**Table 4:** Frequencies and amplitudes derived by a (multi-component) sine fit: variable guide star number (VGS), Guide Star Catalog number (GSC), frequency in mHz, amplitude in mmag
VGS	GSC	freq	ampl
#	#	(mHz)	(mmag)
1	0097301351	0.4417523	2.8164
		0.3704583	1.4488
2	0127000385	0.0989983	19.3090
		0.3066288	4.3830
3	0199501825	0.0505502	0.7653
		0.0011560	49.3609
4	0199502185	0.0182455	0.3170
5	0229301267	0.2572485	5.9219
		0.2927831	4.9748
6	0229301382	0.1110136	20.7980
		0.1987356	7.2022
7	0299600003	0.1978063	18.3684
		0.2403626	3.5217
		0.3337519	4.1131
10	0477400822	0.0132634	84.0723
11	0495601167	0.0621893	27.0865
12	0502500568	0.0375865	1.8812
13	0685403208	0.1362797	155.6936
		0.2725595	43.7994
		0.3975123	20.5502
14	0685601078	0.0516593	8.1490
15	0739300128	0.0122166	18.0785
17	0742602146	0.1373600	20.9450
18	0779501427	0.0076560	2.9164
19	0781801912	0.3195616	1.4698

All frequencies, amplitudes and phases of the variable guide stars are listed in Table 4. The corresponding light curves, fits and phase plots are shown in Figs. 3 and 4. In the remainder of this section, we present detailed information on the guide stars found to be variable.

VGS1, GS0097301351: The light curve was simultaneously fitted with two frequencies (38.16740c/d and 32.00760c/d). The first cycle calls for a slight zero point correction for which, however, no justification can be found. This star is flagged as a variable in the TYCHO catalog and the light curve resembles a multiperiodic $\delta$ Sct star. The SIMBAD archive gives a spectral type of A2, while the TYCHO color suggests A8, which would be commensurate with a $\delta$ Sct variable.

VGS2, GS0127000385: The time base of the observations for this F3 star is rather short and hence a (formal) fit of the data was limited to two frequencies (8.55345c/d and 26.49273c/d). SIMBAD archive and TYCHO color agree in an early F classification which would be consistent with $\delta$ Sct variability.

VGS3 and 4, GS0199501825 and GS0199502185: These two guide stars were observed simultaneously with FGS2 and FGS3, respectively. With a time base of 15.12 hours and homogeneous data, it is possible to study even low amplitude variability. The open squares in Fig. 3 show the mean intensity for each block of data (i.e., HST orbit) and the symbol size corresponds to the standard deviation. For GS0199501825 a simultaneous fit with two frequencies was performed whereas for GS0199502185 a fit with a single frequency seemed to be sufficient (see also Table 4). These stars are examples of cool stars with periods that are too short to be explained by rotation.

VGS5, GS0229301267: For this guide star three different data sets were obtained with FGS1 and FGS2 in 1994, but archived in a telemetry format that increases the noise level. An offset affects the second block of data in set b (see Fig. 3) and we excluded it from further analysis. The reason is unknown, but the intensity drop seems to be of instrumental origin. Two frequencies (22.22627c/d and 25.29656c/d) fit the light curve well. This guide star is also a good example to demonstrate that three individual data sets can have different classifications. Data set a has a time base of 0.22 hours (13.2 min) and hence is classified as "unusable''. The frequency of the highest amplitude of data set b is close to the orbital frequency, and the set was classified as "o4''. The third observation (c) is of better quality and was classified as variable. The two-frequency fit would be consistent with $\delta$ Sct variability, but the TYCHO color indicates F7.

VGS6, GS0229301382: A fit with 2 frequencies (9.59158c/d and 17.17075c/d) was found to be the best solution. This star is clearly variable, but is too hot for a $\delta$ Sct star, if the suggested TYCHO color (A0) is correct.

VGS7, GS0299600003: The two data sets were obtained within 14 days. The first of the ten orbits does not fit the model very well, probably due to instrumental effects. For the frequency analysis this first orbit was ignored. The larger scatter for the middle part of the second set is caused by a different telemetry format, inducing larger noise. This is probably a multiperiodic $\delta$ Sct star, with a TYCHO color of about A5.

