Quality 50 Ohm coax brand ?

[AlexResch]

I apologize for the decidedly incompetent question to follow :

I have a couple of different pieces of coax ( 50 Ohms ) with BNC connectors both ends. When I connect the SA to my sig gen I get different results ( power ) depending on the cable used. Is there a coax which for all intends and purposes is "perfect" ?
And reasonably flexible ? 

[hagster]

All coax is lossy. And gets worse at higher frequencies.

Thicker coax is better. (because the central conductor is thicker too)

Solid core is better than stranded.

Foil shealded is better than stranded.

Solid shield is better again.

Air dielectric is slightly better than foamed PTFE.

Waveguide is better still.

Basically,  if you pardon the innuendo. The thicker and stiffer the better. (So long as it doesnt permit higher order propagation modes)

[taydin]

I have recently bought a set of cables from Pasternack. Didn't have to give up and arm and a leg, and I'm measuring consistent loss when I test a number of cables of the same type.

[AlexResch]

All coax is lossy. And gets worse at higher frequencies.

Thicker coax is better. (because the central conductor is thicker too)

Solid core is better than stranded.

Foil shealded is better than stranded.

Solid shield is better again.

Air dielectric is slightly better than foamed PTFE.

Waveguide is better still.

Basically,  if you pardon the innuendo. The thicker and stiffer the better. (So long as it doesnt permit higher order propagation modes)

Great innuendo !

Being horribly incompetent I thought that for a 30" length the quality of the cable doesn't matter OR it's predictable one way. It seems not. I take the RF from an Anritsu 3642 into a Siglent SA and each cable does something else and does it all over. Should I use a very short cable - like 15" or so ? Could you recomend a brand/specific product ?

[emece67]

.

[Richard Crowley]

It would be helpful to mention the frequency (or frequency-range) at issue here.

It would also be helpful to have some description of your two cable examples.
Is there any name and/or model or spec number on your cables?
Assuming that both of your cables use 50 ohm BNC connectors?

Is it possible that one of the cables is damaged?
Perhaps kinked previously and damaged internally with no external visible evidence?

[Xnke]

Do your 50 cables have 75 BNC's fitted? This is not uncommon, unfortunately.

[AlexResch]

I have recently bought a set of cables from Pasternack. Didn't have to give up and arm and a leg, and I'm measuring consistent loss when I test a number of cables of the same type.

1. Which cable is that ? I saw Pasternak has a lot of cables - it confuses idiots such as myself. The attenuation is just a figure like 4db. Is that per meter ?

2. How does one measure the loss ?

[AlexResch]

Do your 50 cables have 75 BNC's fitted? This is not uncommon, unfortunately.

No, I checked that and was aware of the problem. The BNC connectors might not be the best but I doubt they're complete junk - got them from RS.

[AlexResch]

It would be helpful to mention the frequency (or frequency-range) at issue here.

It would also be helpful to have some description of your two cable examples.
Is there any name and/or model or spec number on your cables?
Assuming that both of your cables use 50 ohm BNC connectors?

Is it possible that one of the cables is damaged?
Perhaps kinked previously and damaged internally with no external visible evidence?

Freq I looked at is what my sig gen can do : from around 100Khz to 2Ghz. I don't think the cables are "injured" and the connectors are 50 Ohms or at least that's what's written on the pouch. I just made a new cable - same thing. Not two cables are the same.

[G0HZU]

Even if you could buy perfect 50R cables that had 1:1 VSWR and were lossless you can still expect to see fairly significant variations in the level displayed on the analyser if the (perfect) test cables were different electrical lengths. This is because of the mismatch uncertainty in your overall system.

i.e. a system comprising your Anritsu sig gen, the cable(s) and the Siglent spectrum analyser. Looking at the specs for your sig gen and spectrum analyser you could typically see +/- 0.6dB mismatch uncertainty depending on cable length. This assumes you are doing tests up at VHF or UHF rather than at very low frequencies where differences in cable length are fairly irrelevant. So one perfect lossless cable might show -0.6dBm on the analyser and a longer or shorter (but still lossless) cable might show +0.6dBm. That's a change of 1.2dB on the analyser even with perfect cables.

