A number of you have asked about running the data link at speeds greater than 2 Mbps. There are at least three issues related to doing this successfully.
The solution to (1) is probably the easiest. I was able to cleanly recover up to 10 Mbps data (Ethernet speeds) with the radio as shown, even without increasing the IF bandwidth. The MC13055 is not specified to operate that fast but it appears that the main limitation is the slew capabilities of the output data slicer. Since I am only asking for a limited range, just enough to drive ECL with everything operating from a single supply, the chip can run much faster.
The last IF bandwidth should be increased to accomodate the modulated signal. How wide this should be depends upon what is done in (2) but Carson's rule is probably a very good start: make sure that the L/C ratio is picked to result in low enough Q to pass 2*(deviation +rate). For example if you choose to deviate 2 MHz with 10 Mbps data, that will translate to a 3 dB bandwidth on the order of 2*(2+5)=14 MHz.
The discriminator also needs to be de-Q'd enough to give relatively linear output over the range of the deviation. Probably reducing the value of the shunt R across the quadrature tank is sufficient to accomplish this.
The answer to (2) depends upon how fast you want to go and which transceiver you decide to use. I found that the simplest door-opener transceivers which use a Gunn oscillator and do not have electronic tuning really are not suitable at 10 Mbps. They might be coaxed into working at half that speed but they simply don't have enough performance with low distortion at higher deviations to be able to do a good job at Ethernet speeds.
On the other hand, the electronically tuned varieties of transceivers like the Ma/Comm Gunnplexers, seem to be fine. If you have a pair of these you plan on using, see part II of my series in Ham Radio Magazine Designing a Station for the Microwave bands in the June 1988 issue, pages 31 and 32, here is a plot from that article of the Tuning Curve of a Ma/Comm Gunnplexer and a simple drive/bias circuit for it. This is a circuit for bias-push tuning of a non-electronically tuned transceiver for comparison. The all-band microwave station allowed a multimode HF transceiver to be used on all the amateur microwave bands. Gunn diode based oscillators were used as local oscillators for the microwave frequency conversion. Here is a View of it. The 1296 MHz and 10368 MHz RF "heads" are shown as well.
How to solve (3) is really out of my territory. If you are planning full Ethernet speed, perhaps you can simply use an Ethernet card. Whether or not there are any problems over long paths, where the propagation delay might cause troubles with the 802.3 protocol I don't really know for sure. N6RCE talked with Ethernet card and chipset designers who indicated that these cards couldn't be moved significantly away from 10 Mbps. I confess that I still don't understand how this can be since I don't see anything hardcoded into the silicon that can have the kind of precision it would take to "know" it was being asked to run at the wrong speed. Perhaps dynamic RAM or some of the gates might fail if run at 10% or less of design speed but I have trouble understanding what else would likely go wrong at, say, half speed.
On the other hand, perhaps there is a required relationship for many card implementation which expects the chipset clock and the host buss clock to be related in some fashion.
If someone spends time on this and discovers some answers, please indicate.
Update March 1997
Since I wrote the above I've had exchanges with some individuals interested in trying Ethernet cards at speeds lower than 10 Mbps. In particular, DG1KJD has made some experiments and provided this very interesting report.
Date: Sat, 5 Apr 1997 07:36:53 -0800 From: dg1kjd@dg1kjd-svr.ampr.org Subject: NE1000 cards @2MBit/sec - success report Hi Glenn, some time ago I promised to to you check out common ethernet cards @2Mbit/sec. I know I'm a little bit late but however, I came to the following results: NE1000 clone cards (8 bit; with the old i8390-type controllers) will work pretty well at 2MBit/sec. I tried it between two linux systems (i486DX2-66 and iP5-150 both with 2.0.29) and achieved about 180kb/sec transfer rate. Wow! The linux systems were stable and the packet loss with ping -f was less than 0.1% . I fired up NT on the P5 system, expecting even higher throughput due to the advanced TCP implementation (oh, I forgot to mention that I used pure FTP while measuring the transfer rate; the file size was about 5 MB), but unfortunately NT crashed completely pretty soon after some hundred KBs. OK, then I tried the same with iEtherExpress (the old type) cards. I always hated these boards because it's normal for them to crash the system at about every 6th reset attempt (a reset is for example necessary when switching on promiscuous mode). As expected, they wouldn't initialize correctly, not even under linux. Now here's what I did in order to make my measurements: I bought some standart 4MHz xtal oscillators from the next electronic store, unsoldered the old 20MHz types, replaced them by a socket and plugged my new ones in. In order to connect the two systems I made a cable which I call "null-MAU-cable" (in reference to "null-modem-cable") which is nothing but a crossover in the RX and TX data lines and some shielding that can be connected to the 15pol MAU port on the ethernet cards. I'm very happy about the results and can't wait until I find enough time (just two weeks military service left :-)))) ) to bring this stuff on the air. I modified an old LNB in a way so that the LO signal, which goes directly to the former input, can be modulated with a signal of about 2kHz-3MHz, I'll try to use this thing as a sender maybe I'll work, if not I'll have to buy one of these expensive GUNN plexers. anyway, thanks for all your information, 55 and 73 Jens David -DG1KJD-
Thanks very much to Jens for doing this. If anyone else tries this or similar experiments, particularly with different cards, please report it.
Other than Ethernet cards, in the amateur world there seems to be only Gracilis and Ottawa PI2 cards. I understand that the former might be expected to go at 1-2 Mbps and from my experience, I think the Ottawa cards might do .6 Mbps with internal clock recovery if 20M components and clock were substituted for the standard issue.
I hope that this might help some of you considering running the hardware at higher speeds.
73
Glenn n6gn
Back to thePrevious Page
