Wed 30 Nov: Meeting (7): Minutes

Read through the technical document and presentation as a group and compiled what is left to do:


TO DO


Presentation: 


- Detector section - Sam will send his part to Natalie once figures have been checked
- Pictures - Do any of the slides need pictures?
- Each learn what you are going to say for a run through on Friday 11.00 before the lecture


Technical Report:


- Draft of the report is done - need to polish the references, contacts page and title page
- Detector section - Sam will fill in the new figures, equations for shot & johnson noise, and any further information on connectors for cable-detector joints
- Final calculations - once the detector parts have been filled in, the tech document will be sent to James to compile the final figures
- Once all of this is done, it will be sent back to Natalie to check and make pretty. 
- Hopefully the full document will be printed and bound by Friday 2 Dec

Fri 25 Nov: Meeting (6): Minutes

What needs to be done


1. Check numbers for detector

2. Need a cost or the optical cable

3. Optical fibre - detector connector

4. Something that distinguishes ours - think of any extras & research them

5. Technical Document (working on this weekend)

6. Presentation Template (for monday)

Wed 23 Nov: Meeting (5): Minutes

1. Discussed final costs


Splicing ~ £500


Detectors ~ £115 (Note: two detectors & delivery & VAT - unable to buy them singularly)


LED ~ £420


Fibre Optic Cable { ~£40,000 w/ Kevlar plating OR ~£20,000 w/o Kevlar}


These final costs seemed reasonable and only a few small bits of information are needed until this is complete.


2. Questions


(1) Kevlar - not necessary, consider just a coating (brightly coloured)
- ask manufacturer for costs w/ coating


(2) Detector to optical fibre connector - consider it and costs (get in touch with manufacturer)


(3) Lengths of cables - email the manufacturers for actual lengths as specs on website seem a bit extreme


3. What to do for FRI 25 NOV (evening at the latest)


Create a document for your section including
1. Manufacturer and spec's
2. Losses associated with your section
3. Costs of each part


Remember to include ALL websites and sources
The technical document may include print screens from the websites and pictures so if you have any particular ones that you want included also send these - if your not bothered or if there is only one then just send the links.
If you have any invoices send these - will be included in the appendix
Check all UNITS!
Also send any spreadsheets. The document itself should include all final figures.


This is so that Natalie can start to put together the technical document over the weekend - at the latest need everything Friday night! We have a meeting on friday so sure this won't be a problem.

Tue 22nd Nov: Optical Fibre Info (2)

I apologise if some of this (and the pictures I have used this week) seems familiar to what I did last week. What I have done this week is taken the additional knowledge I have gained and applied it to better explain some of the concepts and redo some of the calculations to provide more accurate answers.


(I am not entirely sure why this post is all in capitals but something seems to have happened when publishing..)

Taking the statement I made in the previous week

So since we are restricted to using LEDs the best we can use so far seems to be a 4-core (full duplex) OM2 multimode graded index loose optical fibre. (Loose Tube C S T Armoured 50/125 LSOH).

It has come to my attention that we do not need to receive signals, merely transmit them, so we do not need a full duplex system but rather a half duplex one instead.

The reason why I have chosen 4 core is because when considering optical fibres for communications over lengths of 45km I get the feeling it is quite unusual to request a single core. Nearly all of the single core optical fibres that I have found do not come armour plated or have very good resistances to temperature or moisture. However, I have had an idea, instead of only using one core and letting the other 3 go to waste, I propose we use them in the following way. Suppose we want to transmit a file of 20Mb over a line that could transmit at a maximum rate of 20Mbps, now if we had a single core we could transmit 20 million bits per second. However, why not split the file we want to send into 4 pieces of 5Mb. Then we can send each of the 5Mb files down their own optical fibre core. As you can see, one 20Mb file will take 1 second to go down one 20Mbps core, or we can send 4 5Mb files down their own core at a speed of 5Mbps to give the same result. However, if we know that each core can handle a maximum speed of 20Mbps why not simply send all 4 5Mb files at 20Mbps down their own core? This mean we can transmit a signal 4 times faster without having to increase the bandwidth (and hence increase the signal dispersion).

I asked Ross and we are not allowed to do this.

I chose multimode because we are severely restricted in what we can choose here due to the parameters of the problem given to us i.e we are not allowed to use single mode fibres.

I chose the graded fibre because it greatly reduces the amount of dispersion, as explained later.