[AlexResch]

Even if you could buy perfect 50R cables that had 1:1 VSWR and were lossless you can still expect to see fairly significant variations in the level displayed on the analyser if the (perfect) test cables were different electrical lengths. This is because of the mismatch uncertainty in your overall system.

i.e. a system comprising your Anritsu sig gen, the cable(s) and the Siglent spectrum analyser. Looking at the specs for your sig gen and spectrum analyser you could typically see +/- 0.6dB mismatch uncertainty depending on cable length. This assumes you are doing tests up at VHF or UHF rather than at very low frequencies where differences in cable length are fairly irrelevant. So one perfect lossless cable might show -0.6dBm on the analyser and a longer or shorter (but still lossless) cable might show +0.6dBm. That's a change of 1.2dB on the analyser even with perfect cables.

What confuses me here is the following extract from the sig gen manual : "High resolution in level setting
The output level can be set with a resolution of
0.01 dB over the entire level range permitting these
synthesizers to be used as calibration signal sources
for standards such as power meters."

Forgive my ignorance but how can I use the sig gen to calibrate a power meter ? Or the SA for that ? Is there some step in which the cable is somehow "qualified" ?

[rx8pilot]

Even for low-ish frequencies where loss is not a huge concern - I really like the physically more robust cables. I have not had any luck getting the random eBay cables. Pasternack is at least predictable and overall good quality options for general purpose bench cables. Most of my bench work that needs patch cables has been <1Ghz and usually <500Mhz. The RG142 is reasonable performance, and still reasonably flexible. Covers a lot of applications.

https://www.pasternack.com/bnc-male-bnc-male-rg142bu-cable-assembly-pe3495-36-p.aspx

The final call on quality and construction is solely based on your task. The more challenging your task, the more money you need to put into your cables. The price rises very rapidly as the electrical performance goes up. The quality of the terminations and connectors becomes critical at some point. It can get really outrageous.

[G0HZU]

What confuses me here is the following extract from the sig gen manual : "High resolution in level setting
The output level can be set with a resolution of
0.01 dB over the entire level range permitting these
synthesizers to be used as calibration signal sources
for standards such as power meters."

Forgive my ignorance but how can I use the sig gen to calibrate a power meter ? Or the SA for that ? Is there some step in which the cable is somehow "qualified" ?

It's best to look at the specs for the port VSWR for your sig gen. It looks like the VSWR can be worse than 2:1 at sig gen levels in the 0dBm region. Your analyser probably has a typical VSWR spec of 1.5:1 with >=10dB internal attenuation selected. So that would give about +/- 0.6dB mismatch uncertainty.

You can improve things by using good quality external RF attenuators (to improve the port VSWR) or maybe select a lower power level from the sig gen where the output VSWR will be better because of the higher setting of the internal RF attenuator.
 

[G0HZU]

Forgive my ignorance but how can I use the sig gen to calibrate a power meter ? Or the SA for that ?

If you wanted to check the response of your analyser (with a levelled output with low mismatch uncertainty) it might be possible to configure your Anritsu sig gen to provide external ALC levelling as per the diagram in reply #9 in this thread here:

https://www.eevblog.com/forum/rf-microwave/whats-the-difference-between-rf-splitter-divider/

If you make a homebrew 3GHz version of a splitter like the 11667B you could use that system to check the amplitude response vs frequency of your spectrum analyser. The feedback forces port 1 of the splitter to look like a voltage source so the problems with mismatch uncertainty are almost removed if you use a direct connection (no cable) from the splitter to the analyser input. The power meter in the ALC feedback does the levelling and the quality of the 50R resistor in the splitter provides a source impedance with very low VSWR.

But you would have to make sure your sig gen can support this. It might be possible to fudge it in DC coupled AM mode if you invert the slope of the power meter recorder output. This assumes your sig gen has this DC coupled AM mode.

You can see in reply #9 in the link above that I managed to get a fairly flat response from 10MHz to 3GHz using a setup like this

[AlexResch]

Forgive my ignorance but how can I use the sig gen to calibrate a power meter ? Or the SA for that ?

If you wanted to check the response of your analyser (with a levelled output with low mismatch uncertainty) it might be possible to configure your Anritsu sig gen to provide external ALC levelling as per the diagram in reply #9 in this thread here:

https://www.eevblog.com/forum/rf-microwave/whats-the-difference-between-rf-splitter-divider/

If you make a homebrew 3GHz version of a splitter like the 11667B you could use that system to check the amplitude response vs frequency of your spectrum analyser. The feedback forces port 1 of the splitter to look like a voltage source so the problems with mismatch uncertainty are almost removed if you use a direct connection (no cable) from the splitter to the analyser input. The power meter in the ALC feedback does the levelling and the quality of the 50R resistor in the splitter provides a source impedance with very low VSWR.

But you would have to make sure your sig gen can support this. It might be possible to fudge it in DC coupled AM mode if you invert the slope of the power meter recorder output. This assumes your sig gen has this DC coupled AM mode.

