The 50ohm Termination: How, Why and When?

[Mechatrommer]

hillarous but quite informative. dso brand/model which has or hasnt 50ohm internal termination. i dont really mind, i treat it as "general knowledge"

[jahonen]

Back to original topic, why not try to simulate if the transmission line needs termination or not? In attached simulation I have made the line length to 1 meter which can be adjusted my modifying the "llen" parameter (100e-2 = 1). er is relative permittivity (3 is reasonable guess for a microstrip), and c is the speed of light in vacuum.

Regards,
Janne

[vk6zgo]

I think that is easier to avoid that frequency thinking altogether when dealing with digital signals and just think about how much physical space will the signal edge take on a transmission line. This works whether we have widely separated fast edges or very high repetition rate signal. It works even with single transitions where the frequency and harmonics thinking breaks down.

So for example if we have a 1 ns edge and put that onto a stripline structure where signal velocity is approximately about half of that of free space, 1.5e8 m/s. So 1 ns edge will take 1 ns * 1.5e8 m/s = 15 cm length of transmission line (yes, you can put several of those edges simultaneously on the line if the line is long enough and/or edge spacing is small enough!). And now if edge length starts to be 10%…20% of total trace length, then one should start worrying about reflections.

For RF things are a bit different. Although frequencies are lower for that 10 MHz system, but the system is more sensitive to reflections. Digital system may well operate happily with just 10 dB return loss (or SWR of 1.74 or |S11| = -10 dB) but that is probably too little for RF power transmission. Imagine a 1 kW transmitter which gets 900W return power due to mismatch.

Regards,
Janne

Sorry Janne,but you had a bit of a "slip of the pen" here!
10dB return loss equates to just 100W return power,and an SWR of approx 1.9:1.
Not marvellous,but it wouldn't cook the Tx!
900 W "reflected power" is just 0.4576 dB return loss.

I used to do this stuff for a living,but I thought Senility might have caught up with me,so I did my calculations with my antique HP22S,& verified them with these charts.

http://www.jampro.com/uploads/tech_docs_pdf/VSWRChart.pdf
http://www.minicircuits.com/app/DG03-111.pdf

[jahonen]

Sorry Janne,but you had a bit of a "slip of the pen" here!
10dB return loss equates to just 100W return power,and an SWR of approx 1.9:1.
Not marvellous,but it wouldn't cook the Tx!
900 W "reflected power" is just 0.4576 dB return loss.

I used to do this stuff for a living,but I thought Senility might have caught up with me,so I did my calculations with my antique HP22S,& verified them with these charts.

Damn, my last minute edit was wrong after all, my first thought was actually 100 W, like you said

Regards,
Janne

[vk6zgo]

Before free_electron & I had our little disagreement,he,& others, had made the points that:

(1) Two Op Amps operating at 10MHz.& around one inch apart,one driving the other, do not need to be terminated in 50 Ohms(or any other specific value of resistance).

(2)Until the distance between them reaches approx 1/10 wavelength,( in this case 1/10 of the wavelength of a 10MHz signal is 3 metres in free space,a bit less in a transmission line) ,the connections do not constitute a transmission line.

(3)The output impedance of an Op Amp in many cases is very low,so that if a 50 Ohm resistor is connected in line with the output,the apparent output impedance to any circuit driven from that point becomes so close to 50 Ohms that the difference may be ignored.

I think that (3) is the "sticking point" for Mechatrommer,in that it is difficult to visualise an amplifier with nearly zero Ohms output impedance.
I had the same problem,many years ago,when I was first introduced to Video Amplifiers,with their 75 Ohm output impedance.
Here is a link to a site with practical schematics of Video Amplifiers which use the technique in (3).

http://cds.linear.com/docs/Application%20Note/an57fa.pdf

These are 75 Ohm devices,but the concept is the same.

[free_electron]

Oh, it's not a disagreement. It's a matter of perspective. I Can't remember ever using a scope that did not have that capability (and yes the 54645 does not have it. I remembered that wrong. I'm not using that very often. My main staple in the hobbylab is the MSO7104 and the 54832D)
And vk6zgo can't remember ever using one that has.

So it's proven once more that you can't make assumptions on what is 'the norm' based on what you do or use.

Anyway. One comment on opamp output impedance.
A typical opamp will have between 100 ohm and 30 ohm. For a high current opamp it can be lower.
How do you measure it ?
Put opamp in unity gain,Connect input of the opamp to ground ( this is assuming dual supplies. I don't like the usage of 'ground' . The output should be parked at 1/2 the power supply is a better definition.)
 From the output of the opamp, connect a resistor (100 ohm) to a signal generator. Yes from the output. Apply 1 volt pp. measure voltage before and after the resistor.
You have just made a voltage divider. You know two voltages. Go do the math.

Oh, and note that, as frequency increases, the phase will start changing... And that the impedance is not constant either...

[Mechatrommer]

I don't like the usage of 'ground'

i'll prefer "ground" as a meaning for 0V, and "reference" as a meaning for "other potentials", including 0V, if you like
and as ack: opamp output impedance is not constant and changing based on frequency. the stated output impedance in datasheet is only for prefered freq, usually near or is DC. that will make life more miserable

[alm]

(2)Until the distance between them reaches approx 1/10 wavelength,( in this case 1/10 of the wavelength of a 10MHz signal is 3 metres in free space,a bit less in a transmission line) ,the connections do not constitute a transmission line.

I think you may have a problem with senility after all, this should obviously be 30 meter in free space .

[free_electron]

it's not senility. neuron misfires , commonly known as brainfarts.

Why does the robot take the long way around ? They're afraid of short circuits...