VGS8, GS0341102017: This guide star was observed during 1994 and 1995 on six occasions (see Fig. 3, plots a-f) with FGS1. Although the quality of the data is quite good, it was not possible to determine a reasonable sine fit to the data. This may be a cool $\delta$ Sct star.
Note: Data sets d and e are split for this plot, because of gaps of about two days between observations.

VGS9, GS0454701381: The two observations of the eclipsing binary system AZ Cam have been performed during January 25 and February 5, 1994 (see Kuschnig et al. 1997). The light curve is given here for completeness.

VGS10, GS0477400822: A trend with a frequency of about 1.14596c/d is clearly present, with a sudden intensity increase for a small period of time during the sixth orbit. This increase does not happen during a passage of the SAA by the HST (see Zwintz et al. 1999) and it also does not seem to be of instrumental origin. A possible interpretation could be that microlensing due to a planet passage or a flare was detected. This is another example for a K-type star with a period too short to be caused by rotation.

VGS11, GS0495601167: This data set first was formally classified as "t'', but closer scrutiny revealed a clear signal with a frequency of 5.37315c/d and an amplitude of 0.02709 mag. The light curve and TYCHO color would be consistent with a cool $\delta$ Sct variable, but Strömgren indices indicate an F-type spectrum.

VGS12, GS0502500568: This guide star was flagged as "v3'' with a noise level of about 592ppm and it is again a candidate for microvariability. We present the light curve together with the mean intensities per data block (open squares). A fit with a frequency of 3.24747c/d is also shown. TYCHO color as well as Strömgren photometry are consistent with a (very small amplitude) cool $\delta$ Sct variable, but SIMBAD archive gives A3.

VGS13, GS0685403208: VGS 13 is an interesting new variable because its spectral type is listed as M6. It is included in the BMB catalog (Blanco et al. 1984) as BMB224. Strömgren photometry indicates spectral type F, which contradicts the SIMBAD archive classification. Considering the main period and a probable F spectral type, we propose a $\delta$ Sct variability classification. Large amplitudes usually are associated with non-harmonic light curves and result in overtones for Fourier spectra. Hence, we investigated $\rm f1 + 2 \cdot f1$ solutions and found a third frequency close to $3 \cdot {\rm f1}$ to reproduce the observations to first order. Considering the short data set a more detailed frequency analysis does not seem to be justified. Follow-up observations from ground are needed.

VGS14, GS0685601078: The frequency of 4.46337c/d fits the data reasonably well, but with systematic residuals. The number of data points does not allow a more detailed analysis. SIMBAD archive claims that star to be a Mira variable with a period of 218 days, but the FGS light curve looks more like a $\delta$ Sct variable, which would also fit the Strömgren photometry (late A type).

VGS15, GS0739300128: There are two data sets observed in 1992 and 1993, respectively, where the second set has a rather short ( $\sim2$ hours) time base. A frequency of 1.05551c/d fits the data from both years well, but Fig. 4 shows only the observations from 1992. A $\delta$ Sct light curve would be consistent with the TYCHO color and Strömgren photometry (late A), but not with B8 given in the SIMBAD archive.

VGS16, GS0739300524: This star was observed in 1992 simultaneously with GS0739300128. It appears to be a new eclipsing binary.

VGS17, GS0742602146: A fit with the frequency of 11.86790c/d and an amplitude of 21 mmag is shown together with the data, archived with two different telemetry formats. The light curve resembles a multiperiodic $\delta$ Sct variable, which would fit the late A type classification from Strömgren photometry.

VGS18, GS0779501427: VGS 18 is again a candidate for microvariability with a period of about 1.5 days and an amplitude of 3 mmag. The open squares show the mean intensities per block (HST orbit). This may be a cool K type star with a very short period.

VGS19, GS0781801912: The formal classification of this guide star is "o1'', but after a careful stray light correction of the entire data set the lightcurve shown in Fig. 4 remained. The period of about 52 min is close to half of the orbital period of the HST, but definitely seems to be intrinsic. The light curve resembles $\delta$ Sct variability which would fit to the Strömgen photometry, but the TYCHO color is too cool.

VGS20, GS0825203215: All four data sets of VGS 20 have been classified as variable, but unfortunately the last set (d), transmitted with the low noise telemetry format, shows an intensity offset of unknown origin. It was impossible to find a proper fit for all four data sets simultaneously, even after applying a zero point correction for the last set. The last data set resembles $\delta$ Sct variability, but the TYCHO color and even more so the spectral type given in the SIMBAD archive are too cool.

5 Discussion