Loose optical fibre:

http://www.premiumline-cabling.com/download/catalog_PL_FO_final.pdf

We need a loose optical fibre because it provides us with very high resistances to temperature and moisture. The gel acts as an insulating layer which prevents the propagation of water into the system. Also, if the cable is flexed or stressed it reduces the damage done to the cores by acting as a cushion. The reason why I want to choose a high armoured (Kevlar) outer shell is because of its high durability and so there is much less chance of rodents/creatures chewing into or getting to the fibre optic core.

I've looked around and found that the company will provide us with rolls of length 2km.

So to reach a distance of 45km we will need 22 splices and no connectors because this system is designed to be secure.

Power Loss Calculation Again

I am now going to redo the calculation for the power loss in the optical fibre because I have more accurate numbers now.

The attenuation of our cable is 0.5dB/km (at a wavelength of 1300nm), the splice loss is 0.02dB/slice (Can you give me a better value than the one I found Natalie?), no connectors, a 0.3dB error for irregularities in the fibre and finally a 3dB safety margin to ensure that the signal has some strength when it reaches the receiving end.

http://communications.draka.com/sites/eu/Datasheets/MMF%20-%20Graded-Index%20Multimode%20Optical%20Fiber%20%2850_125%20%C2%B5m%29.pdf

This gives me so far a total power loss of 25.8dB(=0.384W). So we need to transmit with AT THE VERY LEAST this power. As more accurate values become available I will continue to improve on this number, but for now this is not too bad.

 

Maximum Bandwidth Calculation

What limits are bandwidth is how badly the signal gets "smeared" or dispersed inside the optical fibre as it travels the 45km distance. There are a few forms of dispersion which need to be considered:

Modal Dispersion:

This is the most important one and is due to the fact that light entering the system at different angles will have to travel different distances and hence will take different times to reach the receiver.

This causes the signal to disperse and means that we lose this sharp "high" "low" signal into a stretched out signal which is less obvious and so we lose data.

Graded index fibres greatly reduce this effect but not entirely.

 Material Dispersion:

This is where light of different wavelengths travel at different velocity through the medium. Even though we are using a monochromatic light source there are always errors in how the light is produced giving us a range of wavelengths about the quoted value which is also emitted (i.e the light source has a spectral width).

http://spie.org/Documents/Publications/00%20STEP%20Module%2007.pdf

Where λ_{0} is the wavelength of the light emitted. We also that that n(λ_{0}) is the refractive index of the inner core, n_{g} is the group refractive index and D_{m} represents the material dispersion in picoseconds per kilometre of length of the fibre per nanometre spectral width of the source.

Waveguide Dispersion:

Seeing as how Material Dispersion results from the dependence on wavelength of the refractive index of the fibre material, even if we assume the core and cladding refractive indices are to be independent of wavelength, the group velocity of each would still depend on the wavelength due to the geometry of the fibre, this is called waveguide dispersion. The waveguide dispersion and the material dispersion together are called intramodal dispersion.

The group velocity is different from the velocity of the individual wave packets.  The group velocity is the speed of the wave packet whereas the phase velocity is the speed of the individual waves.

http://www.slideshare.net/sir2011Anonymous/optical-fiber-dispersion-8807357

http://spie.org/Documents/Publications/00%20STEP%20Module%2007.pdf

However, we do not need to consider waveguide dispersion here because we are using a multimode fibre.

http://www.linktionary.com/f/fiber-optic.html

Just to get an idea of the order of magnitude I used this value:

http://www.teraxion.com/imports/_uploaded/White%20pape-Dispersion%20control%20for%20ultrafast%20optics.pdf

Note that I had to use this example value because manufacturers just don't quote the waveguide dispersion in multimode fibres and you will see in a moment why.

So using the experimentally determined value for pure silica I found (3ps²/(nmkm)), obviously the true value of our optical fibre will be different from this value however I want an order of magnitude. Over 45km, with a spectral width of 30nm this means we will have a rise time of 0.064ns, which is tiny in comparison to the other forms of dispersion.

 

Polarisation Dispersion:

Polarization mode dispersion (PMD) is a form of modal dispersion where two different polarizations of light in a waveguide, which normally travel at the same speed, travel at different speeds due to random imperfections and asymmetries, causing random spreading of optical pulses. Unless it is compensated, which is difficult, this ultimately limits the rate at which data can be transmitted over a fibre. This is more of a problem for single mode fibres rather than multimode fibres.

http://en.wikipedia.org/wiki/Polarization_mode_dispersion

 

The calculation:

The optical fibre I have found has an "overfilled modal bandwidth" of 1200Mhz.km, so over the distance of 45km this implies that the maximum frequency we can obtain just considering the modal dispersion is: (1200/45)=26.6MHz. However, it will be easier to compare all of the different dispersion types if I consider their rise times instead. The rise time is defined as the time required for the signal to change from 10% to 90% of its maximum value. the system rise time is determined by the data rate and the code format.

