-
Notifications
You must be signed in to change notification settings - Fork 1
/
memo_laser_design_scripts.txt
217 lines (129 loc) · 15.4 KB
/
memo_laser_design_scripts.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
2023-09-22T11:59
Here was my message to Tsung-Han:
Hi Tsung-Han,
Here's my attempt at a compact summary lol. In the plots above, I made three cutbacks, pushing the round trip dispersion from 1.89 -> 1.60 -> 1.31, with the rep-rate going from 113 -> 115 -> 117 MHz. I couldn't get mode-locking at 111 MHz, which was the cutback for 113 MHz.
After 117 MHz, I accidentally broke the passive fiber, and so had to jump to 122 MHz, where I lost mode-locking. After a few iterations, I ended up with the gain fiber at 67.5 cm, with a rep-rate of 130 MHz. I thought this should work since this corresponds to a round trip dispersion of 1.32 which should work given the previous cutbacks. However, I don't see it mode-lock, although it shows the usual fluctuations where it's probably close.
Not sure whether to try mode-locking it again, or to just switch to the ER-110, since that’s the one I would have to use anyways.
Notes:
1. If the ER-80 has D = -20 ps/nm/km, then I get mode-locking with round trip dispersion lower than 2.17 ps/nm/km, which I think is a different range than what Tsung-Han had previously considered?
2. The round trip dispersion <-> mode-locking relation doesn't seem repeatable, in that I mode-locked it at 117 MHz with 75 cm of gain fiber -> DRT = 1.32, but can't get mode-locking with DRT = 1.32 at 67.5 cm of gain fiber and fr = 130 MHz.
3. Anyways, I'm following the method I outlined in the group meeting slides that I plan to present next Friday, but no luck :(.
4. It's possible that gain and dispersion are coupled together. In which case, maybe the only reliable way to get there is:
a. at a lower rep-rate with more leeway on pigtails, confirm that you can mode-lock -> the components are good and directionality of the phase bias is correct.
b. splice in the length of gain fiber that you plan to use, with the pigtails spliced to the gain fiber already cut down to length. Then cut back the passive fiber until you mode-lock it at the rep-rate you want.
------------------------------
2023-10-06T12:25
Tsung-Han achieved mode-locking using the ER 80 normal and anomalous, almost half and half, which means round trip dispersion is like D=4.
A little unusual though if round trip dispersion is supposed to be super anomalous, because I cut back to mode-lock, with mode-locking hitting at around D = 1.89.
He says that he also achieved it using the ER 110, with lengths around what I used 35 - 40 cm, which is aroudn what I used, so I'm not sure why mine does't work, he might have estimated numbers.
I think the takeaway is that maybe you need to be quite a bit more anomalous. Like start out at D=5 ps/nm/km
------------------------------
starting gain:
The amount of pump light absorbed in each slice is directly proportional to the population inversion in the slice
----------------------------------------------------------
2023-10-09T16:33
I am having trouble accounting for gain saturation, in that I don't know how to implement it. I also don't know how to implement 110 dB/m absorption at 1530 without having that be an extremely high number...
Also, how do you account for the fact that if you have a strongly peaked spectrum, the narrow portion of it can eat out all the gain?
2023-10-10T11:31
sure there's all this weird stuff, but at the end of the day you only know so much empirically... so you must be using a much simpler model, and it can't be that complicated.
organizing:
My hunch is that if I can understand chapter 1, I can model EDFA's together with what I already know for NLSE, and using the absorption coefficients I got from Ansys.
1. three level scheme
2. accounting for stark splitting in three level scheme
3. propagation
4. accounting for ASE
4. simple case of a step doping, solve full numerical equaton
5. density matrix representation
6. inhomogeneous broadening
Notes:
- the rate equations are all good, but they are not useful if in the end they're all replaced with the experimental absorption and emission coefficients...
- jump to generalized rate equations for pump, signal and ase
2023-10-13T14:23:
I now have forward and backward pumping or both modeled. But only for forward seeding. I now want to allow counter-propagating pulses. So far I have the models running separate of each other. Now I need to make it so that the model also adds the pulse energy of the opposite propagating direction from the previous run.
----------------------------------------
2023-10-14T16:14
Optizone email
Hi Ana,
Thanks, I think just the updated quote would be great! We are working out the dispersion management for a laser cavity. As far as the fiber components are concerned, they are good so long as they have fairly low insertion loss. And I think the shorter fused splitter lengths would help with the short splices.
We've about run out of splitters and phase biases for trying new lasers. We thought 10 of each may be too many for prototyping, so I figured I would just order 5.
