The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> Series termination conundrum
https://designers-guide.org/forum/YaBB.pl?num=1355242519

Message started by Matthias_H on Dec 11^th, 2012, 8:15am

Title: Series termination conundrum
Post by Matthias_H on Dec 11^th, 2012, 8:15am

Dear all,

Over night, I found myself in the middle of a project that requires somewhat more careful RF design than what I'm used to doing on a day-to-day basis (rendering algorithms). Maybe some of you experts can help me out.

My problem doesn't sound all that hard: I would like to take the LVCMOS output of a function generator (Novatech 409B/02; click for spec sheet) and pipe it through a coax cable of 4-6 meters length into a TI CDCV304 clock buffer for further use. The signal is a square wave between 10 and 150 MHz.

According to the spec sheet, the LVCMOS output has an impedance of 50 ohm. The manufacturer further specifies "V_oh >=2.4V and V_ol <=0.4V when series terminated". Given a 50 ohm BNC cable and how I understand series (source) termination, wouldn't this mean that no additional series resistor is needed? See for instance this image here (click image for source article):

So with an output impedance of 50 Ohm, the series resistor should be R = Z₀ - 50 ohm = 0 Ohm, right? However, when I run a 100 MHz square signal through said long BNC cable into a 500 MHz, 2Gs/s scope, I see severe signal distortion until I activate an additional 50 ohm termination on the scope's end. The wave will then be very clean but at an amplitude reduced to less than 1 V p-p.

Any ideas how I can get this sorted out? I appreciate any hint, however slight. Thank you very much in advance!

Cheers,
Matthias

Title: Re: Series termination conundrum
Post by wave on Dec 11^th, 2012, 11:26am

Sounds like you have 'discovered' characteristic impedance matching. (50 ohm source, 50 ohm termination).
With a lot of hand waving, the "50 ohm" cable becomes invisible.

Unfortunately optimal matching is always subject to "maximum power transfer theorem"; effectively a 3dB loss, appearing as a voltage attenuation.

Sounds like you are on the right track. Just search these topics if you want a deeper understanding.

:)

Title: Re: Series termination conundrum
Post by RFICDUDE on Dec 11^th, 2012, 5:44pm

Hi Matthais,

I think your interpretation is correct.

And it makes sense that the oscilloscope produces a clean waveform when its input is 50 ohms. The big question is "What is the amplitude of the waveform from the generator when the oscilloscope is directly connected to the LVCMOS generator and the input impedance of the scope is 50 ohms?"

The difference between the 50 ohm measurement with and without the cable should be an indication of the loss of the cable if the input signal is a single frequency square wave.

If the signal is random data, then determining the loss is slightly more complicated because the signal is spread out over frequency and the cable loss generally increases with increasing frequency. You would need to look at a FFT of the data (hopefully the scope has a fft function) to see if there is some slope over frequency reducing the amplitude of the signal when the cable is used.

Title: Re: Series termination conundrum
Post by aaron_do on Dec 11^th, 2012, 6:24pm

Hi,

from everything you wrote, it sounds like your cable loss is around 8 dB which is a bit ridiculous I think. Maybe your scope settings are incorrect?

Aaron

Title: Re: Series termination conundrum
Post by Matthias_H on Dec 14^th, 2012, 9:33pm

Postscriptum: The manufacturer of the signal generator, has recommended me another 50 ohms in series on the load end. I guess it makes sense this way: of course, the transmission needs to be terminated on both ends. Works like a charm now. Thanks to all responders!

Title: Re: Series termination conundrum
Post by loose-electron on Dec 18^th, 2012, 12:50pm

instead of running a square wave down the cable you may want to also consider sending a sinusoid instead, and "square it up" at the far end.

Fundamental instead of all the odd harmonics will make for an easier signal to transport.