Updated 20020201

This is a "quick-and-dirty" page that I set up to describe the T1-rate modem that I designed several years ago (in 1993, actually...)  Let me point out that unless you are going to use this on wires, this modem is absolutely useless to anyone other than licensed amateur radio operators:  It does NOT intrinsically lend itself to FCC part 15 operation.

Terms of usage of the designs referenced on this and my other pages:

You are free to experiment with this design as much as you want.  All I ask is that you do so legally and, if you make any improvements and refinements, you let me know what you have done so that I can share them with others as well.

Firstly, let me tell you what this modem is NOT:

• It is not a real T1 modem, but a T1-Rate modem.  It does not do things like AMI, B8ZS, and so on.
• It is not a kit.  There are NO plans to make a kit anytime soon.
• It is not plug-and-play.  What you get are schematics and a description.  If you are intimidated when it comes to building things from scratch, then maybe you need to pick another project (or find a friend to help you build it...)
• This modem is not the ultimate in optimized design.  This is a somewhat refined prototype that DOES work as advertised, as long as it is adjusted correctly by someone who understands what they are doing.  This modem uses easily obtainable, off-the-shelf parts that should be available, through only modest effort, anywhere in the world. I am absolutely confident that this design can be improved upon.

• This modem, properly built and implemented, will transport data from one place to another at up to T1 data rates with a minimum of errors.
• This is a baseband modem.  That is, given a medium that has a bandwidth of 1.5 MHz (starting at a few 10's of hertz) and is reasonably well-behaved in terms of amplitude and phase, this modem will push data across it.
• This is a full-duplex modem (well actually, if you only ever wanted to send data in one direction, that would be fine, too...)  Clock acquisition takes a significant fraction of a second, so you will want to have a path that is there all of the time.

Suppose you have wideband FM transmit/receive radio set (Amateur Television FM transmitter and receiver are almost ideal, except that the receiver's IF bandwidth will be wider than necessary...)  As long as you are on the 33cm Amateur band (or higher) you can legally use this (in the U.S., at least...)

Now, let's suppose that we want to link two sites that are, say, 30 miles apart with a full-duplex T1-rate point-to-point link.  To simplify matters, let's go crossband, so on my end I have a 23cm FM video transmitter and a 13cm FM video receiver, and the other end has the opposite- a 13cm FM video transmitter and 23cm FM video receiver.  If the antennas are 10-13 db gain and there is 1 watt of transmit power (overkill, actually...) and a clear line-of-sight path, we'll be fine.  Keep in mind that we've modified the receivers to have a 5 MHz IF bandwidth or so to better match our actual occupied bandwidth.  (If you didn't narrow the bandwidth, it would still work, but you'd have more possibility of adjacent-channel interference issues, plus lower effective receive system sensitivity.)

The baseband output of the modem is simply fed into the transmitter input and the transmit deviation is adjusted so that the occupied bandwidth is on the order of 3 MHz or so.  On the receive side, the recovered baseband output is fed into the modem, the VCXO on the clock recovery is centered, the data-sense jumper is set for proper polarity, and we are ready to go as soon as we do the same for both ends.

The tricky part is to find some sort of box that can handle T1-rate data.  One such box that we have used is the Gracilis Packeten.  The modem supplies the transmit clock to the Packeten, and it reconstructs and supplies the receive data clock to the Packeten as well.

At that point, you can set up your routing tables and knock yourself out.

Well, you were already pre-warned that all you get are some schematics.  The schematic shows an integrated demodulator/slicer system.  What's that?  A demodulator?  I thought you said that this was a baseband modem?  Well, it is...  except that the radio that we initially tested it on (when I drew the schematics) did not have a demodulator, so I used a Motorola MC13055 data receiver IC for its discriminator, and used the built-in comparator/slicer to get a "raw data" stream for clock recovery.  If you already have a baseband source (i.e. from a video receiver) then it's a "no-brainer" to use something like an LM311 as a slicer and some fast op amp to increase the amplitude of the data if necessary...

