Look at this 50 ohm terminator. This isn't right is it?

[T3sl4co1l]

Richard: he said 5.1, not 51!

Also, SMA is usually used at rather high frequencies - no way in hell that cheapo axial is still a resistor above 100 MHz or so...

Nah, it's good up to several GHz at least.  Well, if it were a proper 50 ohm that is.

The coaxial construction means it's a lossy transmission line.

Tim

[c4757p]

Coaxial? What about the spiral cut?

[T3sl4co1l]

The spiral cut will act as a helical transmission line inside the wall.

It will have some dispersion due to being helical, worsening SWR at very high frequencies (wavelength on the order of the spiral length).  The spiral isn't very long, so it's in the GHz.

On second thought, the impedance is probably too low, causing a small hump in SWR probably in the low GHz, or maybe a bit below.  Mainly because the resistor end caps are right up in there.

Through hole resistors are frequently quoted as being terrible, which is absurd.  Many EE "rules of thumb" are.  Only a terrible engineer would take such statements at face value.

Tim

[c4757p]

Hmm. You're probably right about that. One irritating thing with axials is that the greater length often makes it inconvenient to minimize parasitics with them - was just building a pulse generator today for a scope calibration, and an axial resistor attenuator had just too much inductance for a clean pulse but an SMD one was fine. That, I suppose, was more due to the extra loop area caused by the shape of the part when applied across the transmission line than by the part itself.

[T3sl4co1l]

Yup.  It's all about how you apply it.  An axial 1/4W is probably a good match to 100 ohm transmission line over a ground plane, which is fine for thinner traces on a PCB, but not necessarily as practical for 50 ohm or other stuff.  A resistor floating in space (on some elevation, or suspended between terminal strips perhaps) can be even higher, which would be handy for, say, old Tek stuff which typically used ~1kohm system impedances (tube tech, at high voltages, for driving deflection plates).  High impedance, of course, is associated with inductivity against lower impedances, hence the 'rule of thumb'.

If you had really low impedance (e.g., some resistors in an attenuator, anything with a generally low system impedance -- very fast switching converters come to mind), larger, and sideways, resistors (e.g., 1218 vs. 1812) might be better.

Tim

[VK5RC]

I am surprised they didn't use a pair of 100 ohm smt resistors,  probably would have been cheaper!

[SeanB]

I am surprised they didn't use a pair of 100 ohm smt resistors,  probably would have been cheaper!

Harder to assemble though, it is hard to solder inside a hole, while the resistor is a push fit with no soldering needed.

[IconicPCB]

Back in the seventies HP used to make switchable attenuators using carbon composition resistors.

Resistors had adjustable capacitors around the body ( a technician would manualy form a sliver of foil around the body of the resistor  ) to peak the inductive components of the resistor construction.

No reason why a 50.1 ohm resistor inside the SMA shell would be any worse off than the resistor inside the attenuator assembly

[edavid]

Here's some data suggesting that an axial resistor is not good enough for an SMA terminator:  http://defenseelectronicsmag.com/site-files/defenseelectronicsmag.com/files/archive/rfdesign.com/mag/511RFDF3.pdf

[macboy]

Here's some data suggesting that an axial resistor is not good enough for an SMA terminator:  http://defenseelectronicsmag.com/site-files/defenseelectronicsmag.com/files/archive/rfdesign.com/mag/511RFDF3.pdf

The problem with that reference is that it talks about an axial leaded resistor on PCB, not used in a coaxial construction.

[T3sl4co1l]

Curious if the plot there is calculated from the assumed models, or actually measured (with smoothing or curve fit).

As mentioned, these things are all transmission lines on some level, so a lumped equivalent will only be reasonable in the asymptotic limit (frequencies below resonance, cutoff, 1/4 wave equivalent, or however you might describe it), and approximate near the first peak or valley.  A real transmission line structure exhibits harmonics or modes, which a first order lumped equivalent cannot represent.

So, you could spam more lumped components in there to make better and better equivalents, but... why?

I find it amusing that they propose a chip resistor's inductance is entirely due to the end contacts, and that this is three times greater than the inductance of a ball.

You can also place chip resistors upside-down, which is supposed to be of mild advantage over 10 or 20 GHz.

Here's a good modeling example which includes diffusion (skin effect) and one resonant mode (which should be a helical resonator mode of the winding).  The number of lumped components is much greater than the single physical component, as you can see.
http://seventransistorlabs.com/Images/CurveFit2.png
http://seventransistorlabs.com/Images/CurveFit1.png

Tim

[IconicPCB]

The "helix" is made up of resistive , lossy material and any resonance would have particularly low Q. I would not expect any significant salient features in the return loss measurement on a termination made from an axial resistor.
In fact I used to make termination resistors inside F connectors. My favourite tuning tool to improve return loss figures was a pair of combination pliers which I used to squash the connector ferrule.

On the question of lossy transmission line, used to use window putty loaded with iron powder to dampen potential instabilities on input and output ports of high gain broadband multichannel MATV amplifiers.

Do not get too stuck on the resistor construction, if it fits the bandwidth of interest.. You use it,... or try to nudge it into shape