A rather low level question: In an experimental setup, an antenna has to be driven by a digital signal, amplified by two full bridges containing GaN FETs like Qorvo's QPD0020. As usual, the bridges introduce the problem of how to transmit the signal to the high side. The signal is generated by an ECL circuit, resulting in rise/fall times below 200 ps; the FETs aren't slower, and fast switching is an important characteristic of the experiment.
But digital optocouplers seem to be limited to about 150 MHz. In the literature, it is mentioned that optocouplers employing a laser diode could reach a few GHz, but neither a manufacturer nor a type designation is given.
On the other hand, there are transmitters and receivers for glass fibers, reaching 10 GB/s and more. But due to the integration of network protocols they fail to transmit a simple "0" or "1" for arbitrary long duration, or a simple 27 MHz clock signal.
So, if you know of a digital optocoupler which is able to transmit DC as well as 2 to 3 GHz, I would be very grateful for every hint. If it is an obsolete part I might be lucky at AERI or similar shops offering old stock.
You would like to have an opto-coupler, but let me first give you the pure amplification part:
To have DC working, you need a differential amplifier or an OP (which is built out of differential amplifiers). The key value to look at is the gain-bandwidth product. You need 2GHz (I took your minimum requirement you will see why), that means to make a gain of 6db you need a 4GHz gain bandwidth product. The rule is -20db/Decade or 6db/Octave. We are talking about voltages for power half the values, but for the calculation this will do, as datasheets often refer to voltage ratios.
I can tell you this because I designed a DC to 100MHz with 10db in the past (around 1990) so nowadays you can do better, but I doubt you will reach 4GHz gain-bandwidth product.
One problem you have to deal with when doing DC is temperature drift. Your output will be different for same DC levels with different temperatures. The datasheets give you the range. I think in an experimental setup this can be dealt with as you can measure the temperature and adjust your results accordingly.
One OP I found is: https://www.ti.com/lit/ds/symlink/opa657.pdf
We are talking about 1.6GHz.
I am out of this for a long time, but my guess is for fibre transmitters and receivers they do not do DC.
Looking at the problem differently: You will easily find the formula:
"BW = 0.35/tr" with "BW" being Bandwidth and "tr" being the rise time.
A rough calculation for 2GHz gives around 200ps (rounded to one significant digit), which is consistent in the range of what your ECL circuit can do. But looking at commercially available products, we are talking about "tr" in the range of 10ns, which is about a factor of 100 times away, and consistent with what you said about roughly 200MHz.
https://www.mouser.co.uk/c/optoelectronics/optocouplers-photocouplers/high-speed-optocouplers
So the opto-coupler makes your problem even worse.
So here is my advice: If you can avoid the galvanic isolation, you might have a chance of getting this done by building the differential amplifiers yourself using exactly the https://www.qorvo.com/products/p/QPD0020 you mentioned. You can half the gain using two stages, doubling your bandwidth, but doubling your temperature drift.
One trick to limit the temperature drift is to use diodes placed in the emitter circuit to achieve a negative drift which compensates for the positive of the transistor. Maybe something like this: https://www.semigen.net/resources/semigen-beam-lead-pin-diodes-datasheet-.pdf, but please look around a little, this is just the first I found.
Finally, take my answer with the limitation that I am an old guy and did this over 30 years ago, so some things might be different today. But the basic physics hasn't changed, so I am confident this part is still valid.
Johannes Maria Frank
Dear Johannes,
many thanks for your answer!
As to the challenge of DC coupled amplifiers, you are certainly right as long as linear amplifiers are concerned but fortunately, the whole design is digital, and the MC100 ECL devices are even temperature compensated.
I'm looking for a signal path capable of transmitting DC because it would be convenient to switch the output stage resp. to drive it into an idle state via the same path. But this isn't absolutely neccessary, of course.
As long as nobody has pointed me to a suitable integrated optocoupler, my current approach is to assemble one from three parts:
The laserdiode driver ONET4201LD by TI seems to cover the range from DC to about 4 GHz, is inexpensive and still in stock at Mouser.
The 1310 nm laserdiode ML720J11S by Mitsubishi matches well the specifications of a design example in the datasheet of the driver, provides rise and fall times of typ. 100 ps, max. 200 ps, and is at least still available from old stock seller Meredith Instruments.
FCI-H250G-InGaAs-75 by OSI Optoelectronics is a hybrid, containing a photodetector and a transimpedance amplifier, fitted into a TO-46 package, covering 30 kHz to 1.75 GHz, according to an eye-diagram the 20/80% rise and fall times are around 150 ps, sensitive from 1100 nm to 1600 nm. Most TIAs for use with photodiodes are offered as chips (but I have no access to bonding equipment) but a long while ago I got one in a common SMD package, and I never succeeeded in attaining a stable state when a photodiode was connected. So, I'm relieved having found this hybrid which, according to RS Components, will become available in June this year.
Being limited to 30 kHz at the lower end, I will switch the output stage between idle and active state using an additional optocoupler HCPL-0900.
I'm still wondering why, with all the components available, there isn't an integrated optocoupler for this frequency range. Obviously, there is not enough demand. The reason is perhaps that digital engineering and RF engineering are, in a way, separate worlds. No righteous RF developer seems to work with half- or full bridges, and where those bridges are common, they are driven with a few MHz at most.
By the way, we might be of about the same age: I built and programmed my first very basic "computer" containing a CPU 8080 in 1976. :-)
Haha, funny, so we switched sides. At around the same time, I built my first astable multi-vibrator using BC-238C transistors. I started off with electronics, building radios and all kind of electronic circuits to end up programming nowadays. Looks like you went the opposite way. :-)
- Badges
- Science topic
- Similar topics
- Psychology