How to reduce/eliminate charge pump noise?

I use a MAX3232 line driver, which when the RS232 is plugged in, introduces too much noise on the Vcc line (80mv or even 100mV observed). This cause one of the components on the board to fail in such a way that it could still be communicated with over SPI but it did not do its job (basically an RF module, the noise was enough to screw up the receiver).

My question now is - how can I prevent this noise from the charge pumps of the MAX3232 from coming on to the Vcc line??

Thanks

I agree, that MAX232 noise is often underrated. You need suitable bypass capacitors for both MAX232 and RF module, and preferably ferrite beads to filter the supply.

FvM said:

I agree, that MAX232 noise is often underrated. You need suitable bypass capacitors for both MAX232 and RF module, and preferably ferrite beads to filter the supply.

Click to expand...

How do I use ferrite beads?

---------- Post added at 00:56 ---------- Previous post was at 00:56 ----------

FvM said:

I agree, that MAX232 noise is often underrated. You need suitable bypass capacitors for both MAX232 and RF module, and preferably ferrite beads to filter the supply.

Click to expand...

How do I use ferrite beads?

How do I use ferrite beads?

Click to expand...

Generally, they'll be utilized in power supply filters, perferably designed as pi-filters. Filters can be used to isolate an interference source as well as the sensitive circuit part. The problem with RF transceivers and some other functions like PLLs is, that they are sensitive to interferences in a lower frequency range of several 100 kHz. Thus you need rather large bypass capacitors (e.g. several uF ceramic) for sufficient effect. As a first step, I would place a series ferrite bead (e.g. 600 to 1000 "ohm") and a large bypass cap at the transmitter supply pin. If you are not sure about good bypassing of the Vcc net, place a second capacitor at the other side of the ferrite bead and get the said pi-filter.

If you don't have ferrite beads at hand, a series resistor of a few ohms should have a similar effect at low frequencies, of course the bypass capacitor is essential.

P.S.: A good ground connection (preferably a continuous plane) is taken as granted. Otherwise the effect of bypass capacitors may be dwarted.

FvM said:

Generally, they'll be utilized in power supply filters, perferably designed as pi-filters. Filters can be used to isolate an interference source as well as the sensitive circuit part. The problem with RF transceivers and some other functions like PLLs is, that they are sensitive to interferences in a lower frequency range of several 100 kHz. Thus you need rather large bypass capacitors (e.g. several uF ceramic) for sufficient effect. As a first step, I would place a series ferrite bead (e.g. 600 to 1000 "ohm") and a large bypass cap at the transmitter supply pin. If you are not sure about good bypassing of the Vcc net, place a second capacitor at the other side of the ferrite bead and get the said pi-filter.

Click to expand...

Are you suggesting that the pi filter should be placed on the Vcc line close to the Vcc pin of the MAX3232? Can the decoupling capacitor that is already there be part of the filter? And since I have two such line drivers, should each one have its own filter or can they share one?

I suggested to place a filter at the RF module in a first step. But filter at the MAX232 is the other option. The decision depends on the existence of other interfernce sources resepctively sensitive circuit functions.

If the two MAX232 are close togther, a common filter seems reasonable. You need sufficient capacitance on the MAX232 side of the filter to buffer the supply spikes. If the present bypass capacitor serves this purpose, it should be O.K.

I would recommend filtering not only the supply pins of the transceiver but also the charge pump with small chokes. You want to confine all high frequency currents to very small loops, both at the emitting source and the side being interfered with. So put chokes/ferrite beads on both of them.

Also keep in mind that once you connect your the MAX232 to a RS232 cable or whatever, you introduce the possibility of large common mode signals in your board, which are much more difficult to suppress. You may find that completely solving the issue requires designing the board such that the MAX232 and transceiver are partitioned with different supply rails.

You might want to dig a little more for the source of the
noise.

For example this charge pump will push current spikes
at the supply and the ground - which path, or both,
are really perturbing the neighbors?

Very good decoupling from VCC to GND can merely split
the pain, injecting VCC return spikes into GND where it
should have come from VEE (or whatever).

Now you have the switching spikes, and you have the
charge being shuttles as square-ish (or trapezoidal)
slugs. These want different sorts of decoupling, and
not necessarily the highest Q, best ESR / ESL caps.
The better the MAX output (RS-232 rails) cap quality
the sharper the edges on the pumped element of the
supply ripple. Maybe these filter caps don't want to
be so good, maybe you even choose to add some series
resistance if the RS-232 current can stand to be limited.

I'd look at the 'scope waveforms of VCC and current,
see what you can deduce about the supply current and
the nature of the "supply noise".

According to my observations, other than typical switching noise in digital circuits, switching edge rise time isn't the problem with MAX232 interferences, rather than spike area. Bypass capacitor placement matters to some extent, also ESR, but more total capacitance. The below waveform shows the VCC 150 mV voltage drop during each MAX232 charge pump switching event observed in a uP EvalBoard with only moderate bypassing. In this case, it was sufficient to knock out the USB connection of an ATmega USB processor, if the spike coincides with an USB transmission and hit the USB PLL in the right moment.

Some time ago when I started my journey into mixed signal design, I came to conclusion that is is wise to add a PI-filter (like 2 caps and inductor) to power supply of every active component. If a component has several power supply pins, then you may want to bypass/decouple them all separately. In the beginning I just use a single decoupling capacitor (as they are recommended anyway and much cheaper than inductors/chokes) and short the place for the choke with 0R resistor. If some problems arise, then I have possibility of just replacing 0R with choke and adding a capacitor to get pi-filter. This prevents some tedious hacking when you detect noise problems.

Also, i tend to use one LARGE (220+ uF)cap per board if it is under 10x10cm. over that size I use 2 for each supply voltage. I also prefer to place one at least 10 uF tantalum roughly on every 1'' x 1'' square of the PCB.

If your max232 introduces strong noise, you may want to add some kind of RF shielding to it. It is unlikely that max232 generates noise at such high frequency that an RF shield does anything, but it is nice to leave yourself an opening (i.e. a ring of copper without soldermask to which you can solder RF shield). Of course if this is a volume design where every penny and every square mils matters you don;t do such things