[gregariz]

1) if i have 2 high speed opamp very close to each other, should i terminate the output of the 1st amp before feeding to the 2nd amp? like in the 1st picture below?
2) how long coax cable then termination is required? or i should always terminate it no matter how short it is?
3) if i have an opamp output directly feed to (1Mohm || 15pF) dso using BNC connector (no coax cable), should i terminate the opamp output (on the pcb) to avoid dso looks capacitive or displaying odd things on the screen? (2nd picture)
4) i read that by using termination, the coax cable will looks purely resistive (the capacitance of the cable and dso can be neglected, at least thats what i understand). can someone please explain in more elaborate (with math formula if necessary) way. why the capacitance become negligible?

5) more confusing combination, eg:
a) 50ohm series at source, 50ohm terminate at load end (1/2 attenuation)
b) 50ohm series and terminate at source, 50ohm terminate at load (1/3 attenuation)
c) 50ohm series and terminate at source, 50ohm series and terminate at load (1/5 attenuation)
d) 450ohm series at source, 50ohm terminate at load end (1/10 attenuation)
e) 950ohm series at source, 50ohm terminate at load end (1/20 attenuation)
f) etc...
why all the different combination and when best to use which? how come using 450ohm (not coax characteristic) also result in very high freq desirable respond?

too much to ask? well i wish i can post more, but this is all for now.

With regard to 1. its worth noting that typically its quite rare to be able to have full control of the input and output impedance of any amplifier whilst maintaining all other specs you may need ie gains, noise etc. This is because they are often competing requirements. They don't teach this very well in school but when you cascade circuits you have the choice of either impedance matching between sections as you are thinking or adding another section to the circuit to make up any gain loss due to mismatches. Typically designers would simply add another stage, particularly back in the days of transistor design when matching transformers were expensive. the same holds true today however matching sections will likely cost more than another opamp. Its not uncommon to have variable output and input impedances in amplifiers. Ways around this include using a buffer circuit such as an emitter follower or implementing negative feedback.

With regard to all the other points its worth noting that any transmission line only operates as a transmission line (reasonably lossless) when it is terminated at each end by the characteristic resistance, typically 50 or 75 ohms. As soon as the resistance is taken away from this value or goes cap or inductive you will get a standing wave on the outside of the coax and your coax will behave like a complex impedance transformer. You will typically notice this on your scope if when you grab the coax you disturb the impedance of the cable and your measurements change. Placing a 50 ohm termination on a T connector at the input of your scope fixes that end. You'll need to impedance match the other end if your active output impedance of your circuit is not 50 ohms.

hope that helps a bit

[vk6zgo]

Oh, it's not a disagreement. It's a matter of perspective. I Can't remember ever using a scope that did not have that capability (and yes the 54645 does not have it. I remembered that wrong. I'm not using that very often. My main staple in the hobbylab is the MSO7104 and the 54832D)
And vk6zgo can't remember ever using one that has.

So it's proven once more that you can't make assumptions on what is 'the norm' based on what you do or use.

Anyway. One comment on opamp output impedance.
A typical opamp will have between 100 ohm and 30 ohm. For a high current opamp it can be lower.
How do you measure it ?
Put opamp in unity gain,Connect input of the opamp to ground ( this is assuming dual supplies. I don't like the usage of 'ground' . The output should be parked at 1/2 the power supply is a better definition.)
 From the output of the opamp, connect a resistor (100 ohm) to a signal generator. Yes from the output. Apply 1 volt pp. measure voltage before and after the resistor.
You have just made a voltage divider. You know two voltages. Go do the math.

Oh, and note that, as frequency increases, the phase will start changing... And that the impedance is not constant either...

Well caught,I should have referred to the Op Amp  used in a particular circuit,not the output impedance specified in the spec sheet.
The effective output impedance does reduce to a much lower value,or the  Video Amplifiers in the link would not work.

The method of measuring output impedance works with any amplifier,though the first time I was confronted by it in the old Perth Tech School back in the '60s it freaked me out a bit!

[vk6zgo]

(2)Until the distance between them reaches approx 1/10 wavelength,( in this case 1/10 of the wavelength of a 10MHz signal is 3 metres in free space,a bit less in a transmission line) ,the connections do not constitute a transmission line.

I think you may have a problem with senility after all, this should obviously be 30 meter in free space .

I think it effects reading comprehension too!
What is 30/10 ?

[free_electron]

national semiconductor has a nice app note on measuring output impedance. don't rmember the exact number. hang on ....

this is a good one on phase margin and how to measure it:
http://www.national.com/assets/en/appnotes/AppBrief108.pdf

can;t find the 'opamp measurement cookbook. it was appnote 23 or something like that. a big fat appnote with all kinds of practical circuitry...
dang it , where's those blue books when you need them ... ( i still have the whole collection of blue national semiconductor appnote books on the shelf at home... )

[vk6zgo]

Shame the Engineers in the dear old People's Republic who designed the Transmitters we had at my last work didn't read this App Note.
They used an LM358 in a diff amp configuration to control the level of RF drive to a 1kW RF amp.
Due to a capacitive load,at certain power settings,it would "take off"at 38kHz,Amplitude Modulating the drive signal to 100%+
This really upset the switchmode power supplies!
We eventually fixed it "brute force" with a 220 Ohm resistor in  line with the 358's output.
Luckily,the Transmitters were used in an ISM application so we didn't put this crud to air!

[alm]

(2)Until the distance between them reaches approx 1/10 wavelength,( in this case 1/10 of the wavelength of a 10MHz signal is 3 metres in free space,a bit less in a transmission line) ,the connections do not constitute a transmission line.

I think you may have a problem with senility after all, this should obviously be 30 meter in free space .

I think it effects reading comprehension too!
What is 30/10 ?

Could you speak up please? My hearing aid is broken because a damaged coaxial cable is causing an impedance mismatch at high audio frequencies.

You're of course completely correct.