checksum operations are telemetered to the ground every half second. Page 1 contains temporary parameters, and although page 0 contains important normal mode parameters, the used locations on page 0 are not protected like page 2. Approximately 30% of the memory on these two chips is unused. Thus while we may know of some upsets in pages 0 and 1 in the used locations, we cannot be certain that these were the only ones.

The unused locations on all four pages are periodically dumped for a performance check, and so if an upset occurred since the last check, it can easily be detected by comparison. Usually the performance check is carried out once a week, but during the SEP events it was done every day or two. Thus we have two reliable methods of recording upsets: a) the locations on the page 2 chips which are check sum protected (2 chips) and b) the unused locations on the other three pages (30%, 30%, 100% for pages 0, 1,3 respectively representing a total of 3.2 chips).

Finally, in reviewing the TDRS-1 upset data/7/, it was noted that of the 42 upsets in the used location of page 2, four of the upsets were multiple hits. There were two double hits at about the same time.

(10)698 E. Normaad

Protons do not have sufficient energy to induce double upsets, but heavy ions do. We therefore attribute these double hits to the heavy ion component of the solar energetic particles, and subtract them from the total due to protons. Based on the previous TDRS-1 history, multiple hits due to the galactic cosmic rays are very rare, occurring about once every 4 months/16/. We therefore must conclude that these heavy ions were due to the solar flare and not to the GCR.

After accounting for the measured upsets due to protons in known chip locations, we were able to tabulate the actual number of upsets as a function of time for the 19 Oct, 1989 SEP event. These have been plotted in Figure 3. The 2-chip curve accounts for the 90 upsets on page 2 and the 5.2- chip curve accounts for these 90 plus the 142 upsets in the unused memory locations on pages 0,1 and 3, for a total of 232 upsets. The four page 2 multiple hit upsets due to heavy ions, and the 13 upsets in used locations on pages 0 and 1 have not been used. On this figure we have also plotted the predicted number of upsets using the hourly GOES-7 data/4/, in conjunction with equations (4) and (5), the 2-parameter model for the 93L422, and the shielding distribution function /16/. It should be noted that 50% of the path lengths are through a maximum of 250 mils of aluminum shielding, 70% are through less than 500 mils, and 80% are through less than 750 mils. The figure shows that the best agreement with the measured data is by the 500 mil curve, although the 250 mil curve, for the median shielding thickness, is not too far off.

Uncertainty in predicted number of upsets depends on uncertainties in the upset cross section and in the differential proton fluence. As shown in Figure 1, there is scatter in the measured upset cross section measurements and so the cross section fit has an uncertainty of approximately 40%. Figure 2 shows that there is also uncertainty in using equation (4) to represent the differential flux across the entire proton energy range. The calculations indicate thattmost of the upsets are contributed by protons with E<100 MeV, thus the poor agreement in dJ/dE shown in Figure 2 for E>150 MeV should have only a very small impact on the calculated number of upsets. In summary, for the 19 Oct event, the predictions therefore appear to be in good agreement with the measured data, and indicate that the shielding is an important factor in limiting the number of SEUs due to protons.

~0=

"o

34o, CD

$ 3 0 -

=20, b 0

i c c 10,

z

0

• - - .a - - -Umut0d, 2 a ~

- - ~ - - P m d l d ~ O roll M

I I I I I I I

1 2 3 4 5 6 7

Day ~or start of solar proton event

Fig. 3 Comparison of Measured SEU/Chip on TDRS-1 During 19 Oct. 1989 Solar Flare with Predictions.

25,

.J

m

Lo

! , E z

0

~ edlct S00 tr~ AI

i o I I 1 2 3 4

Day lifter start of solar proton event

3S, n bsmurod, 2 i

s °~" I

O . " - - I

1 2 ; 4 , ; Day after sllrt of so~llr proton event

Fig. 4 Comparison of Measured SEU/Chip on TDRS-1 During 1989 Solar Flare with Predictions.; Upper, graph 4a, 29 Sept Flare; Lower graph, 4b, 12 Aug Flare.

