3
\$\begingroup\$

I am assembling nonlinear oscillators on PCB boards for applied math research [Colpitts (100 kHz, 3 V_pp), FitzHugh-Nagumo (100 kHz, 15 V_pp)]. I am working with some other people that want to do ODE simulations, so I am pushing to provide some qualitative results quickly. This means I sometimes solder up components before I rigorously measure them, because an approximate measure (~100 ohm vs 99.24 ohm) is enough for the data I need to hand off.

Can I measure these components (resistance, inductance, capacitance) once they are soldered in, or do the connections negate this possibility? Sorry if this may be a basic question, but my background is not in electronics. I understand that I can remove them, measure them, and then re-solder them, and this is probably the obvious approach--I would prefer to avoid doing this because small qualitative changes in the behaviors are important to me, and re-soldering would probably cause that. Should I just take the time/repeatability hit and always measure isolated components?

Also, do any of you have an intuition on what magnitude of stray capacitance occurs on a PCB board? I know that software layouts like PSpice avoid these problems, but I don't know how to use that functionality yet--if I'm being pessimistic, would I see 10 pF? 100 pF?

Edit: I include a picture of one of the soldered circuits so you can get an idea of spacing and the complexity of the circuit (not pictured--the leads for the V+/- power sources for the op-amp).FitzHugh-Nagumo relaxation oscillator

Edit 2: I include a schematic of the pictured circuit, so I can ask where I might consider that suggested capacitor.

FHN diagram

\$\endgroup\$
  • \$\begingroup\$ You should have a supply cap near the IC to reduce any possible current spike induced noise. voltage. \$\endgroup\$ – Sunnyskyguy EE75 Dec 14 '18 at 21:12
  • \$\begingroup\$ I have to be careful about adding components that add differential equations--the theory people are cranky about that. But if I did add a capacitor by the op amp, would it be on the output? The voltage supply? Again, my familiarity with terminology is incomplete. \$\endgroup\$ – KBL Dec 14 '18 at 21:29
  • \$\begingroup\$ This Op Amp has a slew rate of 2V/ns or 80MHz so adding a cap across pin 7 & 4 reduces supply noise somewhat if there is any or load noise from spikes or does nothing. this is common practice and depends on circuit for choice of caps. With no cap there is a risk of adding spurious noise > 10MHz \$\endgroup\$ – Sunnyskyguy EE75 Dec 14 '18 at 21:37
  • \$\begingroup\$ With the wrong values for R1,R2 and their ratios , it may not work. from excess hysteresis ( positive feedback ratio) and output current cannot drive too low a R load. But it sure looks a lot bigger than a neuron \$\endgroup\$ – Sunnyskyguy EE75 Dec 14 '18 at 22:53
  • \$\begingroup\$ If you stand up the resistors your layout could be 10% of the area which is usually best for LC current and feedback loops. \$\endgroup\$ – Sunnyskyguy EE75 Dec 14 '18 at 23:07
1
\$\begingroup\$

Would have to see your circuit to figure out whether you can measure in-circuit at all without the connections disturbing the measurement. However, if you application requires, as-per your example, 4-digit accuracy on resistance, there's no way you can assume your connections aren't going to disturb the measurement. I'm assuming you're doing this power-off.

Trace capacitance is order 10pF. There are calculators online that will work it out for you. The details of course are dependent on trace length, width and your PWB stackup. For example https://technick.net/tools/impedance-calculator/

\$\endgroup\$
1
\$\begingroup\$

If have ony an ordinary meter then you cannot make accurate in-circuit measurements because the other connections will skew the result, but such measurements are possible with specialised equipment.

Most technicians will un-solder one end of a leaded component it the need to measure it, as this is usually the hardest way to mess up the measurement.

With specialised equipment Any 2 terminal component that is not parallel with another can be measured in-circuit by nulling out the other nodes and measuring it's impedance.

schematic

simulate this circuit – Schematic created using CircuitLab

Here unknown Z1 can is compared against known R1 If both are resistors the comparison will be the ratio of the voltages

Za and Zb do not affect the reading because V1 is a voltage source, and current in those parts does not effect the voltage a proble A. Zc and Zd do not affect the reading probe B is a virtual ground and thus Zc and Zd have no potential difference and so pass no current. the grounds on Za Zb Zc Zd may be applied with aditional probes (usually clip probes).

\$\endgroup\$
  • \$\begingroup\$ I have a standard multimeter, and my expertise is shallow, so it's probably better that I don't get too fancy. But--I need only unsolder one end of a component? This will help, because then I won't have to destroy the circuit so much. \$\endgroup\$ – KBL Dec 14 '18 at 20:49
1
\$\begingroup\$

Unless you have a ground plane or twisted pairs or close proximity, stray capacitance can be < 1pF. It all depends on length/gap ratios of or between conductors.

  • Twisted pair is 0.5~1 pF/cm , single traces much less
  • single trace Inductance can be 10 nH/cm,
  • paired mutual inductance again depends on length/gap ratio.

For 100kHz range this is almost negligible.

For relaxation oscillators, R & C tolerance contributes to errors. For LC oscillators , L & C tolerance contributes to errors.

Just make it neat with correct parts in the correct locations. A final check on power supply resistance helps prevent the obvious with a DMM.

Compare your solder joints with the best and learn to do it fast in < 2 seconds with a clean tip and clean copper board.

I might solder cut resistor wire for probe test points for signal and ground on outputs.

\$\endgroup\$
  • \$\begingroup\$ Could you elaborate on what you mean by relaxation versus LC oscillators? In mathematics, they have a certain meaning, so I would be really curious to know how these terms are used in a more strictly electronics sense. In mathematics, relaxation tends to refer to high harmonic oscillators, like ekg traces or neuron firings. It's opposite, a "smooth" oscillator is sinusoidal. Is there a guideline for intuiting this behavior just from looking at a circuit, broadly? If this is too complicated to quickly explain, could you suggest links to reading? Thanks very much. \$\endgroup\$ – KBL Dec 14 '18 at 20:46
  • \$\begingroup\$ But the FitzHugh-Nagumo uses an RLC model with a negative resistance activation for hysteresis that also acts as a squelch with insufficient or too much energy. . \$\endgroup\$ – Sunnyskyguy EE75 Dec 14 '18 at 20:59
  • \$\begingroup\$ Relaxation types use negative feedback with hysteresis UNLIKE FitzHugh-Nagumo and synapses with an activation potential. LC types resonate with a sine input and amplify to a square output with 180 deg phase shift and 180 deg inversion hence positive feedback and narrow bandwidth so more stable yet as we know body chemistry uses relaxation type so easily more variable frequency to adapt like a voltage controlled frequency. But the FHN model is like firiing neurons according to a limited potential energy range applied yet no fixed threshold. \$\endgroup\$ – Sunnyskyguy EE75 Dec 14 '18 at 21:03
  • \$\begingroup\$ tinyurl.com/y8ph2xv8 FHN model \$\endgroup\$ – Sunnyskyguy EE75 Dec 15 '18 at 0:08
  • \$\begingroup\$ Thanks, this is a very neat demo! Any insight into why three transistors, and why the general architecture? It's very different from mine. The oscillator I am using is adapted from this paper--it does a very nice job of describing the purpose of all the components: ieeexplore.ieee.org/abstract/document/6313098 \$\endgroup\$ – KBL Dec 17 '18 at 17:52

Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.