Thank you,
Peter
Nlight email
- thank you very much, the gvd curves are very useful to have!
- Tsung-Han, cc'ed here, has kindly given us some of the fiber that we asked them to quote us (the ER 110 4/125 and ER 80 8/125). We wanted to purchase more so we don't have to brorrow his.
- However, Tsung-han has had some difficulty using this fiber to achieve high enough gain while balancing dispersion in a figure-9 laser cavity at our target 200 MHz.
- can we reduce the quote to the minimum quantity of 5m? Sorry for the reduction, but we just want a "prototype" quantity for figuring things out.
- can we get the dispersion curve for ER 80 4/125? This is the active fiber he uses to make the lower rep-rate 100 MHz oscillators that we know work very well.
- I'm able to simulate the rate equation and nonlinear pulse propagation, but am missing a few parameters
- besides the simulated dispersion curves are they able to provide the doping info?: curves for absorption and emission cross-section (I'm guessing similar to the attached curve), ER 3+ doping concentration (# / m^-3), and excited state lifetime (I'm guessing about 1 ms?). I can pull default values from elsewhere but would like to simulate as close to the actual fiber as possible.
2023-10-16T13:32
I think I may have had a misunderstanding! The absorption and emission cross-section is kind of per ion! So, it doesn't vary with dopiong concentration (at least not in an ideal world?) I think the lifetime might change because the closer packed they are, maybe they can start to interact.
The calculation for the ion doping is actually pretty straightforward, look at script-3.py, and you can confirm it with simulation (I mean it's just an exponential decay anyways).
However, the corresponding amplification curve doesn't look right! I'm not sure why...
What I can think of is that you haven't taken into account ASE and ESA. Again, for ASE, I can't see why this would be a dominant process unless you seed with very little power. I think ESA is probably what's responsible?
----------------------------------------
2023-10-18T12:31
I wrote the code to do forward and backward pumping, if you provide a direction=1.0 or direction=-1.0, which works just as a multiplier to the output of dpdz. then I modified the code to propagate both a forward and backward pump, and i got some slightly different results when i tested conditions that were the same as before (e.g. only forward pumping). weird
Okay, I reverted back to where I was just using a direction multiplier and tried again from there. I decided to do this rather than debug. To validate, I ran two python consoles: one running the old code with the direction multiplier, and one running the new code which propagates both the forward and backward pump. I had the two console execute scripts testing the same conditions: only forward pumping and only backward pumping. The results are the same :) Happy now???
before, there were some differences in the spectrum and i think the output power. definitely the spectrum. anyways. you can debug that, or just go back to the point before you screwed up, try again and see if the screw up is still there. sometimes that's more effective than debugging you know. i am not about to check out old reverts to figure out what went wrong. point is it works now!!!!!!
----------------------------------------
2023-10-18T13:59
Hi Peter and Tsung-Han,
Thank you both for reaching out! I still have also Tsung-Han’s email on my todo list for responding. I will try to do so soon (= next couple of days).
Attached you find the simulated dispersion curve of the Er80-4/125-HD-PM fiber. Please note that the actual dispersion in your fiber may slightly deviate depending on the core diameter. But I expect it to be within about +/-2ps/(nm*km) of the listed values.
Attached you find the absorption and emission cross section data representative for our Er doped fibers.
The doping concentration is about 56e24 ions/ m^3 with a core diameter of 3.0um.
In Er doped fibers the upper state lifetime is closer to about 9ms.
As stated in the description, the unit price of line #1 of each quote is also valid for 5m order length. Also the 20% discount applies to that unit price at 5m. Or do you prefer a specific quotation that has the 5m quote only?
Best regards,
Steffen
----------------------------------------
2023-10-18T14:04
I don't think you will recover the power conversion efficiency until you account for ASE. I think it's true that ASE doesn't contribute a significant portion of the output light, but it can eat out amplifier gain somewhere in the stretch of gain fiber. So, even if the output ASE is negligible compared to the output signal power, it needs to be incorporated to accurately reproduce the experimental amplifier gain!
Actually, I'm not sure...
----------------------------------------
2023-10-24T11:37
- current parameters are: gamma edfa = 10, D_l = -3 -> -.5, straight section is 8 cm long, pum power = 40 mW
- all of my spectra are peaked on the long wavelength side, whereas I recall all of the experimental stuff was peaked at the short wavelength side... weird. Maybe it has to do with cavity loss? That could be the next thing to try.
do one more read through to look for mistakes!
