How to Better Wire the BNC Connector and GHz issue.

[deephaven]

Where do you live?

In the UK we have UHF PAL TV which are uses 50 Ohm co-axial cable and is fine for most purposes.

If I remember rightly, 75 Ohm is more common for VHF FM radio receivers.

It's normally 75 ohms for UHF TV too.

[Mechatrommer]

You're procedures are confusing to me, mecha.  Some general comments.
efgh.jpg: I presume you mean the images are from left to right, up then down.
ef
gh
All quoted images seem to have your vertical setting at 1V/div. 

you are right on left right up down. i was annoyed when uploaded the pictures individually, vertical arrangement make me hard to trace. you are right, i tried to make it consistent at 1V/div. sorry for making your brain worked very hard.

e = that's what the output looks like if unterminated, ~ +3.5V.  You cannot get a stable sync with that signal, but you can with f, ~ +1V.
g = that's ~ 7VACpp. 
h = as expected, at ~ half of g, but it looks like almost 4 V, it suggests something is wrong; maybe your resistors of 2x "100" ohm in parallel is <> 50 ohms.

you are right, maybe something wrong with my resistors setup, i was doing some dirty tinkering last night (combined the series and parallel resistors together, 4 resistor overall, 2 series 2 parallel, and short the series to get parallel termination type, its quick and dirty, the same to my lab condition right now thanx for the stability advice.

Your reply below sounds like you've forgotten how to read graticules. 
Image b.jpg is ~7VACpp. 
Image c.jpg is correct, that is 3.5VACpp.

i dont know what happened to me again, i tend to stick to positive side (of the voltage), i've not thinking clearly, urgh!
you are right again, now i have re-arrange my mind 3.5Vpp is 1.75V positive max, and -1.75V negative min. i failed at double checking stuffs

Series    Termination means parallel, as in the image I left for reference.

7.jpg is my definition. after the Vpp definition correction, now it start to make sense. the parallel is the way to go (#2 in 7.jpg)

If you use 100 Hz to avoid ringing, you wont' see the effects of correct impedance matching, like it was demo'd with image e and f.  Its clearest when your cable is 1/10th the wavelength of the test frequency, if you use a 1 meter cable, the test frequency should be >= 35MHz.  Using the square wave from the sync output port at 10 MHz, the harmonics are above that, so you can see the effect more strikingly.

quoted for my reference later

@Hero. me in Malaysia, we use PAL format.

[Zero999]

Where do you live?

In the UK we have UHF PAL TV which are uses 50 Ohm co-axial cable and is fine for most purposes.

If I remember rightly, 75 Ohm is more common for VHF FM radio receivers.

It's normally 75 ohms for UHF TV too.

Yes I was talking nonsense. TVs are 75 Ohm in the UK too.

[alm]

Use pure coax whenever possible.  You'll note some have adapters for BNC to alligator or minigrabbers, or as in your photo, for basic AWG wire.  These can cause issues, so if you make your own keep the non-coax segment as short as is practical; like in those photos a good portion of the cable is still coax.  Its flexible to use them in lower frequency applications so long as you are aware of them; jahonen's recommendation is best; direct soldering of cable into the board.  

Be careful with assumptions. I used a cable similar to the left one once, and shielded the grabber leads because noise pickup was a big deal (I think I was dealing with an amplifier with a gain of 10k, so the input signals were very low level). Bandwidth was not an issue, signals were only 1MHz or so. I found out that touching the 'coaxial' part with my hand had a significant effect on the signal: it was not shielded at all, probably just two parallel wires (didn't cut it open to verify). In the end I used a standard RG-58 BNC-BNC cable and soldered some short wires on a female BNC connector.

One thing you can do with your signal generator, is actually test the effect of various connectors [ banana, clips, alligators, BNC splitters etc., ] against one that is just purely coax; then you'll have some idea what artifacts its causing.  Also, the BNC connector and BNC adapters can produce artifacts too at higher frequencies.  

I don't think 75MHz is fast enough to see minor impedance mismatches (of course you'll notice horrible things like screw terminals). The wavelength is about 3m, so you're only just below the limit for transmission line theory with a standard 1-2m cable.

