Thru hole vs surface mount R's at ~100MHz.

0

IMHO stray inductance in bypass capacitors will bite you before stray inductance in series resistance. And there's no avoiding it, through hole parts will have way higher stray inductances than SMD.

As far as leaded resistors go, my feeling (based on RF behavior above the 10W level) is that ceramic composition resistors are least inductive, and seem to completely sidestep the humidity/tolerance crappiness of carbon composition, but I don't think this conclusion is useful for any parts in a PMT amp.

Low-inductance SMD parts are not incompatible with dead bug. You can carve up copper islands with a knife (or Dremel it out if you'd rather) and get the best of both worlds.

Tim.

100 MHz is practically DC. Carbon films should be fine.

I prefer live bug. It's less confusing, and grounding tends to be better.

John

George Herold Inscribed thus:

Beware ! Higher values of surface mount resistors can be quite inductive. Something to do with the way the values are trimmed.

--
Best Regards:
                Baron.

100

Thanks All, I used metal film, and I've got the cutist little pulses at ~20mV out of one X10 gain stage. I'll try adding another stage. This project is almost done, Next!

George H.

100

I've done that, and it works well. L-trimmed flat surface mount resistors have much lower parallel capacitance than spiral-trimmed leaded resistors, and should have lower series inductance (though I've not had practical experience of that).

The cylindrical MELF surface mount resistors are an unknown quantity here - I'm nervous that they might be spiral-trimmed.

-- Bill Sloman, Nijmegen

Interesting conjecture. Got numbers?

and should have lower series inductance (though I've

At 100 MHz, spiral trimmed carbon or metal films are fine. Even 50 nH, a silly overestimate, and 50 ohms is a 1 ns L/R time constant, which is 160 MHz corner frequency.

John

It depends on resistance value, of course. Low value resistors will have more inductive reactance; higher value resistors will have more capacitive reactance. 75 to 200 ohms is (IIRC) the sweet zone.

Note that 100MHz 'scopes (and a bit faster) used to be made with mostly

1/4W sized through-hole resistors. One of the limiting factors was the inductance in the emitter-to-emitter impedance in differential amplifiers, especially in the final output stage where the resistance value is quite low. Above 200MHz this starts to get difficult.

Of course Sloman doesn't have numbers.

I do.

1M 0603 thickfilm 5% resistor 0.32 pF

1M 0805 thickfilm 5% resistor 0.29 pF

1M 0805 thickfilm 1% resistor 0.36 pF

1M 1206 thickfilm 5% resistor 0.29 pF

1M 1/4w axial cf resistor 0.33 pF

1 pF 0603 ceramic cap 1.17 pF

The cap was just a sanity check.

I see no pattern here. Probably the surf mount end cap dimensions dominate, and I'm groveling at my measurement resolution anyhow.

I use cheap surface mount resistors into the several GHz range.

Maybe I should do inductance too.

John

Nothing too specific. Cambridge Instruments merged with Metals Research around 1979, and I got stuck with fixing the Qantimet Image Analyser video camera around 1983 when it was being built into the ChipCheck mask inspection system. The amplifier that detected the video signal fronm the scanning analyser had recently been "improved" by replacing the camera tube with a new tube that we bought from Philips. The new tube came with a built-in Philips pre-amplifier, which used a L-trimmed surface mount resistor as the feed-back around the amplifier. It was a much higher value resistor than the equivalent in the pre-amplifier we'd used on the old tube - IIRR about 680k as opposed to 120k. The old resistor was a standard Philips metal film part with about 0.3pF of parallel capacitance. If the new resistor had had the same parallel capacitance, it woud have rolled off the high frequency response and killed our spatial resolution. Since it didn't, the parallel capacitance had to be below 0.1pF.

All I ended up having to do to fix the problem was to recognise that the new resistor was a higher value part, and rejig the downstream circuit - changing two resistors - to increase the current drawn from the camera tube back to the optimal value.

The Philips data book used to have rf bridge data on their axial lead metal film capacitors - 470R was a magic value, as the stray capacitance and stray inductance seemed to resonate at a frequency where they both had an impedance close to 470R - about 1GHz - which puts the inductance at a not-unreasonable 6.6nH.

-- Bill Sloman, Nijmegen

0 s

In fact I did, but not all that precise - even if I didn't post them until some eight hours after John Larkin had asked for them.

I suspect that a lot of the capacitance measured above is from the leads and the mounting pads. My Philips 680k (or thereabouts) surface mount part was on a very small "board" which might well have been an alumina-based thick film hybrid, and could have been explicitly engineered for low parallel capacitance - a driven shield electrode wouldn't have been obvious during the kind of superficial examination I was in a position to do.

With 10nH per inch of lead length, low inductances aren't all that easy to measure either.

-- Bill Sloman, Nijmegen

I didn't use pads. I have a fixture that's a pc board with copper on top and a slit down the middle. I zero out the c-meter then drop the part across the slit, push it down with a toothpick.

Any idea why all those different sized resistors would have practically the same capacitance?

My Philips 680k (or thereabouts) surface

TDR.

John

100 .

se

ray

gh

ove

n is

an

,

You zero it out with a C-meter without a resistor - or the lump of alumina on which the resistor is mounted - bridging the gap.

What is the capacitance if you put in a resistor from which you have scraped the resistive film? Or an unmetallised chip of alumina - after you have ground off all the metallisation?

The field lines from the copper on either side of the resistor will get a shorter path through the alumina body.

And what would be the capacitance if you used a test board with a single narrow track with a slit in it? It would look more like a real resistor on a real board.

-- Bill Sloman,

Why do you care if a 1M resistor has an extra nanohenry of inductance? You'd have to be working at optical frequencies to even see it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

Right. A 50 ohm 0805 resistor might have a nH or two, so its L/R time constant might be 20 picoseconds or so. If its shunt C is, say, 0.5 pF including pads, the RC tau is 25 ps, same ballpark. That puts the parallel resonance point in the 10 GHz ballpark.

John

100

The ~100 Ohm 0805 resistors behave as resistors in the widest band, as=20 stray L and C are compensating for each other. Higher resistor values=20 are dominated by C, lower values - by L.

VLV

On a sunny day (Mon, 15 Nov 2010 17:04:12 -0500) it happened Phil Hobbs wrote in :

Right, at 100MHz almost anything will work if the resistors are high value. I have build stuff at that frequency with normal high value resistors no problem. It is perhaps only in the 50 Ohms range and maybe lower that a coil could be formed from the turns of the carbon or metal, with still a Q to have some effect. Capacitance is usually fractons of pF, so no real problem either.

On a sunny day (Mon, 15 Nov 2010 14:20:21 -0800) it happened John Larkin wrote in :

With a very bad Q.

But...but... it's a resistor!

John

Here's a cheap 1/4 watt 51 ohm axial carbon film resistor.

Soldered to copperclad and hardline:

ftp://jjlarkin.lmi.net/51R_setup.JPG

and resulting TDR:

ftp://jjlarkin.lmi.net/51R_TDR.JPG

One can sort of imagine an L/R decay in the 100 ps range, which would be 5 nH. The ring is at 9.6 GHz.

The initial blip is clearly net inductive, at least with a 28 ps TDR system. This would be a perfectly good resistor to a couple of GHz.

I could have make the leads a little tighter, but this is more or less realistic for this kind of resistor.

I'll do an 0805 if I have the energy.

John