You can see in reply #9 in the link above that I managed to get a fairly flat response from 10MHz to 3GHz using a setup like this

Thank you - I think I started to get it.  And in retrospect I am ashamed I did not ask right from the beginning how to calibrate the SA using the sig gen.   

[taydin]

1. Which cable is that ? I saw Pasternak has a lot of cables - it confuses idiots such as myself. The attenuation is just a figure like 4db. Is that per meter ?

2. How does one measure the loss ?

I have bought PE-P195 cable assemblies in various lenghts and connector types. I just took three same length cable assemblies with BNC connectors and measured the output of my RF siggen on my scope. I only swapped out the cables and left the rest the same. Sigged generated 1 GHz, 0 dBm, and scope measured -0.8 dBm.

Note that the cable loss probably isn't 0.8 dB. The siggen could be off from 0 dBm, it's not a high performance siggen. The scope's tolerance is also factoring into this.

But the point is, with an identical setup, I have measured almost exactly the same loss for three separate cable assemblies (24" length, BNC male on both ends), which means all three cables have a consistent build quality and hence, consistent loss.

[taydin]

The attenuation is just a figure like 4db. Is that per meter ?

Don't remember the attentuation specs on these. But the pasternack catalog gives an attenuation figure per 100 feet.

2. How does one measure the loss ?

The proper way to characterize the loss of a cable assembly is by using a vector network analyzer. This will give you the amplitude response of the cable over its entire supported frequency range.

[LukeW]

Have the cables been terminated consistently? Is there a chance that the BNC crimps are inconsistent between different cables?

[TERRA Operative]

I use a locally sourced brand of flex cable (Fujikura) that has comparable specs to Belden stuff (Which I would use if I could more easily get it here in Japan), with Canare crimp BNC jacks that I crimp myself.
A bit of $$$ for the stripper, crimper and dies, but the actual jacks are cheap and super high quality.

[BravoV]

What kind of coax ?

For thick coax like RG8 or others that even bigger or better like solid outer jacket, try find local cellular tower contractors, sometimes they carry huge excess high quality coax cable for telco industry. My last purchase locally was Leoni RG8 equivalent with thick stranded cores as its not that stiff compared to single solid core.

[AlexResch]

Have the cables been terminated consistently? Is there a chance that the BNC crimps are inconsistent between different cables?

I ca't say. I did crimp new BNC's on one of the cable and it seems to have gotten better. But I suspect these are cheap, low quality cables. Next week, I'll avail myself of some premium cable and connectors just to have something to compare with.

[TheUnnamedNewbie]

Quality coax? Gore phaseflex, ofcourse.

Now on a more serious note: How much difference are you seeing between cables? What do you want to use your cables for? Measurement? Fixed installation? There are different ways a coaxial cable can be good - it can be rock-solid loss/phase/impedance vs bending and use, like the really expensive cables used on VNAs (Phaseflex, sucoflex 500, etc). But these tend to be quite thick, have a mind of their own, and don't really care how you want them to bend, they will do whatever they damn well please. They also are not the lowest loss cable you can get, that would be something with a solid outer jacket (which is likely also the most stable, because it can't move) The on the other side of the spectrum, you have really thin, flexible cables. They are not great in any way (compared to the xxxx.xx USD/m VNA cables) except for being very flexible and thin. Without knowing what you want you can't really get a good recommendation. I'd say good brands are Gore, H+S, Belden....

Also, keep in mind that BNC was never a great connector for precision stuff. I think general purpose BNC connectors can easily give you .2 dB IL, worse if the termination is not 100%. Factor into that slight missmatch etc, and even the shortest cable possible would give you half a dB of loss.

[AlexResch]

I use a locally sourced brand of flex cable (Fujikura) that has comparable specs to Belden stuff (Which I would use if I could more easily get it here in Japan), with Canare crimp BNC jacks that I crimp myself.
A bit of $$$ for the stripper, crimper and dies, but the actual jacks are cheap and super high quality.

I believe I have some Canare BNC jacks (50 Ohms ). I'll tripple check my stripping tool is adjusted as per their instruction here and give it a new try :

http://www.canare.com/ProductItemDisplay.aspx?productItemID=34

[AlexResch]

Quality coax? Gore phaseflex, ofcourse.

Now on a more serious note: How much difference are you seeing between cables? What do you want to use your cables for? Measurement?

Aprox 2dB at it's worst. I am inclined to believe the cable(s) was absolute rubbish and the BNC jacks not much better. Measurement, yes.