You would need a VNA or TDR to accurately measure them, or at least a fast step (fast compared to the size of your mismatch), otherwise it's just some lumped capacitance/inductance that will have a minor influence on impedance.

For example, BNC  is useful to about 4-11 GHz, and how high you can take them before reflections become substantial can differential a brand name connector from Amphenol versus no-name Chinese brands that 'look' them same.

BNC 4-11GHz? 4GHz is quite high, 11GHz is completely unrealistic IMO. At 4GHz, I would definitely use brand name connectors.

For practical purposes, about ~ 2GHz is the limit of BNC.

Now you're contradicting yourself.

[Mechatrommer]

For practical purposes, about ~ 2GHz is the limit of BNC.

Now you're contradicting yourself.

i believe he meant the whole setup, ie cable and connector. where the cable only capable of the said BW (4-11? GHz). however, looking back at the wiki by Wim_L

For high power radio-frequency transmission up to about 1 GHz coaxial cable with a solid copper outer conductor is available in sizes of 0.25 inch upwards. The outer conductor is rippled like a bellows to permit flexibility and the inner conductor is held in position by a plastic spiral to approximate an air dielectric.

and

The best coaxial cable impedances in high-power, high-voltage, and low-attenuation applications were experimentally determined in 1929 at Bell Laboratories to be 30, 60, and 77ohm respectively. For an air dielectric coaxial cable with a diameter of 10 mm the attenuation is lowest at 77 ohms when calculated for 10 GHz

[saturation]

Hi, yes mecha@ you read me right.  I agree with alm too, but we have different ways of thinking.  

Its more the manufacturers comments regarding the limits of BNC than mine, a lot depends on the quality of components in the cable build, to reach the upper limits of its capability.  This is where you can separate the cheap connectors & cables from the better ones.

See the links I put for Amphenol, it was they who suggest it can reach 11 GHz.  Here's more:

http://www.mwrf.com/Article/ArticleID/8993/8993.html

Likewise, for cables some can be straight through unshielded wire, or not, or even terminate in 75 ohm BNC connectors!  Caveat emptor.  Again, depends on who makes them.

http://www.amazon.com/Pomona-Male-Alligator-Test-Clip/dp/B0000WU364

As for transmission lines, generally you start including such notions when the cable length is ~ 1/10 the wavelength, so 75 MHz ~ 1.3 foot. shooting a 10 MHz square wave puts the harmonics in this park.  But it can happen sooner for more reasons related to impedance mismatch, as the Hantek sync signal clearly shows, QED.

[ArtemisGoldfish]

great! with rigol probe, its getting more and more complicated. now i dont know which reading to trust.
10MHz square synch out from 3x25. rigol set at 1V/div. notice the termination placement (ie close to the rigol and close to the 3x25)
the most notable difference is at 10X setting with 50ohm resistor termination, i dont know what the hell is happening.
maybe this is called ringing (or the reflection), but it will make it different if we put the termination on the different side, i dont know, maybe i'm just wasting time here.
ps: with 3x25 cable, whether the termination is close to 3x25 or rigol, the shape and amplitud is something like e&f in top part of efgh.jpg

Maybe a silly question, but you have compensated your probes, right?

[Mechatrommer]

looking for coaxial cable, found in chinese ebay http://cgi.ebay.com.my/RF-Coaxial-cable-M17-119-RG174-50-feet-/300426266059?pt=UK_Computing_Networking_SM&hash=item45f2cd05cb there is RG400 which is N/A in wiki, from the spec given by the seller, this RG400 is good to 12.4GHz 2x braid outer diameter 4.95mm. seems a good choice for my 3G antenna cable and for measurement with my BNC connector, any opinion if i should buy this instead of RG174 or RG58? is it flexible instead of semi rigid?

ps: i locally bought 75ohm satellite cable at very cheap price for experimentatin later, but would like to get the best for the money and performance cable. so thinking between rg174, rg58 and rg400. there is higher freq cable, but that is semi rigid and i think it will not very comfortable to work with if longer length cable is needed.

