Design notes
Draft notes during design phase
Existing 3U boards
- CTDAB has been originally designed for pulse transmission, but has already been tested in operation with the GMT 500KHz DC-balanced signal.
- CTDPR has been originally designed for pulse reception, but we confirmed in simulations and in the lab on the existing 3U boards that it can achieve reception of a DC-balanced signal up to >10MHz
Optical components selection
- Regarding the transmitter, note that we need high power bandwidth (around 1mW) due to the expected high attenuation and the sensitivity of the receiver
- The HFBR-2316T receiver has been successfully used in CTDPR since years. According to the company Broadcom:
> "There are currently no foreseen plans to obsolete HFBR-1312TZ nor
HFBR-2316TZ.
> The above transmitter/receiver combination is designed for
signal/data transmission over graded index multimode fibers e.g. 50 or
62.5/125µm.
> Only the receiver can also be used with single mode fibers.
> The transmitter is a 1300nm LED with optical power coupled in
multimode fiber between about 10-100µW.
> A 1mW version is currently not available."
Optical transmitter and receiver design
The design is based on the following existing modules:
optical transmitter | CTDAB |
---|---|
optical receiver | CTDPR |
TX
Due to the increased price of the laser transmitter used in the CTDAB, it was decided to switch to a different transmitter (Appointech/Municom FD13R0STSTXS-4BX4, datasheet 1310nm_Sugar_Cube.pdf). The device we order is without an optical isolator, a component that generally just allows light transmission in one certain direction, while the other direction is more or less blocked.
The new transmitter comes in a sugarcube package, similar to the one used for the receiver, which allows to mount both at the edge of the RTM PCB and avoid any fiber patch cables on the PCB, like it was done in the CTDAB.
The new transmitter is more powerful, a fact which has also led us to change the laser driver IC, from the Micrel SY88822V to the more powerful MAX3766EEP+ from Maxim. The Micrel SY88822V was also much harder to procure, so this modification should make it easier to produce the boards.
In place of the jumper which provided two power settings for CTDAB, we introduced a resistor, in parallel with a (non-mounted) trimmer. By default, the power will be fixed with a specific resistor value, but a trimmer could be used instead for experimentation and/or fine-tuning for very special cases.
RX
On the receiver side, the circuit has remained largely the same, with a few exceptions:
- The two transistors were replaced by a single matching dual transistor pack, for better tracking/mirroring under varying environmental conditions.
- The pull-up/down resistors at the inputs of the comparator were changed in order to increase the detection threshold. In the original CTDRP design, which was conceived for very weak transmitters, the threshold was chosen such as to allow detection of pulses of few mV. With CTDAB (and our new optical transmitter which is even more powerful), one can safely expect pulses of more than 100-200mV, therefore the threshold can be increased in order to make the receiver more robust and immune to noise.
- The size of the capacitor on the emitter of the second transistor was increased. This was done to allow longer pulse widths, since the RC circuit formed on the emitter of the second transistor controls the "filtering" of the input signal in order to generate the threshold; if the capacitor value is too small (or, equivalently, the pulse width of the incoming signal too long), the threshold will start rising towards a point where eventually there will be no positive output on the output of the comparator (thus creating shorter pulses than those on the input).
The above modifications were verified with a SPICE simulation, using the SPICE models provided by the manufacturers for the transistors and the comparator.
E2000 Connector type choice
Regarding the E2000 connector, back when the board was being designed, the design was presented to the IN team. Sure enough, they raised the issue of the connector, and the designers explained that this was a very conscious choice, done in order to have a "sugarcube" package for the laser and the diode. The sugarcube package provides three things:
- Robustness. Compared to the pig-tailed 3U cards, there is no dangling fiber
- Simplicity in manufacturing, since the package itself is soldered like any normal component, no need to handle the pig-tail, secure it on the board, make sure that the loop is not too tight, etc.
- Less board space which allowed us to put 3 outputs and 6 inputs on a single board (not possible with pig-tails without making a spaghetti).
Presented with the above arguments, the IN team agreed that it's ok to proceed with this type of connectors and that it will be ok to provide small local patch panels on the racks to convert from/to E2K.
Receiver HFBR-2316TZ input optical power
In the Electrical/Optical and Dynamic Characteristics of the
HFBR-2316TZ we
see
On our setup with CTDAB and CTDPR we noticed that when the optical power
at the receiver CTDPR was ~500uW (CTDAB, without attenuation), the
signal in the output was ~16ns longer.
The following pictures zoom in a rising and a falling edge of a 5MHz
signal (200ns pulse width) at the input of the CTDAB transmitter
(yellow) and at the output of the CTDPR receiver (pink).
The rising edges with or without attenuation show ~40 ns skew.
The falling edges with attenuation (that means optical power at the
receiver ~10uW) show also ~40 ns skew, as the rising ones.
The falling edges without attenuation show 40+16ns= ~56 ns skew, as we
are exceeding the Peak Input Optical
Power.
As CTDPR and CTDAB have been used together in operational installations
for pulse transmission, we believe that this pulse extension when the
optical signal is very strong, is not a problem.
Note that for the GMT signal, these 16ns are a very small fraction of
the 1us-long pulses.
CTDAB-CTDPR system latency
In the setup we connected the CTDAB to the CTDPR with just few meters of
fiber.
We measure in yellow the GMT input of the CTDAB and in pink the output
of the CTDPR.
The system latency as the following figure shows is is
~37us.
Tests on pre-production boards
Files
- 1310nm_Sugar_Cube.pdf
- opt_repeat_measur.xlsx
- CTDPR_LTspice.asc
- rx_peak_input_opt_pwr.png
- rising_-20db.JPG
- rising_no_atten.JPG
- falling_-20db.JPG
- falling_no_atten.JPG
- GMT_latency.jpg
- GMT_latency.jpg
- CONV-RS485-OPT-RTM-RX_sim.asc
- Preseries-testsetup.png
- Preseries-testsetup1.png
- Preseries-testsetup1-oscillo.png
- Preseries-testsetup2.png
- Preseries-testsetup2-oscillo.png
- Preseries-testsetup2.png
- Preseries-testsetup1.png
- Preseries-testsetup2.png