Impact of coax cable input capacitance in non inverting op-amp amplifiers

Morning Alberto, 

Many of the high speed input stages are essentially bipolar emitter followers - a capacitive source gets converted to a negative impedance at the emitter often leading to a >1GHz local oscillation in the input stage - cannot fix with overall loop gain tricks as it is local. We saw this a lot in the earlier Comlinear current feedback days where (for inverting operation) I would always recommend a 20ohm to ground on the V+ input to protect against this. At various times we tried to fix this with internal R's but the bond wire inductance seemed to frustrate that effort. 

Normally, if you come in with a coax you are terminating in matched R - even then, I prefer a 10 to 20ohm series input to the V+ from that termination. 

One of the most memorable here was when I was doing the CLC520 and CLC522 datasheets. The 1st one, I neglected that resistor on the output CFA amplifier and we later found a >1.2GHz oscillation due to that oversight - fixed that in the CLC522 datasheet, The output stage here is a CFA with the signal a current into the inverting node getting gained by the external feedback - that pin 9 is the V+ node where here it is biased to ground through that 25ohm R that fixed the oscillation. 

Alberto,

Micheal's answer makes sense. Note that the op amp feedback loop is stable so the claim is technically correct. The input stage oscillation is at a frequency where the input cable is a transmission line and not a lumped capacitance. My purview is general purpose amplifiers, so I'll transfer the post to high speed ownership.  They will have more insight on this effect.

Dear Michael and Ronald,

thank you so much for your replies. Michael's answer makes perfect sense to me. The issues described fit nicely my observations.

The problem is: How can I know if a specific bipolar op-amp will suffer from instability induced by a piece of coax cable ? May I assume

FET input op-amps are immune ? Your opinion is very welcome.

Kind Regards,

Alberto

Hello Alberto,

Can you clarify as to what amplifier you are seeing these oscillations in and provide more details on the coax cable you are using?

Best,

Hasan Babiker

Hi Alberto,

what situation are we talking about? Is the termination resistor at the input of OPAmp mounted or not?

Kai

Alberto,

A single diff NPN input stage will provide a base current that varies with input voltage. The base current is power and has a voltage dependence.

Chart is input currents versus input voltage (other input grounded)

Bipolar input current, that shares current with the other input, could be a power source for an oscillation. Higher the input current an op amp has the more power available to do something odd. High speed amps can have high input current. A FET input op amp has no input current other than ESD cell thermal leakage. No current and no voltage makes no power which couldn't create an oscillator circuit. 

Dear Kai,

in my application, the 50-ohm RG58 coax cable is NOT terminated on its impedance. This is because I am interested in nearly-DC signals, from 0.1Hz to 1 Khz.

The amplifier driving my post-amp can only drive Rin >10kohm. So, in my case, R1 = 10kohm. I have experienced high-freq oscillations (>150MHz) with

many hi-speed op-amps, from TI and other vendors (i.e. LT1806 from Analog Devices).

Such oscillations vanish if a series resistor is added, In my experience, 100 ohm does the job.

Kind Regards,

Alberto

Dear Ronald,

thank you for your explanation. In your model, the leakage current 'drives' the cable and it can sustain oscillations, possibly due to SWR of

the unterminated cable (on the far end). However, I have no clue why a 100ohm series resistor quenches the oscillations. Have you any

model to simulate in order to undestand the details?

Thank you so much for your help.

Best Regards,

Alberto

So I thought we were done here Alberto. 

This range of amplifiers that oscillate, Keep in mind even relatively slow op amps use fast devices - for instance, our 200MHz type devices use 10GHz transistors. 

So whichever bipolar input op amp you are using even for your low speed signal, that input transistor will be very unhappy with a capacitive source - adding a series R changes that source impedance to subdue the input transistor local oscillation

Somewhere, if I still have that circa 1979 notebook, I have a derivation of this issue by the IBM Fellow I was working with at the time. Its intense, just put the resistor in

As for JFET and CMOS inputs, I am less sure they have this issue, but would it hurt to add the resistor.