3
\$\begingroup\$

What are your thoughts on how to correct this waveform? The pulse should be a 410 ns wide and 5 V high square wave.

Scope Capture

The driver is a 74HCT365 push-pull buffer, supplied from the 5 V rail. The input (3.3 V from an ESP32) to the IC looks perfect.

There is approximately 25 feet of 3C16 cable (12 V, ground, signal) between the buffer and the input of the WS2815 LED strip.

The basic circuit is like this: ESP32->75HC2365->series resistor->connector->25' cable->LED strip. The waveform capture was taken at the connector.

I tried adding 20 μF of bulk capacitance across 5 V and GND about 2 mm from the buffer IC. I've also tried a ~400 Ω series resistor between the buffer and the WS2815. Neither significantly changes the output.

The LED strip functions properly, but this output is going to bother me if I don't clean it up.

Using the suggestion below I lowered the series resistor to ~50 &ohm.

Yellow is between the resistor and buffer. Blue is taken between the resistor and cable. Purple is between the cable and led.

Cleaned up

\$\endgroup\$
6
  • 2
    \$\begingroup\$ Assuming those reflections are measured at the source side of the wire, you can reduce them by impedance matching the buffer to your cable. Do you have a datasheet for the "3c16" cable? Does it spec an impedance? If not, you'd have to guess values and see what zeros the reflection. 400 ohms is way too high, maybe 40-60 ohms would be a better guess. It needs to go as close as possible to the driver, in series with the output. \$\endgroup\$ Sep 14, 2022 at 2:42
  • \$\begingroup\$ Which side of cable the resistor is, at the buffer or LED side? From which point the scope measurement is made, buffer or LED? That is important to know so we don't have to guess. \$\endgroup\$
    – Justme
    Sep 14, 2022 at 5:06
  • \$\begingroup\$ @user1850479 I do not have a spec, cheap Amazon LED cable. Which artifacts are you referring to as "the reflection"? I'll give a smaller resistor a try. Thank you for your patience, it's been a long time since I studied this stuff in college. \$\endgroup\$ Sep 15, 2022 at 0:14
  • 2
    \$\begingroup\$ The second rising edge about 180ns after the first is probably the reflection returning from the high impedance on the far end of the line. The first falling edge is probably the gate driver going low, but it doesn't go to zero because the reflection is still incoming from the far end. Terminating the source will absorb that reflection and keep it from bouncing around. However, to really see how well this is working you should measure at the far end of the wire, since even perfectly terminated you will still see the reflection come in once (just before it is absorbed). \$\endgroup\$ Sep 15, 2022 at 0:42
  • \$\begingroup\$ Is it a problem? The LED looks to be getting a clean enough signal. \$\endgroup\$
    – user253751
    Nov 17, 2022 at 13:01

1 Answer 1

1
\$\begingroup\$

The LED strip functions properly, but this output is going to bother me if I don't clean it up.

What you are seeing is the effect of reflections on your cable. At one of the line you have a <100 ohm output impedance of your driver circuit. On the far end of the line you have a very high impedance MOSFET gate. When you launch a square wave into that wire it travels down the cable taking roughly 90ns (judging from the picture), reflects off the LED input, and then bounces back another 90ns later, and then bounces again off the source:

enter image description here

The reflection from the load (2nd dotted line) is intended and not harmful, so no problem there. The second reflection is not intended (source should absorb the reflection), so adding the resistor helps. If you'd gotten the resistor value exactly right for the impedance of your line, in theory there would be no second reflection.

Looking at the load waveform, your signal looks quite good. I would leave it as is. If you were really concerned or were doing this all over from the beginning, I would use twisted pair cable with a known impedance. Then it would be easier to choose the resistor, and your signal would probably look a little cleaner (most likely the cable you're using doesn't have a well controlled impedance). You're also sharing the signal ground with the power ground, which isn't ideal, but probably doesn't matter too much.

I tried adding 20 μF of bulk capacitance across 5 V and GND about 2 mm from the buffer IC.

Bypassing capacitance would help, not bulk capacitance. You have edges on the order of 1 ns, an electrolytic capacitor (which I'm guessing you used from the odd 20 uF value, although maybe you actually added two 10uF ceramic?) won't do much at those frequencies. A ceramic will help more, but is evidently not needed (or already present).

\$\endgroup\$
1
  • \$\begingroup\$ You have been extremely helpful, thank you. You are correct in your "bulk" vs "bypass" capacitance. I have 20 uF of ceramic caps on the 5 V side of the buffer, two 10 uF in parallel. With that series resistor in there I'm 100% happy with what I'm seeing on the load side. Given the design is required to have a cable of unknown length there's not a whole lot more I can do. Again, appreciate your help! \$\endgroup\$ Sep 15, 2022 at 23:46

This site is temporarily in read-only mode and not accepting new answers.

Not the answer you're looking for? Browse other questions tagged .