This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

THS3491: Impedance matching to drive a 50 ohm transmission line

Part Number: THS3491
Other Parts Discussed in Thread: OPA2677, DRV134


I'm currently in the process of selecting a high current/speed CFA in order to drive a bus with the following specifications : 

-characteristic impedance is 50 ohm

-the transmission medium is a coaxial cable that can be up to 100ft 

-transmitter voltage range is between 6-9Vpp 

-the bus bandwidth is 20Mhz 

-the transmitter end must be terminated with a 50 ohm resistor and the bus end must be terminated with 50 ohm terminator to match the bus impedance

The following screenshot shows the configuration I'm using for the THS3491 

*Note : transformer is 1:2 

I'm using a 1:2 transformer because the headroom on either supply is high and I want to supply it using +/- 7.5V.

Using the above circuit in practice, the output voltage is just fine (9Vpp). 

However, when I insert the terminated coaxial cable a the transformer end, the voltage goes down to about 6.5Vpp and about 6Vpp on the terminated end of the cable. 

I'm suspecting it is due to a impedance mismatch from the source impedance towards the load impedance. 

Here is an explanation about my feedback network : 

I'm driving both inputs with a 1.8Vpp signal with a 50% duty cycle for testing purpose

The feedback network is configured with a 2.5 gain, since the desired gain is 1.25 by considering that the output resistor/bus impedance act as a voltage divider 

The series resistor (R22) is selected according to transformer impedance ratio : Zp/Zs = (Np/Ns)exp2. 

My question goes as follow : Is there any work-around to fix the impedance mismatch or issues I'm currently having, any help would be greatly appreciated. 

Kind Regards 

  • Why do you have that 2nd 50ohm on the output, it would seem to me just get rid of that and use a 12.5ohm source R into the balun and you would get a matched driver without the extra 50ohm. 

  • here is what I had in mind, the bus thinks it sees a 50ohm termination looking back through the balun. here I used one that seems to have about the right frequency span - you probably need a blocking cap on the input side to keep DC current at 0 - driving right into the Balun will create a lot of DC current due to output offset voltage. I did change the R values around the THS3491 to increase loop gain, you have it real bandlimited with the values you used. Also, on the output side you need a DC current path - for the sim at least. 

  • Oops, I forgot to change the load on the secondary to 50ohm, here is an 8.8Vpp swing at 10MHz, I also changed this to +/-7.5V supplies, 

    And here is the SSBW response with correct load, rolled off about 0.6dB at 20MHz, 

    THS3491RGT through balun to 50ohm.TSC

  • Hello Carl,

    I am taking some time to look over your circuit and the writeup; please keep us updated in the thread if you simulate and/or adopt Michael's suggestion.



  • Hey Carl, I got to ruminating on other ways to do this and I remembered this very ubiquitous (but somewhat secret) active impedance differential driver that is used throughout the DSL industry. Here, with cross coupled feedback, we can create more output apparent impedance (I call it synthetic impedance) than the physical buildout Ro. This was done early on in DSL line drivers to get more power to the load and less wasted in the buildout R. Most folks have not seen this, but those of us in DSL lean on it everywhere. The point is with a part like the dual OPA2677, you could easily deliver your desired load power on +/-6V or maybe even lower, I will wait to see where you think I have already gone astray before I try to fire up those design spreadsheets I developed many years ago - I think the last time I looked at them was maybe 2012. Here is the basic topology, and the analysis is very involved, but the results are powerful. I have found those tools, but will not waste any time on them if this is going the wrong direction. The balun would go where Rload is and the required differential input could just be another Balun. This circuit is always driven by a complementary output DAC in the DSL world through a passive filter into the op amp inputs - and the OPA2677 is quite commonly used, not only do you get this nice push/pull doubling of the output voltage swing, but you don't lose 6dB through Ro element anymore. 

  • Hi Michael,

    the audio industry is also using these cross-coupled circuits. The datasheet of DRV134 shows a similar circuit.

    These circuits are a bit critical because they either introduce a positive feedback path or a dual feedback topology which only work stably with a limited range of load capacitance. These circuits can beautifully work with a fixed and constant load, but can become instable with varying loads.

    When I had to do with audio electronics some decades ago, I decided against cross-coupled differential outputs because of stability issues which I observed when a cross-coupled differential output was loaded "non-differentially" (with one output becoming connected to signal ground). In some difficult applications I even used good old audio transformers.


  • There is probably an issue using a VFA topology, the CFA works very well to widely varying loads as the DSL line impedance goes all over the place, 

  • Hi Michael, 

    thank you for the replies, I'm looking at all this during the current morning.

    The reason I have a second 50 ohm on the output may be unclear with my provided schematic. The driver is terminated by a 50 ohm because of the requirements when the bus (coax) isn't connected and the second 50 ohm at the end represent the terminator I'm using when I plug the bus (coaxial cable). It makes sense that it causes an impedance mismatch, however I'm unsure about how to match the impedance with a "floating" driver and a "connected" bus with the coax cable. I will look at your provided schematic below, I'm guessing there is a way to "actively" match an impedance depending if the bus is connected or not? 


  • Well that 2nd 50ohm on the output side of the balun is why you lost gain. The driver does not need to have that termination there for its purposes when the cable is unplugged, I might put say a 500ohm load there to have it see some load when the cable is disconnected. CFA's like the THS3491 or OPA2677 mainly look to their feedback R for stability, wildly different loads will impact their fine scale response flatness, but can normally be made quite stable. 

  • sounds good, I will give it a try and come back to you soon


  • Hi Carl,

    Thank you for keeping us updated.