[Mechatrommer]

and i read about TNC connector, and RP TNC. is TNC can be used to connect to Rigol scope? (BNC compatible?). and why there is RP (reverse polarity) TNC? i was looking for my usb broadband antenna connector, there is no TNC-CRC9, just RP-TNC-CRC9 http://cgi.ebay.com.my/RP-TNC-Plug-CRC9-pigtail-antenna-3G-huawei-PCMCIA-/290543854094?pt=Camera_Cables_Cords&hash=item43a5c3620e

[jahonen]

I made an experiment how well plain wire connection will work under impedance matched conditions As you can see, the result is all over the place. What that measurement does not show, is the sensitivity of this bogus connection to even slightest variations of geometry. That means if you move the wires, the frequencies of those deep notches will shift frequency.

I calibrated the setup using a SMA-SMA-extension as a reference. Result is shown in last picture.

Regards,
Janne

[Mechatrommer]

thanx janne for the precious effort. even though i dont have a clear clue on how to do the precise setup, but from the graph, it proved the inconsistency of loose wires. i'm getting a better cable. maybe i misunderstood the RP-TNC version is actually somekind of female (in male) connection. instead of normal male connector. i'm also eyeing on SMA connectors, as my CRC9 only has "to SMA" adapter. i was playing with rf antenna which i was not aware that its actually using sma connector, i didnt know what i was doing . still, can i assume that TNC "plug" is compatible with BNC "jack"? with better performance? TNC is up to 11GHz where BNC is only 3GHz according to wiki.

[jahonen]

The setup was quite simple, tracking generator output of a spectrum analyzer using a semi-rigid coax to "DUT", and then from "DUT" to the RF input of the same spectrum analyzer using second coaxial cable. Spectrum analyzer and tracking generator can be calibrated to null out any irregularity of test setup, which I did, using a short SMA female-female extension.

As for your question about TNC compatibility, TNC is only compatible with TNC connectors, although one can always get adapters.

Regards,
Janne

[alm]

and i read about TNC connector, and RP TNC. is TNC can be used to connect to Rigol scope? (BNC compatible?).

Only with an appropriate adapter, just like any other connector like SMA and N. It's not compatible with BNC, although they're related.

and why there is RP (reverse polarity) TNC? i was looking for my usb broadband antenna connector, there is no TNC-CRC9, just RP-TNC-CRC9 http://cgi.ebay.com.my/RP-TNC-Plug-CRC9-pigtail-antenna-3G-huawei-PCMCIA-/290543854094?pt=Camera_Cables_Cords&hash=item43a5c3620e

They're required to use a 'non-standard' connector so the consumer can't mess with the antenna, reverse polarity connectors are not compatible with standard connectors. The result of the widespread use of RP connectors for wifi is that some distributors have more RP-SMA than SMA connectors.

I made an experiment how well plain wire connection will work under impedance matched conditions As you can see, the result is all over the place. What that measurement does not show, is the sensitivity of this bogus connection to even slightest variations of geometry. That means if you move the wires, the frequencies of those deep notches will shift frequency.

Looks horrible, as expected. Thanks for doing the actual test. Performance is fairly consistent with the expected lambda / 10 rule-of-thumb (looks OK-ish to 100MHz or so).

[Mechatrommer]

ok thanx guys, i will not into trouble with tnc.
i made the 75ohm cable, this is the best i got:

a.jpg = the cable+connector, soldered 50ohm terminator and male-male
b.jpg = measurement of Hantek 3x25 10MHz synch out (top=original cable, bottom=my 75ohm cable, unterminated and terminated).
c.jpg = while play around, noise/harmonics/distortion is picked up in rigol FFT on 5MHz sine function (the same with my and hantek cable), OT.

from b.jpg, maybe the capacitance of my newly made cable is higher than original cable as the higher frequency is rejected as in (left side, unterminated) picture, but i'm not 100% sure.

[saturation]

This looks great.  Your 75 ohm cable empirically has a better response than the factory 50 ohm cable.  But as you can see the termination gets the best frequency response, given the rising edges of your square wave is still better when terminated.  Your cable construction looks great!  I am much more sloppy, sorry to say  .