Notice that figure (a) has a signal with a sufficient rise time, even though the pulses are rounded the signal is still detectable. However in figure (b) the transmitted signal takes too long to respond to the input signal. This has a strong effect on the data rate:

If the response time is not sufficient then we lose information on transmitted data.

To prevent this distortion, an acceptable criterion is to require that a system have a rise time t_{s} of no more than 70% of the pulse width T_{p}:

t_s <= (0.7T_p)

Now there are many different digital encoding schemes, the most popular being Return to zero (RZ) and non-Return to zero (NRZ) The NRZ requires only one transition per symbol whereas RZ requires two transition for each data bit. This implies that the required bandwidth for RZ must be twice that of NRZ, this is shown below.

A non-return-to-zero (NRZ) code is a binary code in which "1s" are represented by one significant condition and "0s" are represented by some other significant condition, with no other neutral or rest condition. The RZ format uses twice the bandwidth to achieve the same data-rate as compared to non-return-to-zero format. But NRZ requires a more complex demodulator since the clock can't be extracted from the signal as it can in RZ.

http://in.answers.yahoo.com/question/index?qid=20080411132348AARdnVl

To avoid this demodulation problem we are going to use a RZ encoding system. This means that the rise time will be given by:

t_s <= 0.7T/2 = 0.35/B_r

Note that the B_{r} is the bit rate and it equals 1/T.

So as mentioned above, due to the modal dispersion our maximum bit rate will be 26.67MHz, meaning the rise time will be 1.31x10^-8s=13.1ns.

I know that for the optical fibre the material dispersion is given by 0.105ps/(nm²km). After speaking with John he suggests that the spectral width of the light sources will be around 30nm-50nm. Taking the lower end of this estimate over a distance of 45km gives us a rise time of 4.25ns.

So the total rise time will be:

Total rise time t_t = SQRT(13.1^2 + 4.25^2) = 13.77ns

This means our maximum bandwidth considering these two factors will be 25.4MHz, to be on the safe side I recommend transmitting with a frequency of no more than 22MHz.

So our maximum bandwidth so far is 22MHz.

If I were to take the upper estimate given to me by John of a spectral width of 50nm then our maximum bandwidth would be 19.8MHz, but to be on the safe side it should be no more than 17MHz.

Tue 22nd Nov: LED Info (2)

Notes:

Infrared LED (to fall in range of 1300nm)

Optical fibers transmit infrared wavelengths with less attenuation and dispersion

Intensity modulation

LED wavelength of around 1310 nm

An LED has a spectral width of about 25-80nm

Rise time average of 2-10ns

3-dB modulation of 30-180MHz

References for my early LED research:

http://spie.org/Documents/Publications/00%20STEP%20Module%2008.pdf

http://www.fiber-optics.info/articles/light-emitting_diode_led

http://www.linktionary.com/f/fiber-optic.html

http://www.ciscopress.com/articles/article.asp?p=170740&seqNum=11

LED Bands Available:

Band	Descriptor	Wavelength range
O band	Original	1260–1360 nm
E band	Extended	1360–1460 nm
S band	Short wavelength	1460–1530 nm
C band	Conventional	1530–1565 nm
L band	Long wavelength	1565–1625 nm
U band	Ultralong wavelength	1625–1675 nm

LED vs. Laser characteristics:

Characteristics	LEDs	Lasers
Output Power	Linearly proportional to drive current	Proportional to current above the threshold
Current	Drive Current: 50 to 100 mA Peak	Threshold Current: 5 to 40 mA
Coupled Power	Moderate	High
Speed	Slower	Faster
Output Pattern	Higher	Lower
Bandwidth	Moderate	High
Wavelengths Available	0.66 to 1.65 µm	0.78 to 1.65 µm
Spectral Width	Wider (40-190 nm FWHM)	Narrower (0.00001 nm to 10 nm FWHM)
Fiber Type	Multimode Only	SM, MM
Ease of Use	Easier	Harder
Lifetime	Longer	Long
Cost	Low ($5-$300)	High ($100-$10,000)

Intensity Wavelength diagram:

From here examples were taken of potential LEDs from manufacturers in hopes of finding one to fall in with our required specs:

1. 1300nm wavelength
2. 12.5MHz frequency
3. Output power above 0.5W
4. Most efficient rise time
5. Lowest possible spectral width