 
Actually, the hardest parts will be:
• Finding a set of radios that will do this.  We used some 150 and 600 channel Farinon radios and modified them for the closest ham band. NO!!!  ONE PAIR OF GUNN TRANSCEIVERS WILL NOT WORK AS A FULL-DUPLEX LINK!  Look below for the explanation why...  As I pointed out, off-the-shelf FM Amateur Television transmitting and receiving equipment can be used.  I'll provide links to some possible sources later...
• Finding some box that can handle T1-rate data.  Forget your standard PC interface cards.  Actually, it takes some dedicated hardware.  As I mentioned, we use a Gracilis PackeTen box which has a 68302 - a dedicated communications engine with a 68000-like processor and some RISC-type engine on it.  Since we have those boxes on hand, I can't offer any other alternatives since I haven't looked.  If you find some clever alternatives, then please let me know...

Building the thing.  I'll offer some advice, but I won't build one for you!  If you go through the trouble of designing a circuit board, refining the design, and you are of the honest, generous sort, I'm sure that you'll send an electronic copy of your work here so I can make it available to others (and your payment?  I'll give you the credit!)

I have been completely swamped with other Amateur Radio projects (not to mention with work...) so I haven't had time to revisit this T1-rate modem-thing for a few years.  However, there are a few aspects of it that I'll single out for improvement and/or suggestions.  Since I'm writing this off-the-cuff, I'll also throw in some seemingly random comments as well:

• The clock recovery circuit really works quite well.  It isn't particularly fast, but if you have a constant signal from the other end, then that's no problem as you are always locked on.  I have built a variation of this that uses a 74HC4046 VCO instead of a VCXO (I limited its frequency range to prevent it from locking onto harmonics, etc.) that worked extremely well.  The 24 MHz VCXO was chosen for this design since it represented a reasonable degree of oversampling and was a convienient and commonly available crystal.
• The data recovery uses integrate-and-dump filters as its primary means of symbol filtering.  This isn't particularly ideal and you'll likely want to design your own, optimized filter.  You'll not doubt notice that since the LM311's, the 4066's and the HC logic family used in this design is operating near its upper frequency design limit, there is a pair of integrate-and-dump filters used to help relax timing constraints.
• A few years ago I built a 128-256 kbaud modem prototype based on this design to be used as a subcarrier on video (using the 74HC4046 for the VCO as mentioned above) and used only one integrate-and-dump chain.  Unfortunately, I have since lost the schematic and I don't have any idea what happened to the prototype...  One of these days I'll re-create it I suppose...  Now 128-256 kbaud is a data rate that is high enough to be useful, and still can be interfaced to more commonly available PC-type interface cards (like the PackeTwin card from Pac-comm, or the Ottawa PI or PI2 card.)

A completely different modem design that could be implemented as a subcarrier rather than a baseband signal.  If it were to be somewhat spectrally efficient, it would have to be much more complicated than this one.  However, if you are talking about only a 128 kbaud or so subcarrier on video, then the aforementioned design would be quite workable.

    A well-publicized high-speed data link is the "2 mbps microwave data link" article that first appeared in Ham Radio magazine and was later published in the ARRL Radio Amateur's Handbook.  This article describes what appears to be a full-duplex microwave data link that operates on the 24 GHz ham band using Gunn transceivers. Just like coax- based ethernet, this links is NOT full duplex, but half-duplex!

