I'v looked RF amplifier AD8351 for an application of a 1Gbps LVDS transmission system. The system is require fitting to a single-ended output with 50Ohm coaxial cable. I searched the datasheet of AD8351, but no sample of single-ended was on here.

If I terminated partial output line of AD8351 with 50ohm resistance, Should any problem be happened in this application?

Hello ayanaplus

If your input is a differential LVDS and you need 12 dB gain into a single 50 ohm load, you will need to set your amplifier to 18 dB gain. If you choose Rg=10 ohms for the AD8351and your differential load to 100 ohms (50 ohms on each side) the maximum gain is 17.4 dB (from the equation on page 10 of the AD8351 datasheet). This is close enough to your 18 dB requirement but the Bandwidth of the AD8351 will fall far short of your 750 MHz requirement. See TPC1 and TPC2 on page 5 of the AD8351 datasheet and note that the BW will be around 400 MHz. AMplifiers whose gain depends on the load such as the AD8351 (or the ADA4960) might not be a good choice for you.

If you want to achieve a gain of 18 dB with a 750 MHz BW, you might consider the traditional diff-amp architecture. In this case you would need a unity gain stable voltage feedback diff-amp with GBWP of about 6 GHz. Such an amplifier does not exist to my knowledge. You can consider a two stage voltage feedback system each with a gain of about 3. The PBWP of each of these two should be about 3 GHz. This is a little closer to achievable but not quite.

An alternate method you could consider is to use 2 ADA4927 diff-amps. These are current feedback diff-amps and their BW depends on teh feedback resistor value, not on the gain. You can adjust Vocm of the first stage to match you LVDS common mode input levels and you adjust Vocm of the second stage to zero to eliminate any dc current at the output.

Here is the schematics:

Please note. The input termination resistors depend on the LVDS source. If the LVDS source can drive 100 ohm diff load then make these resistors 273 ohms each.

SInce you said you want to put the output into a 50 ohm co-ax, I added a 50 ohm series resistor at both outputs of the diff-amp for a back terminated arrangement.

The gain is slightly higher than your requirement. Increasing all 63.4 ohm gain resistors slightly can reduce the gain.

Here is a picture of a low frequency inputs and output. Note that a 200mV pp differential LVDS will give you 1V pp at the load end of a properly terminated 50 ohm co-ax. ALso note that I put a 500 mV offset on the LVDS input lines to illustrate that the diff-amp removes the common mode at the input (within reasons).

Here is a low frequency pulse response. The inputs have a 5 ns rise and fall time in this picture. Don't have a readily available differential pulse source to show a faster pulse response.

Here is the rising edge of the output expanded horizontally at the end of a 2 feet long RG-58 co-ax terminated in 50 ohms.. Please keep in mind that the input sie time is 5 ns. Don't have ready access to a pulse generator with faster rise time.

Here is the frequency response. The Bandwidth is about 850 MHz. A 0.7 pF capacitor is added across the input pins of the second ADA4927.

Here is another frequency response plot. A 2.7 pF capacitor is added at the input of the network analyzer on this one.

Please note that PCB layout can greatly affects the frequency response at high frequencies. Care must be taken to ensure proper high frequency PCB techniques.

And finally here is the picture of the PCB. Used two of these boards connected with SMA barrels.

I hope this helps.

Zoltan