EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: Mechatrommer on March 31, 2012, 11:40:47 am
-
say we have high impedance source signal. we want to probe with oscilloscope (15pF || 1Mohm input impedance) through 50ohm coax. and i have control to put parallel capacitor for compensation.
i tried to do compensation from a high impedance (1Kohm) source (from c1) for coax cable about say 3 foot (modeled by 3xRLC, is it right?) but changing compensation capacitor (50pF paralled to 1Kohm impedance) will come out with funny bode plot. i tried lumped RLC which give nicer bode attenuation, but...
1) how to properly compensate this?
2) is my coax cable model right? 0.1ohm, 0.5nH, 30pF per foot? (3foot = 3xRLC)
3) can i model coax with just one lump RLC network? (iirc i heard from one high speed transmission line guy here we cant)
4) any better idea or correction here?
???
thanks.
-
Why not make a voltage divider across the high Z source,with 50 Ohms as the lower section of the divider?
Or it you want to be more high tech, make a multisection impedance matching attenuator (pad)?
Then just use the coax with a 50 Ohm inline through termination at the Oscilloscope end.
The cable will be operating at its characteristic impedance,so it should not need any corrections,other than for the fact that its loss will increase with frequency.
This will give better results than trying to correct for impedance mismatches with strange RC networks.
The downside is that your 'scope may not be sensitive enough to give a useable display.
-
Do not try to simulate a coax with L's and C's, you'll need too many of them to model work properly. Transmission line is not a low-pass filter. Use a proper transmission line component instead. I don't know about TINA but at least LTSpice has one. Matching requires that coax sees 50 ohms at either driving or load end (or both, the best case), otherwise there is no match possible.
If you care about DC loading only, then you could do an AC termination at scope end, i.e. 50 ohm resistor in series with a suitable capacitor for the frequency of interest.
Regards,
Janne
-
Build yourself a simple FET probe. What sort of bandwidth do you need?
-
Build yourself a simple FET probe. What sort of bandwidth do you need?
this is somewhat relevant. fet opamp output, will feed signal to coax to dso. this is my OP question ;) except... 1/10x probe (9M divider) is my last option. i try find solution for 1X probing first. or maybe it is just impossible (i guess thats why 1/10X probe were invented) :( ps: actually my signal is already attenuated to some degree, so i dont think further attenuating it for the sake of coax is a better idea. i'll study further, thanks guys for the link, i will need time for that.
edit: forgot to answer, bandwidth.... 100MHz.
-
A SPICE Simulation? Wrong tool for the job. For the n'th time, SPICE if for analyzing, not for synthesizing of a circuit.
The trusted and tried tool for doing the necessary calculations to match one or more impedances is the Smith Chart (http://en.wikipedia.org/wiki/Smith_chart)
Also available in software. http://jcoppens.com/soft/linsmith/index.en.php (http://jcoppens.com/soft/linsmith/index.en.php) And of course, these days the necessary calculations can also be done without the graphical representation in the chart. Although the chart helps to keep focused.
A tutorial how to perform impedance matching with the help of the chart http://www.maxim-ic.com/app-notes/index.mvp/id/742 (http://www.maxim-ic.com/app-notes/index.mvp/id/742)
i will immortalized this good post so it will not come and go until found guilty :P
i thought Smith is only for "really serious RF". i need time for this. thanks for advice.
-
i also figured out (from observation only) that this compensation technique (parallel capacitance) like the 1/10X probe did is somewhat in my term... "cheating". though this form 1/10X divider throughout the bandwidth, but at very high frequency, the divider's impedance will be so low it will load the source. i stand to be corrected.
-
I think you are trying to re-invent what Doug Ford has already figured out:
http://www.dfad.com.au/links/THE%20SECRET%20WORLD%20OF%20PROBES%20OCt09.pdf (http://www.dfad.com.au/links/THE%20SECRET%20WORLD%20OF%20PROBES%20OCt09.pdf)
I'd also like to point out that Smith chart and associated matching methods are usually only useful for a narrow frequency bandwidth around relatively high center frequency, i.e. RF applications. For wideband stuff, only thing what works is resistive matching pads, but unfortunately those are lossy.
Regards,
Janne
-
yes i think i've read about doug ford and other high voltage divider probe. but since this is somewhat different, ie attenuated signal to be sent 1/1X over coax (1Kohm impedance spec, not 9Mohm), so i'm looking at another round of way. but maybe i'm in the wrong direction here. review on the "square" table should be in order. i need to re-read the doug's doc more closely.
