Sending reference carrier/clock over fiber optic cable
Are there any cheaper receiver/transmitter pairs which can effectively transmit a 125MHz carrier?
The only alternative I can see is to send a reference clock at a lower frequency (like 5MHz), and then somehow recover the 125MHz carrier from that. But I'm unsure of whether it's possible to guarantee proper phase locking between all the amplifiers if I do that.
Any suggestions?
You do not mention the length of fiberoptic cable. If the losses are small, you can probably get away with a 5 way optical splitter, and save the cost of 4 transmitters.
The distance is quite short, maybe ten meters max.
I don't think splitters will work for the 125MHz carrier since each amplifier will need a carrier with independent phase (and possible independent frequency). Splitting might be useful for a low frequency reference clock, but again that begs the question of how I use that clock to restore the desired phase and frequency of the carriers on the RF amps.
I've already decided that I'll need some sort of CPLD/FPGA on the transmitters for decoding the fiber signals, and possibly synthesizing a carrier with a DDS. So I've got that at my disposal. But I can't see how it would be possible to accurately synchronize the phase accumulators of the DDS on the amplifiers.
well, 125 MHz for our digital bretheren is sstill pretty fast! That is why the modules cost S20 each, you need to be modulating laser diodes.
If you ran a 5 Mhz or so signal to all the points, and then use a cheap clock multiplier chip to get up to 125 Mhz, then you could either use LED drivers or maybe even copper twisted pair with digital line drivers/receivers.
Hmm, I was hoping that I could get a better price if I found parts that worked at a different wavelength, or worked with crappy multimode fiber, or something. Also Avago is one of the more expensive brands out there (being a daughter company of Agilent), so I though that there might be equivalent devices out there at lower cost.
Yes, but then I again run into the question of how to correctly phase/frequency modulate that 125MHz clock.
Let's say that hypothetically I'm fine with all the RF amplifiers always operating at one frequency (though that frequency can change over time). And that the desired frequency is always an integer multiple of my reference clock. But I still need the RF amps to be phase locked to each other, with some arbitrary phase modulation on each of them. How can I implement that phase modulation? Digital delay lines or something? I'm sure I need my phase to be accurate within <10 degrees, with a full 360 degree range, so analog methods probably won't cut it.
So I'm starting to think that I'll need to use coax for carrying these signals, in order to reduce cost. The potential problem with this is that the amplifiers are driving coil antennas directly on their outputs, so there will be strong RF magnetic fields surrounding the amplifier. The problem is when this RF contaminates the incoming carrier signal, thus creating a horrible feedback loop.
I suppose my problem is similar to the design of a RF repeater. It's already required that the coil antenna be directly attached to the RF amp, so if I use coax there's no way to prevent it from being exposed to the high power RF fields. Are there any tricks to try and prevent coupling between the coil and the coax carrier, or to reduce the effect of this coupling on performance?
well, fiberoptic is the best way to go for signal ingress. I suppose you could send two tones along the coax, like 50 and 75, lowpass filter at the end unit, and then mix the two to form 125 MHz.
I would also look at sending the signal digitally with shielded twisted pair and line driver/receiver.
- - - Updated - - -
line driver/receiver. DS91M047 or AD8320ARP
Doing a mixing operation might work, but I would likely need well designed IF and RF filters. And I suppose I could get away with only one of those tones being modulated by DDS, while the other is generated by one PLL and is distributed to all amplifiers... I'll give this one some thought.
And yes, shielded pair may be more immune than simple coax. Do you think current mode logic have better immunity for this application? Also, specifically what kind of cable would you recommend? I've only worked with shielded twisted pair for use with small personal projects, and don't know much about selecting an optimal type. Like some are basically cat5 cable (with only one pair), while others are very fancy, practically twinaxial cable.
Double post? Seems the boards are still hiccuping...
The twisted pair should help because your transmitted signal will appear like it is common mode, and get a maybe 30 dB of rejection in the line receiver (assuming it is a good one). Also, the line receiver is digital, not analog, so there should be some built-in noise immunity that will also help keep the system from self oscillating. The question is: can you find a line driver that can drive the capacitance of 30' of twisted pair, maybe 500 pF. The data sheets for these line drivers list max operating frequency vs capacitive load. It may be that the line driver will treat the line more like a transmission line and propagate the signal just fine, or it may have a very poor rising/falling edge. You would have to get a fast one and do some actual tests.
So I've been looking at various twisted pair and twinax solutions. I've found some sources for RG108 twinax cable, and Amphenol makes twinax connectors which seem reasonably priced.
But then while searching I found that typical SATA cables are actually twinax, with two pairs per cable. So that sounds perfect, since I have two signals to transmit anyways. SATA cable assemblies and connectors are dirt cheap. Assuming that I'll be able to find long enough cables (they seem to come in lengths up to 2 meters, even longer for eSATA), is there anything wrong with this approach? The only possible issue I see is the quality of the shield. I suspect it uses foil wrapping, but I've actually read that foil is more effective than braided shielding for RFI immunity (though foil is much less physically durable).
Maybe you need some driver amplifier circuits for CLK signals before the cable.
Right, I've had trouble finding something that has low part-to-part skew, as well as good EMI immunity. The best I've found is the sn65cml100 or sn65lvds100, which looks like it could work for both the receiver and transmitter. The weird thing is it's specified for driving two individual 50ohm or 25ohm cables, rather than a single 100ohm twisted pair or twinax. Even the dev board they sell for it has two separate coax connectors for the differential outputs. Seems strange, but I assume it would have no trouble driving a 100ohm twinax cable.