The experimental setup of my actual toy project consists of several units, all connected to the same DC power supply (5V). The noisiest one is an oscillator (11MHz) and LED driver (38mA). Both are realized with one old TTL IC, the 74F02.

To measure the ripple, I inserted a 1 Ohm resistor into the power line and measured a voltage ripple of 105mVpp across it. An oscilloscope screen shot and its fourier transform are shown below. The fundamental frequency is 11MHz as expected, and several harmonics have about the same strength.

In order to suppress it, I came up with the schematic shown below, where the square stands for my oscillator and driver circuit. My reasoning as follows: The energy source is at left, but the ripple source is at right, so the LRC circuit should be read from right to left, and represents a 2. order lowpass. I calculated the values 2uH and 1uF for -40dB at 11MHz, but it doesn't work. So I tried several combinations from 1 to 2000nF and 1 to 220uH, but did not even achieve a factor of 2 in ripple reduction.

So I guess my concept is wrong. How would it be done right? By the way, I am mounting through-hole.
right? By the way, I am mounting through-hole.

 

Are you sure it's not the power supply?
Normally a 100nf cap connected from Vcc to gnd and as physically close as possible is all that's required.
Post the schematic of everything using the 74F02.
105mVpp indicates 105mApp
Seems you maybe overdriving the chip.

 

Are you sure it's not the power supply?

Of course. A power supply delivering 38mA with a ripple of 105mApp would be a pretty strange equipment.

Normally a 100nf cap connected from Vcc to gnd and as physically close as possible is all that's required.

Actually, it is only 10nF.

Post the schematic of everything using the 74F02.

Attached below. The LED is connected at JO02.

105mVpp indicates 105mApp
Seems you maybe overdriving the chip.

That might be, but I don't care because the circuit does what it shall do.
Let me add that I don't want to merely solve the ripple problem, but I'd like to know why the lowpass described in my original posting does not work. After all the problem should be common in short wave RF circuitry, which exist for 100 years now. So well approved standard solutions should exist.

 

The decoupling capacitor across oscillator supply rails is required and needs to be sized accordingly to its current consumption. Having 10nF is clearly not enough. This is not just to reduce ripple: if circuit drains a lot of current, it will do it through all series components (inductor, resistor) and high frequency current will travel all the way to the battery radiating the frequency you are trying to suppress. Having local bulk capacitor will provide local supply for this current (especially if some chips are in overdrive), after that remaining current will still try to find its way through inductor, but will be limited by its impedance and capacitor on the left of it will kill the harmonic off and your resistor will see clear current with very low harmonics.
So your LC filter needs to be PI-filter: C-L-C.
Another tip: using inductor may lead to oscillations (less likely if capacitors are big enough). To prevent that you can put resistor across the inductor (say 100R) - it still will be effective, but it will not be able to oscillate. In old tech such inductors sometimes were wound on the body of through-hole resistor for exactly that reason. Ferrite beads have similar characteristics, they are not just inductors.

Here is a simulation results for your circuit compared to the one I just described - just adding one capacitor (Green is voltage ripple spectrum, blue is current from battery spectrum):

 

Of course. A power supply delivering 38mA with a ripple of 105mApp would be a pretty strange equipment.

Actually, it is only 10nF.

Attached below. The LED is connected at JO02.View attachment 299508

That might be, but I don't care because the circuit does what it shall do.
Let me add that I don't want to merely solve the ripple problem, but I'd like to know why the lowpass described in my original posting does not work. After all the problem should be common in short wave RF circuitry, which exist for 100 years now. So well approved standard solutions should exist.

The decoupling capacitor across oscillator supply rails is required and needs to be sized accordingly to its current consumption. Having 10nF is clearly not enough. This is not just to reduce ripple: if circuit drains a lot of current, it will do it through all series components (inductor, resistor) and high frequency current will travel all the way to the battery radiating the frequency you are trying to suppress. Having local bulk capacitor will provide local supply for this current (especially if some chips are in overdrive), after that remaining current will still try to find its way through inductor, but will be limited by its impedance and capacitor on the left of it will kill the harmonic off and your resistor will see clear current with very low harmonics.

Thanks Oleksandr to take the time for such a detailed analysis.

What has been puzzling me for a while ist that capacitor value. In a lot of schematics around you see this 100nF value. On the other hand, the usable frequency range of capacitors decreases with capacitance. The only generic site with specific values I foundon the internet shows an intrinsic resonance frequency of just above 10MHz for 100nF, which is about my fundamental frequency. 10nF is a little below 100MHz. Searching sites of manufactorers doesn't help, because they specify only SMD capacitors and nothing about obsolete THT components. Any comment on this?

So your LC filter needs to be PI-filter: C-L-C.
Another tip: using inductor may lead to oscillations (less likely if capacitors are big enough). To prevent that you can put resistor across the inductor (say 100R) - it still will be effective, but it will not be able to oscillate. In old tech such inductors sometimes were wound on the body of through-hole resistor for exactly that reason. Ferrite beads have similar characteristics, they are not just inductors.

I will definitely try a Pi filter. I doubt somewhat your simulation because of the use of ideal components, but with some experimenting i should achieve some progress. I will post the results here. Thanks again.

 

The decoupling capacitors are not only selected by their frequency characteristics. They provide a local power supply delivering energy via small loop as they are positioned close to the load. That is why every IC must have its own decoupling capacitor very close to it. The value also matters: the more energy is consumed at high frequency spikes, the larger value is required. That is why generic logic IC usually uses 100nF value. Something like RF module or large MCU would require 1uF or 10uF. You would also put 100nF in parallel because capacitors of very different values may have different frequency response and doing so reduces resulting impedance further in wider frequency range.
You can read a bit more about decoupling capacitors here: https://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf

As for the simulation results, you are right, they are only as good as the input information provided. But in this case absolute numbers are not important, it just demonstrates general effect of having or not having a decoupling capacitor. For that purpose those results should be good enough