I noticed those harmonics on your 5 MHz, the top of the sine wave seems clipped for some reason.  I noticed sometimes the Hantek needs a warm up period, believe it or not.  If you turn it on, and just type in a MHz frequency, sometimes it has more distortion.  But if you start with the default, 1kHz, let it output, then slowly ramp up after each  waveform stabilizes before the next increase: 10kHz, 100kHz, 1MHz, then 5 MHz, it produces the best response.  What I do is just type a zero and enter to move up the frequency, then 5 to overwrite the 1 to make 5 MHz.  Not sure why it does this, but it only happens when its started from a cold start, if you let it warm up for at least 10 minutes, you can do as you please! 

ok thanx guys, i will not into trouble with tnc.
i made the 75ohm cable, this is the best i got:

a.jpg = the cable+connector, soldered 50ohm terminator and male-male
b.jpg = measurement of Hantek 3x25 10MHz synch out (top=original cable, bottom=my 75ohm cable, unterminated and terminated).
c.jpg = while play around, noise/harmonics/distortion is picked up in rigol FFT on 5MHz sine function (the same with my and hantek cable), OT.

from b.jpg, maybe the capacitance of my newly made cable is higher than original cable as the higher frequency is rejected as in (left side, unterminated) picture, but i'm not 100% sure.

[Mechatrommer]

better response? hmm, if you say so. i'm not capable of proving that, but thanx. yes, not much difference in terminated cable (which is the way when measuring or real application i think). the trick in sloppiness is in the black rubber, and i'm not intending to show whats inside thanx for the tip for stable output saturation. Cheers

[saturation]

NICE!  Thanks for taking time to do this.  You have top quality connectors, they look gold plated!  One could stabilize the response by twisting the wire to emulated twisted pair, this should put a more repeatable fixed inductance and capacitance between conductors.  Would be interesting to see what kind of response you get.  After getting one set of readings say with 5 twists for that cable length, twist the wires more and double the twists/inch, this should improve the response even more.

I made an experiment how well plain wire connection will work under impedance matched conditions As you can see, the result is all over the place. What that measurement does not show, is the sensitivity of this bogus connection to even slightest variations of geometry. That means if you move the wires, the frequencies of those deep notches will shift frequency.

I calibrated the setup using a SMA-SMA-extension as a reference. Result is shown in last picture.

Regards,
Janne

[Mechatrommer]

i guess thats the idea Mr Oliver Heaviside came up with 1.3 centuries ago.

[jahonen]

NICE!  Thanks for taking time to do this.  You have top quality connectors, they look gold plated!  One could stabilize the response by twisting the wire to emulated twisted pair, this should put a more repeatable fixed inductance and capacitance between conductors.  Would be interesting to see what kind of response you get.  After getting one set of readings say with 5 twists for that cable length, twist the wires more and double the twists/inch, this should improve the response even more.

That was also suggested elsewhere, maybe I'll try that in near future. Also, I thought I'll also try similar setup but with RG-174, that should be considerably better. BTW, I used a SMA torque wrench (0.50 Nm) to tighten the connectors, important thing to get consistent readings At GHz frequencies, even slightest sloppiness at connectors or cables, will show up as a measurement error.

Mechatrommer: Your 75 ohm coax is very short for any reflections to show up for those rise times, try something like 5-10 meters, to amplify the effect. Cable capacitance has nothing to do with frequency response, only thing it affects is the signal velocity in the cable. It is sometimes difficult for people to refrain from substituting a lumped LC-filter in place of the cable, this of course leads to completely wrong results, a transmission line is not a low-pass filter! In real cable (or in any transmission line), L's and C's are distributed evenly along the whole length, so if one wants to replace the cable with L's and C's, it must be done with large number of small L-C-sections.

Oh, I forgot to mention that one can use a TDR technique to determine the cable impedance, as mentioned by others. If you have reasonable length of open-ended coax (or whatever transmission line structure) and feed a fast rise time step of U0 to that from impedance Zs, then what you'll see right at the feed end of the coax is a step function, which has initial height of U=Z0/(Z0+Zs)*U0, i.e. if source impedance is equal to the coax impedance, the initial step will be at half of the final height (it is left as an exercise to reader to solve the expression for Z0). Cable is basically resistance equal to the nominal impedance from source point of view. The voltage will eventually rise to its full height, after the reflected wave from the open end travels back to the source. Of course, this measurement technique requires that you have relatively non-intrusive way to measure the signal at source without significant loading.