----------------------------------------
2023-10-25T16:16
You can do a massive parameter space search, but I'm not very inclined. Here are some notes for the 200 MHz figure 9 cavity:
Notes to take from stepping parameters:
- the lower the loss, the more anomalous the cavity needs to be to keep the same straight section
- the longer the straight section the more anomalous the cavity needs to be
- If quite anomalous, the more "slammed" the spectrum is to the 1600 nm side, which can be strongly peaked. At some point, it can't stabilize. Closer to ZDW the spectrum looks better, but the instability becomes worse below some D value, below which the pulse cannot stabilize
- this means that some loss is actually better!
- because otherwise you need to be very anomalous, and the very anomalous spectrum doesn't look good.
- 0.7 - 1 dB loss is good, I think that's typical. With round trip dispersion between 6 - 7, and
- too much loss means the cavity needs to operate too close to ZDW, at some point it never stabilizes.
- if you assume say 0.8 dB splicing and insertion loss across all devices (excluding pm1550 -> pm1550), then the maximum straight section length falls at around 15 cm (preferably less). A number it seems to like is 11 cm with D=6 round trip.
- not totally sure what the effect of changing nonlinearity is. It seems like gamma_edf can take any value between 4 - 9, and it will still mode-lock but at a different threshold and with a different spectrum.
- it claims that it can mode-lock with 28 cm of edf.
- I get that you can use the ER 80... but you have to get 500 mW into the gain fiber, whereas with ER 110 you can be at 300 mW in the gain fiber. With loss taken into account, that means you need to output close to 800 mW, which I think might be more than what I recall measuring with the current driver turned to full. So in other words, the ER 80 is really pushing it, possibly not doable?
----------------------------------------
2023-10-27T15:52
validation: The forward propagation matches the re_only (rate equation only) exactly. Then, I calculate the backward propagation using only forward propagation code + optimization to guess the value of the output pump. This value is back propagated using the forward propagation code. The output matches the results from using the iterative shooting method. This validates the iterative shooting method by matching it to the forward propagation code which is matched to the re_only code. So, granted that the re_only code is correct, the whole chain is validated :) !!!!
The above is done using re_only_5level.py, 5level_amplifier.py and scratch_2.py
-----------------------------------------------
2023-11-02T11:17
I realized I was missing something in the above. However, now I solve it exactly the same way as Barmenkov et al. and again I ran the validations.
2023-11-02T11:17
splicing ER normal to anomalous, 0.7 dB splicing loss between the two edf's, but no splicing loss from anomalous to pm1550. I get that the cavity dispersion needs to be much more anomalous than before. And, the pump current ends up being about the same... This is with 40 cm total, however. I get the same spectrum at 500 mW 8 ps/nm/km as 400 mW 4 ps/nm/km when only using the ER 110
I think the natural thing to try next is to increase the amount of anomalous fiber.
2023-11-02T11:49:
I tried 50 cm total of edf, and there isn't a regime where the pulse can stabilize ...
2023-11-02T12:18:
I had previously tried 40 cm of edf, and so I think I'll try 35 cm now and see if I can mode-lock with lower current... I'm thinking increasing it was probably the wrong direction? There should be an optimum length of anomalous gain fiber.
2023-11-02T15:33
Things are a lot better if I remove the splicy loss between the 8 um core edf and pm1550. 40 cm is then the best. 50 cm I think I had tried with the splice loss removed, but I double checked briefly just now without a huge range of values and it wasn't stable. I tried 45 cm with a wider range of values pushing normal and anomalous and it also isn't stable... So, 40 cm is around the max? +- less than 5cm.
-----------------------------------------------------------
03-09-2024
current problem:
The splice simulation doesn't match the no splice simulation in the limit that one of the fiber lengths go to zero. Okay I have like 1mm of it, but seriously that shouldn't matter. I already set the splicing loss to 1.0
actually i'm not sure if i ran the sanity check test correctly.
Okay yeah the convergence is still quite slow compared to the other scenario
okay, that's because in scratch 3 i was testing out what happens if you account for the pigtails before the gain fiber, so that the gain fiber is not totally on one side only...
i didn't think it made sense to include t_shock and raman in some fibers but not the others. So, i removed it from all of them haha.
okay so everything is matched now to the "simplest" case of how I set up the cavity, with t_shock and raman turned off. And the only additional factor i'm accounting for is multiple gain fibers.
also the way I have it set up, it's easy to calculate for any combination of gain fibers, including combining normal and anomalous gain. I think the simulation in the past was off there, since I know Tsung-Han got it to work pretty well!