LED Product Examples

http://szjn.en.alibaba.com/product/487202935-212788870/High_power_infrared_led.html

This is an initial Infrared LED, but the power output is much lower than needed.

http://www.optekinc.com/datasheets/OPF395A.pdf

Offers only 850nm, but does give a good frequency of 75MHz.

http://www.optekinc.com/datasheets/OP130_SERIES.PDF

Only offers a 935nm wavelength range.

http://pdf.eicom.ru/datasheets/tyco_electonics_amp_pdfs/259012,_13,_15/259012,_13,_15.pdf

This fell in line with the required wavelength, but as with the majority of these LEDs, the power requirements maxed out at 75microwatts.

http://prizmatix.com/docs/UHP-Mic-LED-White.cfm

This was added purely out of interest. It is a 29 watt LED, showing the possibility of producing the extremely high powers needed for our fiber optic device, but unfortunately it fell short of our wavelength. Ideally this would have been a good guideline for adapting a custom LED to give similarly high power output, but falling in range of 1300nm..

From here I became interested in the possibility of customising an LED for our system with a high drive current, therefore allowing us to produce extremely high output power

http://www.diodes.com/_files/design_note_pdfs/zetex/dn95.pdf

This example of boosting current into the system led me onto a device which can produce a 2.8A drive current with an 18V drive voltage. Using this in a suitable LED would produce 50.4W of power for our system.

From here I knew that our spectral width would most likely be too high for the system, and we would need a way to lower it in our system. I researched novel ways of reducing it, including this article.

http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1072479&userType=inst

It became clear that a simple LED would not be enough to provide the power we were looking for, so I researched super luminescent LEDs, with encouraging results.

http://www.superlumdiodes.com/pdf/sld_overview.pdf

These diodes are able to emulate the high power levels of laser diodes, without actually being a laser. I looked into several of these product types in hopes of finding a suitable example.

http://www.denselight.com/15_DL-BP1%20Series.pdf

This offered a good example of SLDs, offering 1310nm wavelength, 38nm spectral width, but the power output was still at a very low 50mW.

With further research I was able to find varying specifications for SLDs, and on average:

Wavelength: 400nm-1700nm

Spectral width: 5nm-100nm (5nm-50nm should be achievable)

Lowering frequency range to 12.5MHz optimise power for LED

1mW-60mW range ordinarily

60mA-1500mA (1.5A) ordinarily

Allow for optimal beam quality (minimal beam divergence)

The lowest value (around 127 dB/Hz) is attained by the most powerful SLEDs in the 1310 nm window and in the frequency range limited to values less than 500 MHz

They are similar to laser diodes, containing an electrically driven p–n junction and an optical waveguide, but SLDs lack optical feedback, so that no laser action can occur

Unlike standard power supplies, High Output Wide Range LED Power Modules deliver a fixed current to the output. The output voltage will vary as required to maintain the specified output current with differing forward drop voltages of LED junctions

Blending this with an LED driver would have been preferable. An LED driver is a self-contained power supply that has outputs matched to the electrical characteristics of an LED or array of LEDs. There are currently no industry standards, so they can be customised to fit any need.

Using 2.8A high driving current technique (LED driver), with a range of 4-18v, offers a power of 11.2-50.4 watts, which would have been brilliant!

Final Super Luminescent Diode

Wavelength -             1310 nm (Range: 400nm-1700nm)

Spectral width -         38nm (Range: 5nm-50nm)

Frequency -               12.5MHz (Range: >500MHz)

Current Drive -          1.5A (Range: 60mA-1500mA)

Voltage Drive -           Based on other models, we should expect this to be 1V

Output Power -         1.5W (Possibility of 50.4W using LED Driver of 2.8A/18V)

Rise time -                  2-10ns

*These results are based on an SLD system I discovered with no mention of power output. I used the current drive and power output of the average SLD system (being 60mA and 60mW) to get an average voltage drive of 1V. From here I used the upper availability of 1500mA (1.5A) in the SLD in question, with this 1V average, to produce a power output of 1.5W. Dependant on the diode, this could be a wrong assumption, but it is the closest SLD I could find to producing anywhere near the amount of power we need*

From here I looked into surface emitting diodes, in the hopes of finding a suitable power/wavelength blend, as SLDs offer either a very good power output, or a very good wavelength, but never both at the same time. I summed up the information as:

• Novel surface emitting long-wavelength LED structure with narrower output spectrum

• The LED power spectrum is narrowed by growing an integral filtering layer which absorbs the power emitted

• Resulting LED spectra can be as narrow as edge-emitting diodes

Some good examples have cropped up of ultra high power IR LEDs, which again suffer from the power/wavelength trade-off. If this LED had a 1300nm wavelength, we could have scrapped the SLD idea

http://docs-europe.electrocomponents.com/webdocs/09d9/0900766b809d9191.pdf

(850nm, 40nm, 3.6W, 1A DC, 10ns)

An article by L. Goldberg claims that 470mW of power can be output when 165mW LED is seeded with low power SLD. I will look into this as an option for our fiber optic source.

