New PAS Normal mode description
Since FSW 3.4.? we are using a New Normal mode, that allows to produce PAS 3D distributions with a better resolution than 4s.
This new PAS Normal mode is characterized by some parameters:
The DPU software is organized in cycles of 100s, related to the SCET (spacecreaft even clock).
At offset 0 (SCET modulo 100 = 0), the cycle starts with a 3D full-sampling (FS), typically [64 energies x 9 elevations]
Then, a peak-tracking algorithm is used to adapt dynamically the energy and elevation windows of PAS measurements.
The 3D full-sampling parameters are given in a configuration register #3006
ParamID = 3006 : Full3DParId
- normalFull3D.EnergyNumi = 64
- normalFull3D.ElevNumi = 9This FS and peak-tracking algorighm, are using a fixed 4s time interval.
Then, there will be up to N = 93 3D samplings, in the same 100s cycle, with typically [64 energies x 7 elevations]
These normal sampling parameters are given in a configuration register #3005
Param ID = 3005 : StatNormParId (42 bytes)
- NormalSampling.EnergyNum = 64
- NormalSampling.ElevNum = 7
- NormalSampling.K = 1
- NormalSampling.N = 93The main improvement with the new PAS normal mode is characterized by the parameter N.
For each DPU 100s cycle:
-
when N > 1: the DPU generate a full-3D sampling (4s) and N normal sampling at 1s rate, with N = 93 as usual value
-
when N = 1: the DPU generate a full-3D sampling (4s) and 24 normal sampling, also at 4s rate
Another important parameter is K :
When K > 1, in NM :
-
PAS FPGA send K sub-samplings per second
-
DPU software compute and produce the sum of K sub-sampling per second.
When K = 1, the DPU produce an unique 3D sampling per second.
Another parameter is the PAS cadence, that can take 3 levels : NORMAL, MEDIUM, HIGH, where:
-
High : allows PAS samplings at 1s resolution
-
Medium : 2s resolution
-
Normal : 4s resolution
This parameter can be modified by a dedicated TC, part of normal IOR commanding.
It indicates the maximum rate that can be used (to decrease the size of TM allowed by the SOOP planing)
Typically, we are currently using in flight a FS [64x9] + 93 [64x7] samplings at 1s resolution (HIGH cadence).
PAS Burst-mode description
The PAS burst-mode has not changed with the recent FSW update.
The main parameters of PAS Burst-mode are given in register 3009:
Param ID = 3009 : DynBstParId
DynamicBurst.Channeltron = 0
DynamicBurst.FirstEnergy = 4
DynamicBurst.EnergyNum = 90
DynamicBurst.FirstElev = 0
DynamicBurst.ElevNum = 9
DynamicBurst.EnerWinSize = 36
DynamicBurst.ElWinSize = 5
DynamicBurst.K = 4
DynamicBurst.N = 19
DynamicBurst.L = 15The PAS burst-mode intervals have a fixed duration of 300s (5 minutes).
There are divided in L sequences of (1 FS + N sub-sampling) as 1s rate.
Typically L = 15 x (1 + N=19) to cover the 300s interval.
-
EnergyNum and ElevNum are the parameters used for 3D full-samplings [90x9]
-
EnerWinSize and ElWinSize are the parameters used for sub-samplings [36x5]
-
Channeltron can be used to compute the number of channeltron (0 ⇒ 11, 1 ⇒ 9)
More over, PAS 3D samplings can be divided in smaller ones, in order to be generated at higher rate, up to K=4 sub-samplings per second.
So, there can be a sequence of L=15 cycles (1 FS [90x9x11] + 19 x 4 smaller sub-sampings per second [36x5x11]), generated each 20s
The DPU software is using a dedicated memory buffer that allow to register all the 3D generated during this 300s cycle.
So it can receive PAS 3D at a very high rate, register them in real-time in this buffer, and compress them later.
After the end of burst-more cycle:
-
DPU software automatically restart in Normal mode.
-
while in parallel it can compress and send the 300 samplings stored in the dedicated buffer.
PAS Burst mode data size
It can be possible to compute the TM size used by a full PAS burst mode sequence (300s):
size = L * (EnergyNum * ElevNum + N * K * EnerWinSize * ElWinSize) * nb.channeltron * sizeof (int_16)As an example with the values given above:
size = 15 * (90 * 9 + 19 * 4 * 36 * 5) * 11 * 2Total size = 4.781.700 bytes for 300s
That give 15939 bytes/second
Due to limitation of DPU software, only 296 samplings are generated rather than 300 possible.
