Commit ·
bca5b64
1
Parent(s): eada2c7
Added citation section, updated directory tree structure diagram, plus small cosmetic changes
Browse files
README.md
CHANGED
|
@@ -9,15 +9,15 @@ tags:
|
|
| 9 |
---
|
| 10 |
# Simulated 1 m multimode-fibre transmission matrices
|
| 11 |
|
| 12 |
-
5,000 simulated transmission matrices (TMs) of a 1 m step-index multimode fibre with random bends, generated with the transfer matrix method (TMM), plus the matrices that convert them to the pixel basis. They were used to pretrain the autoencoder in the code repository
|
| 13 |
|
| 14 |
## Simulated fibre
|
| 15 |
|
| 16 |
| | |
|
| 17 |
|---|---|
|
| 18 |
-
| Fibre | step-index, core diameter 50
|
| 19 |
-
| Length |
|
| 20 |
-
| Wavelength | 632.
|
| 21 |
| Guided modes kept | 386 (scalar LP modes, one polarisation, one azimuthal orientation) |
|
| 22 |
| Bends | random bend radii and directions, saved alongside each TM |
|
| 23 |
| Filter | only TMs with a 2-norm condition number below 250 were kept |
|
|
@@ -25,16 +25,18 @@ tags:
|
|
| 25 |
## Contents
|
| 26 |
|
| 27 |
```
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
|
|
|
|
|
|
| 35 |
```
|
| 36 |
|
| 37 |
-
Each `TM*.pkl` is about 1.
|
| 38 |
|
| 39 |
## From a mode-basis TM to a pixel-basis TM
|
| 40 |
|
|
@@ -44,7 +46,7 @@ With `u_in` the input patterns (flattened to a column per pattern):
|
|
| 44 |
TM_pixel = dist_pim_to_real @ tm_mode @ prox_real_to_pim @ u_in
|
| 45 |
```
|
| 46 |
|
| 47 |
-
For a composite
|
| 48 |
|
| 49 |
## Loading
|
| 50 |
|
|
@@ -59,15 +61,29 @@ with open("TransferMatrixMethod/TM0000.pkl", "rb") as f:
|
|
| 59 |
|
| 60 |
## Related
|
| 61 |
|
| 62 |
-
- Code, weights and splits
|
| 63 |
-
- Experimental TMs and RMs of a 5 m fibre
|
| 64 |
|
| 65 |
-
## Acknowledgements
|
| 66 |
|
| 67 |
The TMM simulation method and the original MATLAB fibre simulation code were developed by David B. Phillips and Robert J. Kilpatrick.
|
| 68 |
|
| 69 |
Licence: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).
|
| 70 |
|
| 71 |
-
Citation: *to be added when the preprint is online.*
|
| 72 |
-
|
| 73 |
Contact: Callum J. Wilson, cw2011@cam.ac.uk
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
---
|
| 10 |
# Simulated 1 m multimode-fibre transmission matrices
|
| 11 |
|
| 12 |
+
5,000 simulated transmission matrices (TMs) of a 1 m step-index multimode fibre with random bends, generated with the transfer matrix method (TMM), plus the matrices that convert them to the pixel basis. They were used to pretrain the autoencoder in the [code repository](https://github.com/callucallum/mmf-tm-pca-autoencoder), where five 1m TMs are multiplied to form one simulated 5m "composite" TM.
|
| 13 |
|
| 14 |
## Simulated fibre
|
| 15 |
|
| 16 |
| | |
|
| 17 |
|---|---|
|
| 18 |
+
| Fibre | step-index, core diameter 50µm, NA 0.22, refractive index 1.457 |
|
| 19 |
+
| Length | 1m |
|
| 20 |
+
| Wavelength | 632.8nm |
|
| 21 |
| Guided modes kept | 386 (scalar LP modes, one polarisation, one azimuthal orientation) |
|
| 22 |
| Bends | random bend radii and directions, saved alongside each TM |
|
| 23 |
| Filter | only TMs with a 2-norm condition number below 250 were kept |
|
|
|
|
| 25 |
## Contents
|
| 26 |
|
| 27 |
```
|
| 28 |
+
mmf-tm-simulated/
|
| 29 |
+
├── TransferMatrixMethod/
|
| 30 |
+
│ ├── TM0000.pkl ... TM0999.pkl TMs made by the CPU script
|
| 31 |
+
│ ├── TM1000-GPU.pkl ... TM4999-GPU.pkl TMs made by the GPU script
|
| 32 |
+
│ ├── dist_pim_to_real.pkl distal-facet matrix: propagation-invariant modes (PIMs) -> pixel basis
|
| 33 |
+
│ ├── prox_real_to_pim.pkl proximal-facet matrix: pixel basis -> PIMs
|
| 34 |
+
│ ├── fibre_metadata.pkl parameters of the simulated fibre
|
| 35 |
+
│ └── orthonormal_random_inputs_371x371_145.npy the 145 random orthonormal input patterns (371 x 371)
|
| 36 |
+
└── README.md
|
| 37 |
```
|
| 38 |
|
| 39 |
+
Each `TM*.pkl` is about 1.2MB and holds the TM in the 386-mode basis under the key `tm_mode`. All `.pkl` and `.npy` files are stored with Git LFS.
|
| 40 |
|
| 41 |
## From a mode-basis TM to a pixel-basis TM
|
| 42 |
|
|
|
|
| 46 |
TM_pixel = dist_pim_to_real @ tm_mode @ prox_real_to_pim @ u_in
|
| 47 |
```
|
| 48 |
|
| 49 |
+
For a composite 5m fibre, `tm_mode` is replaced by the product of five 1m TMs in propagation order. Measurement noise is not stored in the files; the code repository adds it during training. See `training/` and `src/` there for the exact procedure.
|
| 50 |
|
| 51 |
## Loading
|
| 52 |
|
|
|
|
| 61 |
|
| 62 |
## Related
|
| 63 |
|
| 64 |
+
- [Code, weights and splits](https://github.com/callucallum/mmf-tm-pca-autoencoder) (GitHub)
|
| 65 |
+
- [Experimental TMs and RMs of a 5 m fibre](https://huggingface.co/datasets/OPTIMlab/mmf-tm-rm-experimental) (Hugging Face)
|
| 66 |
|
| 67 |
+
## Acknowledgements and licence
|
| 68 |
|
| 69 |
The TMM simulation method and the original MATLAB fibre simulation code were developed by David B. Phillips and Robert J. Kilpatrick.
|
| 70 |
|
| 71 |
Licence: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).
|
| 72 |
|
|
|
|
|
|
|
| 73 |
Contact: Callum J. Wilson, cw2011@cam.ac.uk
|
| 74 |
+
|
| 75 |
+
## Citation
|
| 76 |
+
|
| 77 |
+
If you use this dataset, please cite it.
|
| 78 |
+
|
| 79 |
+
```bibtex
|
| 80 |
+
@dataset{wilson2026mmftmsimulated,
|
| 81 |
+
author = {Wilson, Callum J. and Gordon, George S.D.},
|
| 82 |
+
title = {Simulated 1m multimode-fibre transmission matrices},
|
| 83 |
+
year = {2026},
|
| 84 |
+
publisher = {Hugging Face},
|
| 85 |
+
version = {v1.0.0},
|
| 86 |
+
doi = {TODO},
|
| 87 |
+
url = {https://huggingface.co/datasets/OPTIMlab/mmf-tm-simulated}
|
| 88 |
+
}
|
| 89 |
+
```
|