Finally, the average cost of a 1300nm multi-mode LED comes to $695.00

Tue 22nd Nov: Photodiode Info (2)

InGaAs Photodiode

A silicon photodiode is incapable of detecting the wavelength we intend to emit our data with. Silicon photodiodes can typically detect a maximum of 1000nm. Our system will be using a wavelength of at least 1300nm.

Other materials such as Germanium (Ge) and Indium Gallium Arsenide Phosphide (InGaAsP) were considered. Ge photodiodes have a high dark current – this is the major noise factor in reverse bias systems and hence this material will not be used.

InGaAsP is a modified Indium Gallium Arsenide Phosphide (InGaAs). It has a smaller spectral response range (1000-1350nm). InGaAsP are rare since InGaAs photodiodes cover the spectral range of InGaAsP and are of similar prices.

Indium Gallium Arsenide (InGaAs) can detect wavelengths from 900 – 1700 nm with peak responsivity at 1550nm. This makes it ideal for our optical system.

PIN or APD

PIN:                                                                                                                             

PIN (P-type Intrinsic N-type sandwich diode) is a P-N diode with an intrinsic layer in between the P and N regions.

The image above shows the basic structure of a PIN diode.

The P-type region consists of ‘holes’ and electrons are in the N-type region. In a reverse bias system, the depletion region decreases.

This means it is easier for the electrons from the N-region migrate to P-region in the diagram to the right. (note: diagram to the right is for P-N junction and is included to aid help understand the hole-electron equilibrium)

In a PIN junction the holes and electrons both travel into the I-layer. The applied voltage helps speed up the transfer of charge carriers, hence speeding up the diode’s operation time.  Once the electrons and number of holes reach equilibrium the PIN diode will conduct a current.

APD:                 

                                                                                                                          

An APD (avalanche photodiode) is similar to PIN except that it uses an internal gains system which requires a high voltage. This consequently means that APD has a higher dark current and overall noise current.

The very high voltages (100+ volts) causes the electrons initially generated to rapidly accelerate. As these electrons travel through the APD’s active region they collide into other electrons present in the semiconductor material. A fraction of these electrons become a part of the photocurrent because of the collision.

Source: http://www.tpub.com/neets/tm/111-3.htm

APDs are in general more expensive than PIN diodes when comparing like for like models. Another problem was that mass produced InGaAs APDs had a lower cut-off frequency. The lowest one I have found had 600MHz lower cut off frequency. Our LED source is capable of supplying a maximum of 100MHz. It is possible to custom order an InGaAs APD, however that is an unnecessary addition to the cost of the system as we do not require an APD for our system to function. The InGaAs PIN diode is able to detect a frequency below 400MHz.

An APD would provide a higher sensitivity. However we are not required to transfer the signal received into any form of output and therefore our only priority is to detect the signal emitted by the LED transmitter. An InGaAs PIN diode is perfectly capable of doing so.

A PIN diode is cheaper, has a lower noise and is compatible with our spectral and frequency. Therefore, our fibre communications system will use an InGaAs PIN diode.

Noise

In a photodiode there are two types of noise that have to be considered. These are ‘shot noise’ and ‘Johnson noise’.

The shot noise is dependent on dark noise. Dark noise is present in every photodiode with a reverse bias (voltage) applied across it. The dark noise increases with the voltage, it is also linearly proportional to the temperature of the detector.

The Johnson noise includes shunt, series and load resistance.

The total noise current is the mean square root of shot and Johnson noise.

Noise current = (I_j² + I_s²)^1/2

Units of noise current is A/Hz^1/2 (amps per square root hertz)

This can now be used to calculate Noise Equivalent Power (NEP). NEP is the power of the noise in the system. In other words, the detector must receive more power than the NEP to be able to generate a photocurrent which is higher than noise current.

NEP = Noise current / Responsivity

The units of NEP is W/Hz^1/2.

In reverse bias systems, dark current (therefore shot noise) is the dominant factor of the noise current.

Source: http://home.sandiego.edu/~ekim/photodiode/pdtech.html#NEP