Would it be possible to replace these combinations of ceramic and electrolytic capacitors with just a pair of low-ESR 10 μF ceramics?

Conversion from 12 V to 5 V, minimal consumption, only MAX485 power supply.

enter image description here

  • 2
    \$\begingroup\$ MAX485 sounds like the opposite of "minimal consumption"? RS485 implies a rapidly changing, many-mA bus draw? And at the price of a single MAX485, why are you eager to save 4ct on ceramic capacitors? \$\endgroup\$ Sep 7, 2022 at 13:15
  • 2
    \$\begingroup\$ Without taking time for analysis, I can just say that higher ESR electrolytics may have important zeros in the transfer function that help control the gains in certain terms. That more applies to LDO types. The 7805 is an NPN type regulator and doesn't require input or output capacitors (memory serving.) It's stable under almost any condition as it uses a Darlington emitter follower arrangement with voltage gain near unity and the current gain high, which doesn't introduce any low freq poles into the loop gain. Dominant pole compensation is really easy so external compensation isn't needed. \$\endgroup\$
    – jonk
    Sep 7, 2022 at 13:29
  • 1
    \$\begingroup\$ What data sheet are looking at, the ones I see typically show them. Some state:". All characteristics are measured with a 0.33-µF capacitor across the input and a 0.1-µF capacitor across the output". My recommendation is to follow the manufacturer's data sheet. If you will use other brands put them in. \$\endgroup\$
    – Gil
    Sep 7, 2022 at 13:41
  • \$\begingroup\$ 7805 is pretty stable and it will work. Output capacitor big value need to compensate load current change. If load stable, value can be lower. Input capacitor need to reduce pulsation, min voltage should not be under approximately 8V, so the value depends on max load current. \$\endgroup\$
    – user263983
    Sep 7, 2022 at 13:48

2 Answers 2


The MAX485 is a 2.5MBaud, non-slew-rate-limited transceiver or line-driver. It is designed to send very fast, high-current pulses through a serial cable.

It's very nature implies that it will be demanding of a power supply. It may be off 99% of the time (minimal consumption)... but the 1% it is on and sending pulses, this could thrash and abuse these bypass caps, especially if powered from a weak source, like a 9V / PP3 battery.

The LM7805 shows very low-value bypass capacitors. This is fine if powered from a strong source (and powering a weak load.) If you're experiencing sags in the power due to a weak source and strong load, increase the incoming bypass capacitance first. How much, depends on how "active" the load is and thus how bad the sag is. And if the voltage sags are due to overloading the LM7805 (it can't regulate enough or fast enough) then increase the output bypass capacitance. Careful though, too much capacitance can make even linear regulators unstable. Will need to use an oscilloscope to "see" this.

Typically, the incoming 10µF is electrolytic for higher voltage rating per size. A ceramic actually has to be rated for 2x that voltage at least, because a ceramic's effective capacitance drops significantly as the voltage increases. It might even make sense to use two ceramics of half the capacitance but twice the voltage, in parallel. This gives the same net capacitance, but lower ESR and better value stability at nominal voltage. Or use a hybrid electrolytic/solid cap (those seem to have very nice specs.)

So if designing this on a printed circuit board, suggest adding pads/holes for a variety of input and output capacitors. Then assemble a prototype, populate some caps, and measure the performance in operation. May find that more or less capacitors work better. (If designing using a "plug-in breadboard", the contact resistance and other parasitics are likely influencing operation.)

  • \$\begingroup\$ It's a differential load current and thus constant supply current except for RF transitions and unbalanced loads \$\endgroup\$ Sep 7, 2022 at 14:05

It's also possible to add nothing or just use only 0.1 uF as it all depends on your load spectrum and step load noise tolerance.

The 7805 and alike are internally compensated up to a unity gain bandwidth that may or may not be less than the step load bandwidth of your application.

If you want more details, add your specs for cable capacitance, data rate and any step load error tolerance measurements.


Use 0.1 on input and output only. 10uF may reduce a 100 mA step error but extend the duration time. Your differential RS485 load is unlikely to experience any significant step loads until enabled as differential loads tend to be constant supply currents.

It's not that critical for logic levels on RS-485.

The bulk caps were recommended by Adafruit for hobbyists who might step anything on it and also another answer on this site.

You could verify my assumptions, add it now and delete it later, but I might not as I have done these types before.

other info FWIW

What an old 7805 looks like from the inside

enter image description here

The big 0.3 ohm series emitter output resistor on the right below is what gives the amplifier stability but also step errors. 16 mohm at 1kHz due to negative feedback but Re=300 milliohms so a 1A step (5 Ohm) is guaranteed to drop at least 300 mV unless a bulk cap low ESR holds it up a bit more but causes ringing a bit longer. enter image description here A low ESR 10uF ceramic may be < 10 mohm.


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.