1
\$\begingroup\$

The circuit purpose is to control square signal frequency from 100Hz to 1kHz (generated by MCU GPIO) and be able to control amplitude from 100mV to 2.3V. It is like a very simple signal generator. The goal is to keep edges <100ns. Switch is SN74LVC2G53. The schematic:

enter image description here

Problem: I get ringing on rising edge (ringing Frequency 3MHz), and on falling edge going negative about -600mV. enter image description here enter image description here Layout(i know it's bad, but i had no choice at this time. I will redesign the PCB later ): enter image description here

How can i reduce ringing on falling edge?

\$\endgroup\$
3
\$\begingroup\$

You can try wiring it dead bug over a ground plane:

enter image description here

Source of the picture is here: Ugly is Beautiful!

The chip's VCC pin should be decoupled with a very close surface mount ceramic cap, which will give you a nice low inductance supply.

When the switch switches it will inject some charge into the input. Since the DAC is far away, and probably buffered by a very slow opamp, or unbuffered, its output impedance will be high. So, a 1-10nF C0G or film cap should be placed at the input of the switch to soak up the injected charge.

If the DAC is buffered by an opamp, a low value resistor like 22R will prevent it from having an allergic reaction to a capacitor at the output.

Now, the output impedance of the switch is its internal resistance (6.5 ohms) plus 0R if it's set to output GND, or the 10nF cap if it's set to output the DAC voltage. In both cases this is too low to drive a 50R coax or even a trace without some ringing, so source termination is required: you can add a 33R-47R resistor at the output.

Dead bug can yield extreme performance if you do it right. For example this is a 3.3V 100mA DIY voltage regulator. Ain't it cute? I used Kaptan tape (chinese ripoff of the well known Kapton brand) and adhesive copper tape.

enter image description here

This is the transient response to a 50mA current step with a few ns risetime, generated by a 74AC14 with all gates paralleled. Yellow is output voltage. Blue is current in the pass transistor. The time scale is indeed 20ns/div.

enter image description here

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ That's only a 3mV overshoot for ~10ns? Nice! \$\endgroup\$ – marcelm Sep 27 '17 at 20:28
  • \$\begingroup\$ Yes, the overshoot is di/dt times the inductance of the decoupling caps and plane which is about 0.2 nH total. I put a piece of laser transparency film and double sided adhesive tape between the copperclad and the power plane adhesive copper tape to emulate the 0.2mm layer spacing between toplayer and layer2 in a 4 layer board. \$\endgroup\$ – peufeu Sep 27 '17 at 21:19
  • \$\begingroup\$ My point is, by using a bit of copper clad pcb as a base and building layers with kapton tape and copper tape you can get very low inductance, in fact lower than on a PCB, but it requires a bit of creativity, a sharp xacto knife, and not drinking too much coffee. \$\endgroup\$ – peufeu Sep 27 '17 at 21:24
  • \$\begingroup\$ Thanks for the comment! Which part of the circuit must be designed using your technique (low spacing between signal and ground layers) ? In other words, which traces need to have low inductance? It it enough if i put analog switch, decoupling caps on copper clad PCB? All other connections to MCU and VCC would be made by wires. \$\endgroup\$ – Andrius B. Sep 28 '17 at 5:23
  • \$\begingroup\$ nothing super fancy really -- just copperclad as ground plane, solder the decoupling cap on the copperclad and onto the chip's pin... same for the dac's output capacitor, should be alright... \$\endgroup\$ – peufeu Sep 28 '17 at 7:07
2
\$\begingroup\$

With aerial wiring as shown at the above picture, the happening of parasitic components is inevitable. One of the most likely ways to reduce this oscillation is by twisting the wires which carries that signal, reducing as much as possible the "area" between them, since the inductance is proportional to this.

|improve this answer|||||
\$\endgroup\$
2
\$\begingroup\$

The first thing to do is ensure your probing technique is good. You need a very short ground on your scope probe or you will see ringing on your scope even if the signal on the PCB is good.

As other answers state the wiring of your circuit is not ideal. As far as possible reduce the length of the leads especially to ground.

One test I do to check if the probing is good is to firmly grip the body of scope probe with your hand - if the scope display changes then your scope ground is not good enough.

This uses your body as damping to reduce any ringing present on the scope ground - there shouldn't be any.

There are a number of guides to effective signal probing on the web here is one:

Short grounds give better scope traces

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ The probing is good, i use short ground spring. Touching the ground probe doesn't change the ringing. Thanks for the link! \$\endgroup\$ – Andrius B. Sep 28 '17 at 5:15

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.