Driving multiple laser diodes with GHz arbitrary waveforms

[ktyszka]

Hi, I do not have much experience with high-frequency electronics and I am looking for some advice on how to drive multiple laser diodes with GHz arbitrary waveforms.

For the purpose of a scientific experiment, I need to illuminate a sample with multiple laser diodes (LD) arranged in a 1D array. These LDs need to be driven with analog signals in such a way that the arbitrary electrical waveform is converted to light intensity emitted by each diode and each diode has a specific intensity in time and these intensities vary with GHz frequency.

In general as high driving frequency, as many LDs, as much optical power as possible is my goal.

I have access to TEC AWG7082C (8GS/s - 3 Ghz, 2 channels). As LDs I wanted to use VCSELs typical for ethernet communication (like HFE6X92-X61-10Gbps-850nm-VCSEL-LC-TOSA-Package-Data-Sheet). I thought I could build a kind of electro-optical serial-to-parallel converter in which I would drive serially connected VCSELs with one channel of AWG. By selecting the lengths of the wires between the VCSELs I could control the driving signal delay. So if I drive 10 VCSELs like that and program an arbitrary waveform with 10 consecutive analog levels I should theoretically get a proper output intensity configuration for every 10th sample. This way I have a kind of electro-optical serial-to-parallel conversion which depends on AWG sampling rate and number of VCSELs used.

My question is if this would work and if it is possible to drive an array of VCSELs with AWG. I can imagine I need a current source for VCSELs (or typical circuit with a resistor?) and some kind of tee-bias (?). Could you advise on how to connect such a setup? Should I use a high-frequency splitter with delays (https://www.minicircuits.com/pdfs/ZN16PD-0563-S+.pdf) instead of serial connections?

An alternative approach would be to do something similar but use only one high-power high-frequency VCSEL and carry out the conversion in the optical domain (by beam splitting and delaying), although I could not find this type of VCSELs.

I have a budget in the range of 2000 USD to upgrade this idea somehow.

Solutions I have considered:
Some kind of FPGA board or DAC with multiple channels to drive VCSELs
Some kind of SFP development board to drive multiple SFP transceivers
Buying some old-school multichannel pulse/pattern generator (if something like that exists)

I will appreciate any suggestions and your opinions. Best regards, Chris

 

[BradtheRad]

* At least one experimenter put up a Youtube video, demonstrating how he measures the speed of light (and electricity) with a GHz-capable oscilloscope, by observing the delay through 12 inches of wire.

* Old-fashioned color tv's brought the chroma into sync with the luma by using the delay of sending one signal through 200 feet (or thereabouts) of wire.

 

[ktyszka]

Thank you for these tips. I do know that it is possible to use the propagation delay. My question comes down to whether it is possible to drive VCSELS in series with GHz AWG (in a simple benchtop setup as shown in the figure) without damaging the components.

 

[FvM]

The series connection scheme achieves terrible impedance matching with reflections at each diode. You can get a basic idea with a LTspice simulation, modelling the diodes with the specified series impedance of 60 ohms.

The parallel splitter will probably work well in terms of AC impedance matching. The datasheet suggests that it acts as low ohmic DC short, respectively it misses a means to balance the bias current. Series resistors with bypass capacitor might work.

 

[biff44]

the string of diodes in series will certainly turn on quickly, but might not turn off quickly, destroying the light envelope pulseshape.

the best thing would be to drive each diode independently with its own semiconductor driver chip.

the 2nd best thing would be to try a five way RF power splitter with a BIG BANDWIDTH.
for a 3 Ghz digital input, i would probably want a power splitter bandwidth of 2 to 12 Ghz.
and then i would drive the snott out of it with a very high power amplifier.

if that kind of worked, then i might try five independent passive matching circuits, to try to raise the diode load impedance to something higher to work better with the power spliltter's 50 ohm impedance.

 

[dick_freebird]

When you say AWG this implies to me an analog waveform.
VCSELs may not be especially power-linear, especially if close
to lasing threshold?

Laser diode drivers for comms I believe use a bilevel output
which has both logic states putiing out current above lasing
threshold, big and little for 1/0. If you let the laser quit lasing
it becomes slow.

A question is whether youd like to separate the time and the
amplitude setting, like put a control voltage / word to whatever
sets the "big end" laser current, gangwise or per driver, and
use a good timing generator to do the phasing? You might
be able to use a "spare" VCSEL as a feedback element for
the "big end" current setting, with a calibrated phototransistor
or photodiode to measure.

AWGs likely run slower than a b@lls-out-for-speed digital
timing generator, because DACs are slower than plain
drivers; analog-capability "baggage" which maybe you can
"divide and conquer"?