-
If you want to get up to 100MHz bandwidth over a meter or more standard 50 ohms coax, you have to terminate the 50 ohm coax with 50 ohms. That means that if you have a 1K resistor on the input, the signal on the output will be divided by 20. If you attempt any other solution, it is just going to be ugly. In a divide by 20 mode, with a very tiny capacitance in parallel with the 1K resistor, you can get good performance to over 500MHz at least.
Sure oscillosope probes seem to get away with it, but they do not use standard coax, and I do not know where you can buy the specialized resistive coax cable used in probes.
Richard.
-
yes. i read about doug's doc, its not what i think it is. reading closely reveals this resistive coax. now hunting is in order. anybody knows where to get this? at least shipping to asia? 1/20X divider now thats the problem. my signal is attenuated to 1/50X, i cannot amplify it its already the opamp fullswing. with this 1/20X divider, it will make it 1/1000X system/scope. so much for a small (less than 1KV) signal. ???
janne mentioned about resistive pad matching, not sure what it is. anyway, does anybody know other supplier that sale resistive coax ship to asia, pleaseeee?
-
i think this will be a good read on 3 type of probes (normal 1/10x, fet and 1Kohm resistive)...
http://www.signalintegrity.com/Pubs/straight/probes.htm (http://www.signalintegrity.com/Pubs/straight/probes.htm)
and this catalogue shows all fanciness of cute little coax and variety, but not sure how to get this kind of supply. from hewtech.com.japan...
http://www.hewtech.co.jp/products/pdf/high_frequency_e.pdf (http://www.hewtech.co.jp/products/pdf/high_frequency_e.pdf)
1Kohm probe seems promising with viability up to GHz range, passive. i think i've come with a solution for this, but the cost of probing system will be increased, ie both active at both source and scope end. :|
-
Just so I understand your question.
You want to use a digital Storage Oscilloscope (DSO) to measure the output of a hi z signal source. I assume that you are concerned that the 50 ohm coax will cause a mis match with the output of the signal source? leading to bad measurements?
I think you are up for a Z match. This should not be too hard if the output impedance of the signal source is known. How did you get the value of the output Z. Is it calculated or is it specified by someone else? If there is no complex component a resistive match is not too hard. If there is a complex component in the output Z it gets a bit more complex (pardon the pun) try a conjugate match if you know the complex component.
Lot of factors here as in what you are trying to achieve, an absolute measurement as in 2V RMS at a particular frequency with a known accuracy or do you want to do relative measurements like a rough bode plot? Trying to RF measurements with home brew equipment that are accurate is not something for the feint hearted.
Good luck and let us know how you went
regards
SW
-
my intention is something similar to fet probing. ie i have an (instead) attenuated signal out of opamp. and "i alone" specified the output to be 1Kohm, since if i put 50ohm, it will be loaded by dso at high freq increasing the opamp's output distortion. i cant afford cheaply by further attenuating the signal by using probe's divider since it will reach the dso at 1/500X - 1/1000X and thats nasty for a small signal (minimum 2V/div graticule?). yes i need to
matchcompensate the coax to avoid bad/funny measurement. there's no complex component, just an opamp and a 1Kohm resistor that i want to put on the output to be specified as its maximum load. and also maybe i've put the word wrongly, maybe its just compensation that i want to achieve, not "matching". sorry for that confusion.
i know this is funny and embarassing (and maybe highly nontechnical), but for the sake of learning, i've assembled 1Kohm source (feed from cheapo FG) feed through 2 feet 50ohm coax, 50ohm "noninline" termination (poorman's method), the resulting bode shows (2nd picture) the cable only speced as somewhere 70MHz (3db down the 26db 1/20X attenuation) i believe, and thats without parallel caps yet (quick assembly), maybe if i have proper pcb circuit and adjustable caps, i might get better respond? i hope :-|
other setup such as nonterminated coax resulting at merely just over 1MHz BW (4th pic). cheating a bit, setting 10X at dso, showing 2nd picture (-26db) at 10X (-6db) in 3rd picture, still showing 70MHz BW, so i hope its not cheating as well when i'm doing active amplification at the dso end later.
-
my intention is something similar to fet probing. ie i have an (instead) attenuated signal out of opamp. and "i alone" specified the output to be 1Kohm, since if i put 50ohm, it will be loaded by dso at high freq increasing the opamp's output distortion.
