1. Title Instrument Simulator and L1B Processor Meeting
Swarm CEFI LP
Instrument Simulator and
L1B Processor Meeting
ESTEC, June 7-8, 2006

2. What we expect from this meeting
Understanding of requirements and context
Schedule for IS/L1B development
Definitions of interfaces and who does what

3. Basic principles of Langmuir probes
Bias voltage Vb applied on probe
Resulting current measured
Depends on ne, Te, Vsc, ram speed, and ni and Ti for all ion species
Derive ne, Te, Vsc from probe response
Many variants depending on how bias is varied

4. CEFI LP particulars Normal mode: one probe operated in each gain state
Two Langmuir probes
Two gain states in h/w to cover all parameter range
,000,000 cm-3
Normal mode: one probe operated in each gain state
Allows largest dynamic range

5. LP functionality A: Sweeps
Most detailed information using probe bias sweeps
Vary Vb, record current-voltage relation
Drawbacks:
large data volume
time consuming
e- retardation region
e- “saturation” region
ion “saturation” region

6. LP functionality B: Fix bias
Constant Vb (usually positive)
Current variations depend on ne and Te
Advantage: very high time resolution
Drawback: cannot disentangle variations in ne and Te
Occasional sweeps can be used to calibrate large scale trends in ne/Te, but there is no chance to catch rapid variations

7. LP functionality C: Track
Software can be used to track a constant current, i.e. to vary the bias so as to keep the probe current constant
Allows tracking of Vsc variations at high time resolution
Sheath variations also influence result
Use occasional sweeps to calibrate result

8. LP functionality D: Ripple (1/3)
Rippled Langmuir probes used to measure I and dI/dV simultaneously at a few points on a probe curve
Quasi-sinusoidal dI
Sinusoidal dV

9. LP functionality D: Ripple (2/3)
2nd point:
positive side,
e- saturation
3rd point:
e- retardation
region Vb
1st point:
negative side,
ion saturation

10. LP functionality D : Ripple (3/3)
Ripple technique advantages:
Fast operation
Small data volume
Low noise (averaging) compared to sweeps
If points are judiciously chosen, variations in ne and Te can be disentangled
CEFI LP: Digital implementation in FPGA
Flexibility in design and operations
Resource economy
H/w stability

11. CEFI LP particulars Normal mode: one probe operated in each gain state
Two Langmuir probes
Two gain states in h/w to cover all parameter range
,000,000 cm-3
Normal mode: one probe operated in each gain state
Allows largest dynamic range

12. LP mode concept High-level modes:
Normal
Calibration
Single-submode modes TBD (see below)
Submodes composing the high-level modes:
SM: Sweep mode
ZM: (Zero-)tracker mode
HM: Harmonic (ripple) mode (including Vsc track)
FM: Fix bias mode
Diagnostic modes

13. LP normal mode Normal mode is a combination of fundamental mode blocks
Sweeps (SM) every 128 s
Time between sweeps divided into 0.5 s blocks
In each 0.5 s block:
Track constant current (ZM) for ~0.2 s
Measure I and dI/dV (HM) for ~0.3 s
Resulting data
High resolution ne, Te and Vs from HM
Independent dVs from ZM
Independent dn from probe current (HM)

14. LP HM submode Uses three points One in each (±) saturation region
One in retarded electron region (gives unique ne/Te separation)
3rd point:
e- retardation
region Vb
1st point:
negative side,
ion saturation

15. LP simulator/L1B processor status
Used for design purposes (h/w & s/w)
Probe size
Gain resistors
FPGA filter evaluation
Operational submode sequencing
Algorithms for operational modes
Data analysis algorithms
Error analysis

16. Present simulator s/w (1/4)
Works on one 0.5 s segment at a time
Input for each time segment:
ne, Te, ion species, Ti, ram speed
Instrument setup
S/c potential calculated from input plus a 0.5 V rms random value added
Bias voltage Vb time series generated using specified instrument setup

17. Present simulator s/w (2/4)
Probe current time series calculated from Vb and plasma using Langmuir probe theory (Mott-Smith & Langmuir, 1926)
Combined ram+thermal expression for each ion species (Medicus, 1962)
Sheath model for electrons by Walker (1961)
Photoelectrons by Pedersen (1995)

18. Present simulator s/w (3/4)
Analog electronics
op amp noise (the dominating noise source) added
ADC quantization noise added

19. Present simulator s/w (4/4)
FPGA operations
Full simulation of FPGA code at internal resolution
Mixing and filtering (including correct CIC filters)
NCO (numerically controlled oscillator) as sine table with truncated values
Flight s/w functionality included except timing and ISP production

20. Plan

21. Present L1B s/w No ISP unpacking Analyzes normal mode data
Nominal analysis of HM
Does not yet use ZM
SM implementation needs update
Uses simplified probe equations
Leads to systematic errors...
... which are collected in tables
Output parameters corrected using the tables (TBI)

22. IS/L1B example result: Random errors
Contours indicate IRI parameter range
8 orbits at 450 km
Conditions for 1999 (i.e. 2010)

23. Systematic errors
Systematic errors are less important: correction to be included in analysis

24. Errors in s/c potential
Small errors in important parameter range

25. Plan

26. Plan

27. SM L1B algorithm (1/3) Find Vsc from max(d2I/dV2)
Noise reduction in differentiation
Model of ion and photoemission current:
Use the 75% of the probe curve lowest in potential
Linear least square fit
Positive values replaced by zeros
Subtract modelled current from data

28. SM L1B algorithm (2/3) Get Te from electron retardation part
Non-linear logarithmic least squares fit
Do linear fit above Vsc
Ie(Vb) = e + f Vb
Calculate ne = c sqrt(Te) f
Do a full non-linear fit
Values above used as input
First leave only Vsc free (with necessary coupling to other params)
Finally adjust all parameters

29. SM L1B algorithm (3/3) Correct for systematic errors
Errors from using simplified model, imperfect algorithms and noise
Calibration table determined from simulations
Calibration table updated post- commissioning for identified on-orbit error sources

30. HM L1B algorithm (1/4) Estimate Vsc:
Get line from I and dI/dV for point on positive side
Get a second line from I and dI/dV in electron retardation region
Use intersection of the two lines as Vs

31. HM L1B algorithm (2/4) I Vb
Vsc estimate

32. HM L1B algorithm (3/4) Model of ion and photoemission current:
Use I and dI/dV from point at negative side to derive linear expression
Electron temperature:
Derived from I/(dI/dV) at electron retardation region (intermediate point)
Corrected using linear expression from ion side

33. HM L1B algorithm (4/4) Get ne and Te from di/dV on electron side
ne = c sqrt(Te) dI/dV
Correct for systematic errors (table lookup)
Summary of algorithm:
“Lite version” of sweep analysis
Good Te, ne and Vsc at (better than) 0.5 s resolution
Can fail in extreme situations (highest and lowest ne, highest Te). Fallback solutions:
ZM data is fallback for dVsc
I data from HM is fallback for dne
Both are calibrated using the succesfully analyzed data points just before and after the failed point (HM or SM)
Conflict with ISP-to-L1B data mapping requirement