A more accurate formula should be:
size = (L * EnergyNum * ElevNum + (296 - L) * K * EnerWinSize * ElWinSize) * nb.channeltron * sizeof (int_16)That gives with the values above:
size = (15 * 90 * 9 + (296 - 15) * 4 * 36 * 5) * 11 * 2Total size = 4.718.340 bytes for 300s or 15727.8 bytes/second
What’s new in FSW 3.4.9
The FSW 3.4.9 has been updated to solve a time-shift problem we encounter with the new PAS Normal mode.
Due to the size of PAS 3D packets sent by PAS FPGA, there were some time synchronisation problem between PAS FPGA and DPU software.
The IRAP team has written a new version of PAS sequencer, that split each PAS 3D packet in 6 consecutive SPW sub-packets
The new FSW has to be modified by DPU team to handle the new behaviour.
Test done on SIIS 2024-10-30
We tried the NEW FSW 3.4.9 on SIIS to validate both the new PAS sequencer and DPU software.
During this first test, we tried various PAS configuration, but with compression disabled.
-
FS [64x9] + 93 [64x7] samplings at 1s rate
-
FS [64x9] + 93 [48x5] samplings at 1s rate
-
FS [90x9] + 93 [64x9] samplings at 1s rate
-
FS [90x9] + 93 [90x9] samplings at 1s rate
-
FS [64x9] + 93 [56x7] samplings at 1s rate (witk K=2 sub-samplings summed per seconds)
-
FS [64x9] + 93 [64x7] samplings at 1s rate
Then we entered in burst-mode (BM) :
-
FS [90x9] + 19 x [36x5] x K=4 samplings per second
|
between each configution change, there is a cycle of 100s, with 3D data at 4s resolution. |
A synthetic view of PAS configuration and 3D data rate can be found at:
During this test, without compression, DPU was able to process PAS data at the maximum allowed rate (1/s for NM, 4/s sub-samplings in BM)
However, there was a problem with PAS commanding, and the DPU software could not restart using PAS normal mode after the end of burst-mode cycle.
Summary
The FSW 3.4.9 has solved the time-shift problem encounterd previously with new PAS Normal mode.
It also solved, which was not really expected, a similar time-shift bug that we had in BM and snapshot data.
We had a problem to restart in NM after the BM cycle, but it was a commanding problem, not technical.
The DPU was able to produce PAS 3D samplings as 1s rate, even with higher configuration [90x9]
Test done on SIIS 2024-11-05
We tried the same kind of test than on 2024-10-30, but with compression enabled.
Also limiting the duration of the BM sequence to 300s, which is the maxumum duration allowed (due to the size of the dedicated internal buffer) to ensure a proper restart in NM.
We tried same various configuration of PAS NM:
-
FS [64x9] + 93 [64x7] samplings at 1s rate ⇒ OK
-
FS [64x9] + 93 [48x5] samplings at 1s rate ⇒ OK
-
FS [90x9] + 93 [64x9] samplings at 1s rate ⇒ not possible (only 2s rate allowed)
-
FS [90x9] + 93 [90x9] samplings at 1s rate ⇒ not possible (only 4s rate allowed)
-
FS [64x9] + 93 [56x7] samplings at 1s rate (witk K=2 sub-samplings summed per seconds) ⇒ OK
-
FS [64x9] + 93 [64x7] samplings at 1s rate ⇒ OK
Then we entered in 300s burst-mode using 15 cycles of :
-
1 FS [90x9] (1s)
-
19 x 4 sub-samplings [36x5] per second
And finally automatically restart in NM (100s cycles), using FS [90x9] (4s) + N=93 sampling [64x7] per second
A synthetic view of PAS configuration and 3D data rate can be found at:
Summary
Like during previous test, the new FSW 3.4.9 has solved the time-shift problem encountered previously.
Limiting the duration of BM cycle to 300s as also allowed the automatic restart in NM.
The main difference was due to the compression of NM data:
-
it was not possible to generate N=93 [64x9] samplings at 1s rate, but only 93/2 at 2s rate
-
is was not possible to generate N=93 [90x9] samplings at 1s rate, but only 93/4 at 4s rate
After discussion with Gennaro, it seems that it’s an expected limitation of DPU software (rather than a technical problem)
