[Mp4-tech] time_scale and num_units in tick: what am I missing?

[Glenn Connery]

Okay, so I'm having trouble parsing a seq_parameter_set_rbsp which contains 
vui_parameters, so I'm getting lost in the following bits.  Everything I 
compute agrees with another program up to the timing info.  Here's the MPEG-2 
transport stream packet:
47 41 2c 39 01 60 00 00 01 e0 00 00 82 c0 0a 31 
57 e1 5a 21 11 57 e1 3e 01 00 00 00 01 09 10 00 
00 00 01 67 4d 40 1e 89 99 82 c1 22 c1 bd c1 41 
41 7c 00 00 03 00 04 00 00 03 00 ca 10 00 00 00 
01 68 ea 88 35 c8 00 00 00 01 65 b8 00 20 00 af 
fe d4 a7 99 63 a3 da 16 1b c2 4b b2 86 de 8f 0d 
58 fb 4f 60 9e b0 8e 20 00 00 24 b0 8c 91 50 b2 
7e 90 54 df 21 13 74 7d c0 0b 09 8a 95 d5 81 d5 
8a 32 04 0e a7 35 bb 21 06 9f af 00 02 fe 72 88 
05 7c ac 89 0f b0 68 3e 80 a6 cf 3b 8a 47 87 d6 
4f 1d b4 8d 39 43 72 d0 d7 11 a2 63 55 9a fd 53 
ff e9 68 23 e9 5b 91 7c d3 31 b0 f7             .
which has a PES header (00 00 01 e0), an AU delimiter (00 00 01 09) and then 
the 26-bytes of the SEQ following the 00 00 01 67:
4d 40 1e 89 99 82 c1 22 c1 bd c1 41 41 7c 00 00 03 00 04 00 00 03 00 ca 10 00
Here's my parse, by HAND of this, which agrees with my code.  And agrees with 
this other program up to the timing info.  Oh, by the way, its a PAL sample 
(25fps):
00: 4d  0100 1101
01: 40  0100 0000
02: 1e  0001 1110
03: 89  1000 1001
04: 99  1001 1001
05: 82  1000 0010
06: c1  1100 0001
07: 22  0010 0010
08: c1  1100 0001
09: bd  1011 1101
10: c1  1100 0001
11: 41  0100 0001
12: 41  0100 0001
13: 7c  0111 1100
14: 00  0000 0000
15: 00  0000 0000
16: 03  0000 0011
17: 00  0000 0000
18: 04  0000 0100
19: 00  0000 0000
20: 00  0000 0000
21: 03  0000 0011
22: 00  0000 0000
23: ca  1100 1010
24: 10  0001 0000
25: 00  0000 0000
[ 01001101 ]  profile_idc = 77			// u(8)
[ 0100 ] CF0 = 0 CF1 = 1 CF2 = 0 CF3 = 0	// u(1) u(1) u(1) u(1)
[ 0000 ] Reserved					// u(4)
[ 00011110 ] level_idc = 30				// u(8)
[ 1 ] seq_parameter_set_id = 0			// ue(v)
[ 0001001 ] log2_max_frame_num_minus4 = 8(12) // ue(v)
[ 1 ] pic_order_cnt_type = 0			// ue(v)
[ 00110 ] log2_max_pic_order_cnt_lsb_minus4 = 5(9) // ue(v)
[ 011 ] num_ref_frames = 2				// ue(v)
[ 0 ] gaps_in_frame_num_value_allowed_flag = 0 // u(1)
[ 000010110 ] pic_width_in_mbs_minus1 = 21(22) // ue(v)
[ 000010010 ] pic_height_in_map_units_minus1 = 17(18) // ue(v)
[ 0 ] frame_mbs_only_flag = 0			// u(1)
[ 0 ] mb_adaptive_frame_field_flag = 0		// u(1)
[ 1 ] direct_8x8_inference_flag = 1		// u(1)
[ 0 ] frame_cropping_flag = 0			// u(1)
[ 1 ] vui_parameters_present_flag = 1		// u(1)
[ 1 ] aspect_ratio_info_present_flag = 1 	// u(1)
[ 0000 0110 ] aspect_ratio_idc = 6 		// u(8)
[ 1 ] overscan_info_present_flag = 1		// u(1)
[ 1 ] overscan_appropriate_flag = 1		// u(1)
[ 1 ] video_signal_type_present_flag = 1	// u(1)
[ 101 ] video_format = 5				// u(3)
[ 1 ] video_full_range_flag = 1			// u(1)
[ 1 ] colour_description_present_flag = 1	// u(1)
[ 0000 0101 ] colour_primaries = 5		// u(8)
[ 0000 0101 ] transfer_characteristics = 5	// u(8)
[ 0000 0101 ] matrix_coefficient = 5		// u(8)
[ 1 ] chroma_loc_info_present_flag = 1		// u(1)
[ 1 ] chroma_sample_loc_type_top_field = 0		// ue(v)
[ 1 ] chroma_sample_loc_type_bottom_field = 0	// ue(v)
[ 1 ] timing_info_present_flag = 1			// u(1)
[ 0000 0000 0000 0000 0000 0000 1100 0000 ] num_units_in_tick = 192	// u(32)
[ 0000 0001 0000 0000 0000 0000 0000 0000 ] time_scale = 16777216  // u(32)
... 11 0000 0000 1100 1010 0001 0000 0000 0000
Now the other program says the num_units_in_ticks is 1 and the time_scale is 
50.  So 50 fields per second, which is obviously correct.  And my values make 
no sense at all.  But from what I can see the spec is straight-forward--these 
two values are simply u(32)'s, with no masking or other interpretation of any 
kind.
What am I missing?  Was there some kind of addendum to the spec since the 
version I'm looking at?  Or some other text I'm missing?  
Thanks for any help,
Glenn