OK George, I'm happy to take Rod Elliott's assertion: 'Impedance matching is quite accurate, and it will load the transmitter to almost exactly +/-0.25V.' and it's on that basis that I recommend the circuit.
w
w
I can not help about Rod's assertions, I'm going by analysis of the circuit and by the test result confirming it.
It behaves like shown, scale it down to 15% reflection. Not 30, it was my error.
Ciao, George
It behaves like shown, scale it down to 15% reflection. Not 30, it was my error.
Ciao, George
Andrew I beg to differ, if you are goiung to split a 75 ohm co-ax with a y splitter you require 150 ohms at the other ends, and as these are parallel terination they are better at the end, otherwise you are going to have an impedance mismatch. Reference Chapter 6 Terminations, "high speed digital design" Howard Johnson.
We do this stuff all day on PCBs and cables with high speed design, it is what the signal sees as it travels down the cable, track etc that matters.
Have Fun
Marc
Had a word with our resident expert, 150Rs will work for short runs, only way to do it properly for cables is use balums for the best matching of everything.
We do this stuff all day on PCBs and cables with high speed design, it is what the signal sees as it travels down the cable, track etc that matters.
Have Fun
Marc
Had a word with our resident expert, 150Rs will work for short runs, only way to do it properly for cables is use balums for the best matching of everything.
Last edited:
If what you have stated is correct then you are explaining it in a form that can be misunderstood.
The way I interpret what you said does not terminate the coax correctly.
Can you explain more clearly what you mean?
Missed the other posts, woops.
As said earlier I am thinking transmission line, I'll go away and have a think about the cables effects.
On a pcb transmission line if you spilt a track into 2 lines you terminate the far end with 2 150R for a 75R match, if you split it into 4 you use 300R.
As said earlier I am thinking transmission line, I'll go away and have a think about the cables effects.
On a pcb transmission line if you spilt a track into 2 lines you terminate the far end with 2 150R for a 75R match, if you split it into 4 you use 300R.
Reading a bit further, if you use middle terminators (ie terminate at the y split) the figures are Zo/3, which gives 25 ohm. The position of the termination has an effect, which is why I said 2 x 150R at the far ends to gnd.
Last edited:
Marce,
Think of the line characteristic impedance as a real physical quantity.
You take a cable, lets say, 20nsec length. A signal arriving at it will see it like a real (like:resistive) 75ohm load. For a period of 20+20nsec. The termination issues from the other end will not show up at the input before this 40nsec. If you had terminated at the far end then there will be nothing coming back, ever, so it will continue to appear like a resistive 75ohm. If you did not terminate it well, like shorted it to gnd then it will reflect back inverted, and will appear as a short also at the input. After 40nsec.
So, when looking at the star connection of 3 * 25ohm shown by AndrewT, one has to simply terminate each branch with 75ohm.
It's working alright, in reality. We are doing it daily here at the job. Even the reflections can be kept very low, if done with care.
Ciao, George
Think of the line characteristic impedance as a real physical quantity.
You take a cable, lets say, 20nsec length. A signal arriving at it will see it like a real (like:resistive) 75ohm load. For a period of 20+20nsec. The termination issues from the other end will not show up at the input before this 40nsec. If you had terminated at the far end then there will be nothing coming back, ever, so it will continue to appear like a resistive 75ohm. If you did not terminate it well, like shorted it to gnd then it will reflect back inverted, and will appear as a short also at the input. After 40nsec.
So, when looking at the star connection of 3 * 25ohm shown by AndrewT, one has to simply terminate each branch with 75ohm.
It's working alright, in reality. We are doing it daily here at the job. Even the reflections can be kept very low, if done with care.
Ciao, George
Now I had seen your last post.
Star of 3*25R, and connect the cables directly to it, without any "middle" termination.
Star of 3*25R, and connect the cables directly to it, without any "middle" termination.
At last!
We are addressing the splitting enquiry.
If the signal ends up being attenuated too much, only then consider changing gains in the system.
I am not an RF engineer/technician. I am not an electronics engineer/technician. I am not in any way a professional in the electrical industry. I am a Civil Engineer.
I have cobbled together various RF splitters (VHF & UHF) that work without needing extra gain.
TV signals are probably in the uVac range. Splitting these tiny signals correctly so that all viewers in the "block" receive their share of the uV is important. They would soon tell me if the resistor values were wrong.
We are addressing the splitting enquiry.
If the signal ends up being attenuated too much, only then consider changing gains in the system.
I am not an RF engineer/technician. I am not an electronics engineer/technician. I am not in any way a professional in the electrical industry. I am a Civil Engineer.
I have cobbled together various RF splitters (VHF & UHF) that work without needing extra gain.