Comparison of Geosynchronous Satellite Measurements (10)699

Logs of the upsets on TDRS-1 during the 12 Aug and 29 Sept, 1989 solar flares are also available /5,6/, along with the proton fluences from GOES-7 for these two events/4/. This data was combined into a plot of the measured daily SEU rate compared to the predicted rate for three shielding thickness in the same manner as for the 19 Oct, 1989 SEP event shown in Figure 3. Figure 4a shows the comparison for the 29 Sept, 1989 event and Figure 4b for the 12 Aug, 1989 event. Again the agreement of the predicted daily upsets with measurements for 500 mils shielding is very good for both of these event.

CONCLUSION

Solar proton integral fluxes during SEP events, as measured by the proton detector on the GOES-7 satellite, were utilized to calculate proton-induced SEU. The satellite data were converted into differential proton fluxes, dJ/dE, which were integrated with the 2-parameter proton SEU cross sections to calculate the expected number of upsets. This procedure was applied to the intense solar energetic particle events of 12 Aug, 29 Sept and 19 Oct, 1989 which produced a large number of single event upsets in the eight 93L422 RAM chips onboard the TDRS-1 satellite. The logs of actual RAM upsets recorded during the three SEP events were analyzed to produce plots of the SEU upset/chip for the first few days of each SEP event. These compared well with the overlaid prediction curves generated using the GOES-7 proton data, the 2-parameter proton upset cross section model and the shielding distribution curve for the chips on TDRS-1.

REFERENCES

1o

2.

.

McNulty P. J., et al,, IEEE Trans. Nucl. Sci., NS-38 1642, (1991)

Adams, L., et al, Proton Induced Upsets in the low Altitude Polar Orbit, IEEE Trans. Nucl. Sci., NS-36 2339, (1989)

Wilkinson, D., et al,, IEEE Trans. Nucl. Sci., NS-38 1708, (1991)

.

.

.

.

.

.

10.

H. H. Sauer, R. D. Zwicld and M. J. Ness, Summary Data for the Solar Energetic Particle Events of Aug through Dec, 1989, Space Environment Laboratory, Boulder, CO, (Feb, 1990)

Contel memorandum, by P. Darling, TDRS-1 Single Event Upset (SEU) Rate Increase on Day 224, (Aug, 1989)

Contel Memorandum by P. Darling, Summary of TDRS-1 Single Event Upsets (SEU) Following the Day 272 Solar Flare, (Oct 1989)

Contel Memorandum by P. Darling, Summary of TDRS-1 Single Event Upsets (SEU) Following the Day 292 (1989) Solar Flare. (Feb, 1990)

Normand E., and Stapor, W. J., Variation on Proton - Induced Upset Rates from Large Solar Flares Using an Improved SEU Model, IEEE Trans. Nucl. Soc. NS-37 1947, (1990)

Stapor, W. J. et al, Two Parameter Bendel Model Calculations for Predicting Proton Induced Upset, IEEE Trans. Nucl. Sci, NS-37 1966, (1990)

Bendel, W. L., and E. L. Petersen, IEEE Trans. Nucl. Soc., NS-30 4481 (1983)

11. Shimano, Y., et al,, IEEE Trans. Nucl. Sci., NS-36, 2344, (1989)

12. R. Zwicld, private communication

13. R.E. McGuire and T. T. von Rosenvinge, The Energy Spectra of Solar Energetic Particles, Adv. Space Res., 4, 2, 117 (1984)

14. M.A. Forman, et al, The Acceleration and Propagation of Solar Flare Energetic Particles. in: The Physics of the Sun, ed, T. Holzer, (1984)

15. M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions with Formulas, Graphs and Mathematical Tables, National Bureau of Standards, (1964)

16. E.C. Smith and T. R. Simpson, Prediction of Cosmic Radiation Induced Single Event Upsets in Digital Logic Devices on Geostationery Orbit, Final Report, TRW Defense Systems Group, (November 1987)

JASR 14: IO-TT