Regards,
Janne

[saturation]

Hi jahonen,

I hope you do decide to test with a twist    You can just take both ends of the cable with the connectors mounted and simply twist the whole assembly; the geometry will be more stable that way and it could make a difference.  As you tighten the twists, the geometry will be even more 'compacted' thus stabilizing the conductors position together and thus, the capacitance, at the very least.

Here's an actual measured test of RG-174 vs RG58, seems like its feasible for short cable lengths to make more ergonomic lab cables, given the measured values of these folks.  I think their actual data is close to spec sheet values.

http://www.dxing.info/equipment/rg_174_coax_bryant.dx

As for the torque wrench, I wouldn't doubt it, how did you know to torque it to 0.5Nm?  I found a good supply of Amphenol Connex BNC splitters for the same price of no-name Chinese BNC splitters, and oh my, even at MHz speeds, I can see the difference, albeit its small, its clear approaching GHz speeds, the Amphenol craftmanship has the better chance of getting the proper result.  It would seem for best results too,  you'd need to confirm the cables Zo unless its a good brand premade cable from Pomona, Tek, Agilent or such.

The Zo technique is nifty, its familiar but seems buried in my memory somewhere, I'll give  measuring Zo a whirl.  See the link above, they measured it for each frequency too, but a step function puts all those sines together for a good square dance

[Mechatrommer]

thanx for the link, will get back to it later.

[jahonen]

I tested the twisting, no dramatic effect on the response. It seems that the geometry is still simply too irregular for >700 MHz (quite obviously). I think that even below that, the impedance is probably all over the place.

The torque wrench I used is actually Huber&Suhner 74_Z_0-0-79, 0.45 Nm (remembered wrong), which is "SMA economic, standard style" according to H&S. I'm not sure if the actual tightening torque is absolutely critical as long it is consistent. It seems to vary quite a bit from manufacturer to another (even H&S makes three kinds of SMA torque wrenches, 0.45, 1.00 and 1.95 Nm), and also which kind the SMA connector is. Greatest advantage is IMO that it is easy to tighten the connectors enough without having a fear about damaging the threads with regular spanner (0.45 Nm is really low torque).

Regards,
Janne

[saturation]

Nice, thanks a bunch for giving a whirl, jahonen.  The change in geometry did improve the response, when looking at the curves side by side, certainly not the level of a proper cable.

Its a great demo of the effect of the types of cabling, wires and construction, on GHz transmission.

Will keep a look out for those wrenches. 

I tested the twisting, no dramatic effect on the response. It seems that the geometry is still simply too irregular for >700 MHz (quite obviously). I think that even below that, the impedance is probably all over the place.

The torque wrench I used is actually Huber&Suhner 74_Z_0-0-79, 0.45 Nm (remembered wrong), which is "SMA economic, standard style" according to H&S. I'm not sure if the actual tightening torque is absolutely critical as long it is consistent. It seems to vary quite a bit from manufacturer to another (even H&S makes three kinds of SMA torque wrenches, 0.45, 1.00 and 1.95 Nm), and also which kind the SMA connector is. Greatest advantage is IMO that it is easy to tighten the connectors enough without having a fear about damaging the threads with regular spanner (0.45 Nm is really low torque).

Regards,
Janne

[Mechatrommer]

so how about the correct coaxial cable? or even the 75ohm one? saturation's link only show test up to 20MHz. maybe there are lot of info on the net, but i prefer the one that is done with the same setup same person to better compare but i wont ask too much. i thankful for what i already got.

btw. how actually this test is conducted? i'm assuming the device will give a signal at certain frequency and then check the attenuation on the output, and then increase the frequency bit by bit, check the attentuation and so on until we get the nice graph? is that what is called "frequency sweep"?

[jahonen]

Next, I'll try these. It was quite difficult to spread the braid to four separate pigtails, but I think that the result was acceptable. Length is about the same than in one with loose wires.

Second one is just a SMA-SMA cable (RG-174) which I made a while ago. I expect that to be reasonably good.

Regards,
Janne