TV signals are probably in the uVac range. Splitting these tiny signals correctly so that all viewers in the "block" receive their share of the uV is important. They would soon tell me if the resistor values were wrong.
Marce,
It's an interesting approach. What You probably forgot to include is that you cannot do this without changing track impedance.
It could even work in this way: you have your input line, 75ohm track impedance. It meets the split point, from there are starting two branches, but now 150ohm characteristic impedance each track! Then you have to terminate each in 150ohm.
But still it's not valid backwards - any signal coming back down the 150 ohm line will not meet the right impedance, so it's better not to have any reflections in those lines?
So it kind of works, but not properly.
You can easily do it in a pcb, but not with cables. One needs a proper power splitter like the one described by AndrewT.
It's an interesting approach. What You probably forgot to include is that you cannot do this without changing track impedance.
It could even work in this way: you have your input line, 75ohm track impedance. It meets the split point, from there are starting two branches, but now 150ohm characteristic impedance each track! Then you have to terminate each in 150ohm.
But still it's not valid backwards - any signal coming back down the 150 ohm line will not meet the right impedance, so it's better not to have any reflections in those lines?
So it kind of works, but not properly.
You can easily do it in a pcb, but not with cables. One needs a proper power splitter like the one described by AndrewT.
Set up this basic scenarios, as shown, also set up the 3 25R star one with 75R at the reciever ends, which gives a much better result.
Will post some waveforms tommorow, and set the correct drivers (any suggestions welcome).
Finding that with a mimic of a 500mm cable the 3 way with 75 at the end is best. Normally I do PCB's and I am just getting into modeling cables so my appologies basing my replies on what works fine on a PCB.
The cable models are basic at the moment, and no connectors have been modelled yet, though they can be.
As said Andrews method with added 75R to GND gives a very nice waveform at both ends, my option was not as good.
Will post some waveforms tommorow, and set the correct drivers (any suggestions welcome).
Finding that with a mimic of a 500mm cable the 3 way with 75 at the end is best. Normally I do PCB's and I am just getting into modeling cables so my appologies basing my replies on what works fine on a PCB.
The cable models are basic at the moment, and no connectors have been modelled yet, though they can be.
As said Andrews method with added 75R to GND gives a very nice waveform at both ends, my option was not as good.
Once again I have to remind members, I am only the messenger.
I am not trained in any RF.
The information is all in this Forum.
That is why I get annoyed to have to correct folk that try to tell us something that is different from what the real experts tell us.
I should not need to chip in and point out what is wrong. It has all been done before !!!
I am not trained in any RF.
The information is all in this Forum.
That is why I get annoyed to have to correct folk that try to tell us something that is different from what the real experts tell us.
I should not need to chip in and point out what is wrong. It has all been done before !!!
waki, with those resistors output impedance is ~ 73R. Or little more. But DMM will give you 91R ?
input Z?
We have 100R terminating resistor in parallel with 100R + (8K2/Aol+1).
Aol (open loop gain) can be 10...or 20...or 30...or 40...or 50.
In this case, input impedance is somewhere in range 73R to 90R. Not even close to 100R.
But DMM will always show 100R.
We have 100R terminating resistor in parallel with 100R + (8K2/Aol+1).
Aol (open loop gain) can be 10...or 20...or 30...or 40...or 50.
In this case, input impedance is somewhere in range 73R to 90R. Not even close to 100R.
But DMM will always show 100R.
When the power is applied the output of the inverter becomes active and the 360 is in approximately in parallel with the 91 so the output impedance approaches 75.
It doesn't bother you that the '04 is a logic device and changes the pulse widths because the slice point is not well defined? It doesn't bother you that the signal levels are no longer on target?
Andrew, RF levels are a totally different deal because your receive devices have AGC and will restore levels internally. The DAC is not intended to perform this function. Nevertheless, the impedances on the transmission line must be correct to keep VSWR at 1 and add no response 'bumps' an ringing.
You people sit here with hundreds of dollars of PCs and many dollars a month for internet service and then grouse about a $7 IC that actually performs the task properly. Exasperating.
I promise not to pay any more attention to complaints about clock jitters because it clearly means nothing.
G²
Your circuit almost certainly performs better than the one I copied, but there is a difference between engineering and over-engineering. If the OP is of the 'paranoid' school he will doubtless choose to spend the $7, if of a trusting nature he should have the option of saving $6. The jitter in this case is random, and in a comparatively recent study random jitter of 250nS proved inaudible to a panel of expert listeners.
I don't intend to open a debate about the audibility of jitter here, and you should not be discouraged from offering the products of your obvious expertise.
w
stormsonic is just pulling my leg about the impedances and the DMM.
Since I am a certified paranoid schizophrenic with OCD I don't mind spending a few extra dollars for an IC.😀 Especially so since it will do precisely what I need it to.
- Status
- Not open for further replies.