Measuring Digital Cables - Canare L7CFB

[House de Kris]

Another digital audio cable to test as supplied from DKak. This time a cable made of Canare L-7CFB (RG-11) terminated in BNC/BNC. Not sure who made it, but I believe it is Markertech. I have no idea about the price. A directional cable was affixed, and I respected it for these measurements. This is a pretty fat cable without added fancy sleeves, hoses, or jackets.

Attached time domain and frequency domain pictures. The impedance is a smidge high at 77.38ohms, but extremely uniform throughout the length. The 3dB down point is 5.4GHz. Another screamer.

[DKak]

Yep, this L-7CFB cable was made by Markertek.

I wonder why the bandwidth is less than the L-5CFB? It does have larger conductors for lower loss over greater lengths, but I don't know how this relates.

I find it interesting, too, that Kris' measurements show this (all BNC-BNC):

L-5CFB is about 8.7GHz bandwidth, 337ps risetime, 75 ohms
LV-61S is about 7.5GHz bandwidth, 342ps risetime, 74.3 ohms
L-7CFB is about 5.4GHz bandwidth, 337ps risetime, 77.3 ohms

Canare's published cable specs:

LV-61S (RG-59)
24 awg stranded bare copper
braided copper shield
PE dielectric
PVC jacket
Vprop 66%

L-5CFB (RG-6)
18 awg solid annealed bare copper
braided tinned copper shield plus AL foil
Foam PE dielectric
PVC jacket
Vprop 79%

L-7CFB (RG-11)
15 awg solid annealed bare copper
braided tinned copper shield plus AL foil
Foam PE dielectric
PVC jacket
Vprop 79%


LV-61S has the smallest conductors, followed by L5-CFB and the L7-CFB, but that's not the order of widest bandwidth. Is bandwidth then, in part, a function of the dielectric? What does Vprop mean in any of this? What does the risetime mean? Stranded vs. solid? And the dreaded skin effect...anything here relating to that?

Can't wait to read Kris' take on all of this!

[House de Kris]

I think an important aspect to consider in this three-way comparison is that the length of the LV-61s is about half that of the other cables. Let's make another table, based on what you have so far:

L-5CFB 3.0ft 8.7GHz 337ps 75.0 ohms 3.61ns
LV-61S 1.5ft 7.5GHz 342ps 74.3 ohms 2.01ns (x2=4.02)
L-7CFB 3.0ft 5.4GHz 337ps 77.3 ohms 3.46ns

When comparing the L-5CFB to the L-7CFB, everything from the manufacturer looks about the same, except the gauge of the center conductor. I, too, would have thought that a larger conductor would have given a wider bandwidth, due to the greater amount of conductive material per same amount of transmission depth. Let's revisit how we've defined bandwidth. For all the comparisons I've added text to, I've always used the -3dB point since that is a very common bandwidth definition point for most any network. But, the -3dB point, also called the half-power point, is indeed a pretty severe attenuation of signal when issues of signal integrity are at hand. I get the impression many interface engineers define bandwidth more alone the lines of the -1dB point. This represents about a 10% reduction in voltage (around 20% power reduction). If we look at the -1dB points on these two cables, we'll find the L-7CFB is 1.53GHz and the L-5CFB is 1.43GHz.

As far as I know, the dielectric has the greatest effect on propagation time, or propagation velocity as Canare referes to it. In this case, both cables with foam PE (L-5CFB and L-7CFB) are claimed to be 79% the speed of light where as the solid PE dielectric (LV-61s) is claimed to be 66% the speed of light. Likewise, the measurements of propagation time for both foam dielectrics is pretty much the same (~3.5ns) with the solid being slower (caluculated to be an equivalent of 4ns). Without knowing the physical length (or better yet, the electrical length) very accurately, we can't relate this to the speed of light very easily, but the numbers appear to make sense.

The risetime measurement on the TDT pictures is the risetime of the edge coming out the end of the cable. The faster the better here for resistance to jitter induced by the interface. Risetime is quite related to bandwidth, but not always directly related. If we could assume a textbook perfect 1st-order system, then the simple BW=0.35/Tr equation works. This is also pretty good for a second order system. Beyond that you have to fudge around with the numerator to get anywhere close to reality. You'd think a cable should easily be a first-order system, but with reflections from mismatches and whatnot, it has the appearance of being more complex.

As far as stranded vs. solid, traditional thinking has been that for very high frequencies, solid is best. More than likely though, for digital audio applications either is acceptable. The "no compromise" types amongst us may just opt for solid then. But, stranded is more flexible and may be more desirable in some applications. The issue with stranded is that it is very hard to get a smooth surface with multiple conductor strands all squished together. Even for solid conductors, the surface polish is very important. If the signal has to travel up and down surface roughness, it will actually have an electrical length that is longer than the physical length at very high frequencies. Hmmm, now that I type this, I'm starting to wonder...If a cable has a unpolished center conductor which makes its high frequencies travel a longer length than the low frequencies traveling over the same cable, then from the destination's point of view, the high frequencies would have a phase shift WRT the lower frequencies. A phase shift usually indicates a pole higher up in the frequency band, and a pole helps define the bandwidth. As such, it is conceivable to me that the surface polish on the center conductor can influence overall cable bandwidth. It would be interesting to find out if Canare specs the surface smoothness of these cables. If they do, it would be in the order of a few 10s of microinches. Everything in this paragraph deals with skin effect.