    Perhaps you have used Gunn transceivers and thought "hey, these are neato!  They're full duplex like a phone!"
    That's the problem! Like a telephone, you can hear yourself talking...
    Gunn transceivers receive at the same time that they transmit by virtue of the fact that the transmit oscillator is also the local oscillator for the receiver.  Because they modulate that oscillator for transmit, what you are transmitting also appears on your receive signal.  So, if you are transmitting data, the receive data from the other end will also have your own transmitted data on top of it.  The result will likely be interpreted as garbage by your modem!
    There are three obvious solutions:
• Use a different set of Gunn units for receive and transmit.  The disadvantage is that it will take two antennas and twice as many Gunn devices. This is quite do-able, but it certainly adds to the hardware. Perhaps one could use an orthagonal feed and utilize opposite polarities for the two directions, thus allowing the use of a single dish...
• Modulate the data on one (or two) subcarriers. One direction could be baseband (from 0-1.5 MHz or so) and the other could be modulated on a subcarrier to occupy baseband from, say, 2-5 MHz. If you were clever, this could be done (maybe some 74HC4046's may be modulatable at the data rate, but I don't know for sure - they do work up through at least 256 kbaud, so I can at least vouch for that... but you'll probably have to up-convert the subcarrier to a higher frequency to get a decent demodulator, or use good filtering and use, say, a UA733 and some inverters for the demod, but I digress...) The obvious disadvantage is that you've increased your occupied bandwidth tremendously and will suffer the obvious effects from doing that (i.e. less range, more susceptible to nonlinearities, fading, etc.) By doing this, they two signals, while they will be present at each receiver, will be spectrally different and can be (relatively) easily filtered out. Hey, you might want to do this anyway for lower-speed channels, or some analog audio links, anyway!
• Subtract out your own transmit modulation from your received signal.  This is the most likely to be doable.  Since you know what YOU are sending, just use an inverse copy of it to cancel out your own transmit signal from the received signal.  This assumes, of course, that both your modulator and demodulator are adequately linear to allow good cancellation.  If you do a simple subtraction, keep in mind that your receive bandwidth will be approxmately doubled since you now are trying to receive two signals and that will increase the system's noise bandwidth.  To avoid this, do a broadband conversion of the Gunn's mixer diode output to an IF, then, using a local oscillator modulated with your own transmitted waveform (but in the oppposite sense) cancel out your transmitted signal.  After mixing down to a new IF the resulting signal will now be (relatively) free of your transmitted signal and you can do your optimal IF filtering and demodulation.

• Originally, there was an archived file that contained the drawings in HP-GL and PCL4 format.  Since those formats aren't particularly easy to use these days, they have been converted to 2-color .GIF bitmap images instead.  Not being able to find the original .DXF files from which these images were orignally created, I was forced to convert the GL file to the aforementioned raster format at a fairly high resolution.
• Note that these bitmap images are quite large and they may or may not print properly directly from your browser.  When you do print them, please do so in landscape mode for best results. Errata in the previous drawing/documation package has been used to update the drawings.  If you have the previous versions of drawings and documentation, please discard them and use those in the links below.

The images are as follows:
• t1mdmrxa.gif - A sample baseband demodulator using the MC13055 chip.  This shows how the "raw data" (for clock recovery) as well as the slicing levels and baseband signals are processed prior to handing them off to the modem.  (Ver. E.1, 154102 bytes)
• t1mdmtxd.gif - The transmit portion of the modem.  (Ver, E.1, 158611 bytes)
• t1mdmcdd.gif - The demodulator/clock recovery portion of the modem.  (Ver. E.1, 212845 bytes)
• There is also a text file that describes the theory of operation  and adjustment of the modem that may be found here.  (Text Rev. 1.4, 25754 bytes)


I would expect that you will read this page AND the documentation before you start asking questions.  If the answer isn't in the documentation (possible...) or you can't make sense out of what I have written (likely...) then I can be reached via email..

Perhaps I'll get some time and add links to this page, but for information about Amateur FM ATV transmitters, go to the Utah ATV Home Page and follow the link to the Houston Amateur Television Society (HATS). These people sell some FM ATV transmitter kits. You can also look at some of the vendors listed under the Where do you go for parts/kits/information on VHF/UHF/Microwave projects? section. In particular, go to the Mini-Circuits page and look at their VCOs: These VCOs can be used as the basis of transmitters and LOs for receivers - keeping in mind that they have an intrinsic limit on the highest frequency that can be modulated on them (translation: You probably can't push them at T1 rates...)