Why would the scope load it at high frequencies? Are you talking about the capacitance that will be in parallel with the 50 ohm termination? The bandwidth of your scope (Rigol DS1052E?) is 100 MHz when 'hacked'. The impedance of the 15 pF you quoted at 100 MHz is ~100 ohm, resulting in a response that's -33% down. Not that much a worry for instruments rated at -3 dB at that frequency, I would think. Scopes with a higher bandwidth will usually have real 50 ohm inputs.
What kind of frequencies and voltages are we talking about? You would probably accomplish more by stating the problem instead of your proposed solution.
What's wrong with a regular 10x or even 1x high-impedance probe? You insist on using a 50 ohm cable, do you need more than the 3m of cable that regular probes offer? I don't think the resistive coax is easily available, and you would basically be designing your own high-Z probe. Be sure to read the Tektronix Circuit Concepts book 'Oscilloscope Probe Circuits' (should be easy to find as PDF/DJVU) if you intend to go this route, it was published way back in ~1960, but this is by far the best reference on this topic. Much better than Doug's article in my opinion. Note that the center conductor is very thin and needs some aggressive flux to solder it (some people use Aspirin, make sure you don't inhale the vapors).
If you insist on 50 ohm cable and your circuit can not drive a 50 ohm load, you can either go for passive (resistive) or active. Passive probes usually have a low DC resistance and high attenuation. That 1 kohm resistive probe will attenuate 20x. Active probes, which have a FET or op-amp driving the 50 ohm cable, don't have this problem, but are more complex, especially high bandwidth. There are various DIY designs floating around. Bob Pease published one in his Troubleshooting Analog Circuits book (not sure if this one has a 50 ohm output impedance). Jim Williams published one in one of the LT appnotes, and there is also a more modern Elektor design floating around.
Cable length should not be critical when terminated in its characteristic impedance, although eventually skin effect losses become a problem as the frequency goes up. Commercial FET probes compensate for this.
If it's your own circuit, adding a buffer capable of driving a 50 ohm load would be the easiest solution.
-
i have all the docs you've mentioned, including the vacum tube active probe tekprobe.pdf. and i thought i've mentioned my problem clearly. maybe lets put it another way...
i have a signal A, my circuit will (and only can) produce 1/50X of A. i want the scope to show A. i want the dso -3db bandwidth will set the limit, not the transmission line (coax) which make the whole system BW = sqrt(probe^2+dso^2) how? and to make thing worst. i've set the output impedance of the circuit as 1Kohm.
or maybe another way...
i want to build a fet probe, but instead of normal 1/1X buffer fet probe 50ohm output, i'm building 1/50X fet probe 1Kohm output (for some reason). now for 1X fet probe, thats easy, we can do whatever attenuation down the coax, the scope has the multiplying setting 10X, 100X. what coax attenuation i can afford with this 1/50X signal? 1Kohm output impedance?
fair enough, seems i cant get rid of 50ohm termination (like all people always said) or normal 1/10X Zo probe or 1/20X resistive probe, so be it :P i guess.
-
out of curiosity. OT but i've plotted rigol probe at 1/10X compared to direct FG connection. peculiar but maybe coincidental 3db peaking at 30-50MHz. any idea why? or just maybe measurement error, not sure which.
-
I would imagine that is HF probe compensation. The intention is that it compensates for any falloff in other parts of the measurement circuit.
-
...in other parts of the measurement circuit.
you mean the oscilloscope?
-
In the scope or more likely the probe. It's fairly common is you're trying to push a system for maximal bandwidth and overcompensate a bit, to get peaking like that. Some probes include trimmers for high frequency compensation to adjust this.
If you can't do anything about the source being high impedance, the best solution probably is to put an high input impedance buffer (FET input, 50ohm out) right at the output of the source and send that through the coax to a 50 ohm inline terminated coax to the scope.
-
, i'm building 1/50X fet probe 1Kohm output (for some reason).
oh my, we're being fuzzy again aren't we ? why the 1K output impedance. you are driving from a THS3062 opamp. that hing has no problem driving a 50 ohm terminated pathway.
1 meg scope probes are made to be terminated into 1 meg...
10 meg scope probes are also made to be terminated into 1 meg...
terminate them in 50 ohm and all bets are off.
-
that hing has no problem driving a 50 ohm terminated pathway.
yeah, you should put your finger on it when its steadily running the 100ohm load. THS3062? i burnt few of them already ;) just like that! (snapping fingers)
1 meg scope probes are made to be terminated into 1 meg...
50ohm characteristic impedance coax cable need to be terminated with 50ohm (its char Z) in order to put the reactance away and away.