1
\$\begingroup\$

In my first realization of a common source JFET preamplifier, I encountered a rather predictable problem but for which I was unable to find any documentation: the power supply ripple and noise are reflected in the output signal, which comes out amplified and audible in the audio output!

The ripple has a frequency of about 120Hz and a Vpp of 7mV, this is the picture of the power supply output: enter image description here

This is the ripple at the output of a 4.7μH π LC filter and ~4mF placed after the 600kHz step-up, which falls within the audibility threshold. This ripple also contains noise at 265kHz and 2MHz, which, being well beyond the audibility threshold, should not be a problem.

So I wonder: what are the best ways to obtain a clean output of the signal to be amplified? Would an RC filter of 200Ω and 47μF only attenuate the issue, but not be a real solution? I would like to use a solution that requires discrete components and does not overly complicate the circuit. Can the noise at 265kHz and 2MHz have a negative impact for an audio application?

Additional information required:

\$\endgroup\$
5
  • 2
    \$\begingroup\$ Please show the full exact circuit and explain what the power source actually is and, describe how you have built this. Provide also a data sheet link to the inductor. What signal in particular is the scope picture showing? \$\endgroup\$
    – Andy aka
    Commented Jan 13 at 17:22
  • \$\begingroup\$ I have added additional details you requested and some extra information \$\endgroup\$
    – boromyr
    Commented Jan 13 at 18:12
  • \$\begingroup\$ 3.7V input and 14V output,600kHz,4.7μH Is the 3.7V source a battery? \$\endgroup\$ Commented Jan 13 at 18:21
  • \$\begingroup\$ You have a regulated boost converter. This has very good rejection at low frequency. Moreover the booster is battery powered isn't it? Mains frequency can't come from there. So I would expect some probing issue and also mains interference pickup in the guitar signal. \$\endgroup\$
    – tobalt
    Commented Jan 14 at 5:01
  • \$\begingroup\$ Yes, the power supply is battery operated. To study the ripple and noise present, I should probably only solder the components of the power management, and therefore nothing beyond the output capacitor of the step-up. At that point, I could proceed to more accurate measurements, following the advice of @bobflux. \$\endgroup\$
    – boromyr
    Commented Jan 14 at 19:06

3 Answers 3

2
\$\begingroup\$

Those high-frequency components could be filtered with an LC filter, but the low-frequency 120 Hz component requires a different approach. One must be careful that the 120 Hz component doesn't ride in with the input signal - a ground loop might cause this.

A capacitor multiplier might address the 120 Hz component, if it really does enter via Vdd. It does consume some Vdd, dropping a few volts, but has good attenuation of 120 Hz:

schematic

simulate this circuit – Schematic created using CircuitLab

One has to be careful that Q1 doesn't oscillate at hundreds of MHz. Even printed-circuit construction can oscillate at low amplitude. Ferrite beads on base and/or emitter might be advised.

\$\endgroup\$
2
  • \$\begingroup\$ Great solution, I have done some simulations in Multisim and achieved a reduction in ripple of up to a few tens of μV, also it is simple and inexpensive. I wanted to ask if there is any transistor more suitable than the 2N3904 for this purpose, or if it is already sufficient. Do you recommend placing the ferrite beads before or after R2 and R4? \$\endgroup\$
    – boromyr
    Commented Jan 14 at 15:06
  • \$\begingroup\$ Transistor might be chosen for high \$H_{FE}\$ so that DC voltage loss across it is minimized. To avoid very high frequency oscillation, a low gain-bandwidth product might help. Component selection for C1 might make ferrite beads unnecessary but risky. Any added ferrite should be close to Q1. \$\endgroup\$
    – glen_geek
    Commented Jan 14 at 16:27
2
\$\begingroup\$

If you see a 120Hz sawtooth on the output of your boost converter, and the device is battery powered, the question is whether it is a measurement issue or not, and in the latter case, how did it get in there...

So first set the scope to trigger on mains. It's probably in the "trigger source" menu. When that's done, the scope will trigger on mains voltage zero crossing. If your "120Hz" remains steady on screen, it means it's synchronized to mains. If it sweeps then it's not synchronized and it does not come from mains.

Assuming it is synchronized, next step is to check if it's a measurement artifact or not. You can do that by probing the circuit ground with your scope. You should get a flat trace, with a little bit of switching noise from your DC-DC. Then without moving the scope ground clip, probe VCC. If you see the same sawtooth when probing GND or VCC, then... the noise is not on VCC, it's probably being picked up by your probe.

Note most scopes are Earthed. The guitar isn't Earthed but the guitar pedal you're working on is probably connected to a guitar amp that is Earthed. So you have a ground loop with the scope. You can check if the noise you're measuring with the scope goes away when you disconnect the output jack, which breaks the ground loop.

If you conclude the 120Hz sawtooth really is present on VCC, then it can either come from the boost converter:

  • feedback resistors picking up some noise
  • or it operates in light load power saving mode

...or it can come from variable output current wiggling the DC-DC's output voltage. I mean if there is a 120Hz signal on the input jack, your JFET amp is going to pull variable current from the supply, which will introduce some ripple on the supply. But in this case the supply ripple is not the cause of the problem, only a symptom of hum being picked up at the input.

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

This circuit does not reject supply noise. VOUT is equal to VCC minus the drop across R17 (which is proportional to the input signal); so VCC directly feeds to the output.

You need some way to filter or regulate VCC. This circuit doesn't draw much current, so a simple zener diode shunt regulator would work. You might try adding a 1k R between VCC and the top of R17, then a 5 V (5.1 is available) between that node and ground.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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