| ??? 06/20/08 02:24 Read: times |
#156056 - How would that come to pass? Responding to: ???'s previous message |
Frieder Ferlemann said:
unless you have a very good reason.
What does a resistor in the GND connection mean? Let both outputs T1OUT and T2OUT be on -10V and R1OUT be at 0V and R2OUT be at 5V and R1OUT being connected to RX of an 8051. Now switch both T1OUT and T2OUT to +10V (against a resistive or capacitive load). A load of 3 k-Ohms is the specified maximum. The current needed to drive both lines will cause a voltage drop over your resistor in the GND connection. This very voltage drop will cause a lower potential for R2OUT (where it does not hurt) and a lower potential for R1OUT (where it does hurt). I'm not sure I understand how the receiver on R2 can be pulled down to a level at which it will be a problem. The standard threshold is at 2.65 volts, for these MCU's, isn't it? Just how much current do you suppose has to flow through those resistors in order to make an appreciable difference to either the nominally TTL levels on the MCU side or the minimally 3-volt level of the RS232? How could such a condition occur? Essentially you are feeding (part of) the two output driver currents into the protection diodes of a single 8051 input. Yes, and what are those currents? Keep in mind that with a ratio of 2:1 between the R in Vcc and that in Gnd, (looking at is as a loop current, which it's not) and the current specified is only 1.6 max for a high and 1.0 for a low. Plus the rationale "lower the noise" is only partly met:
you'll find part of the switching noise on R1OUT and R2OUT (which typically are connected to the heart of your circuit). Likewise you'll find part of (external) noise applied on T1OUT and T2OUT to propagate on R1OUT and R2OUT. The on-board regulator, a darlington-input 7805 or similar, can easily deal with the smoothed noise imposed on the primary circuit after the R and C components at the MAX232, so the worst-case issue will be a very slight diminution of Vcc to Gnd potential. Unfiltered, this would probably not be manageable by a regulator. A switcher couldn't deal with it at all. Switchers are too slow to respond to changing output circuit conditions. (ESD not yet mentioned)
In short: don't use a resistor in the GND connection unless you have a very good reason. One very good reason is that the circuit under discussion is the original MAX232, which, from my observation, is much noisier than any MCU or other logic circuit. Secondly the currents surging in and out of the MAX232 have to be handled by the capacitor(s) attached to its Vcc and Gnd leads. Hopefully there's a low-value cap within 1 mm of each supply pin and a larger one very close as well. The key is to match the resistors with the capacitors at the nominal frequency of the oscillator in the MAX232. That will tend to keep the switching noise currents recirculating via the cap's on the MAX232 side of the resistors and thereby substantially smoothed out. Yes, there will be an aggregate level shift, but I doubt it will typically exceed the noise amplitude of about 150 mV and considerable power that would otherwise be imposed on the entire circuit by the MAX232 with its 1 uF caps. I've done this sort of thing from time to time, and find it easier to tolerate the slight impact on Vcc to Gnd potential as opposed to 150-200 mV of additional noise between Vcc and Gnd. I've observed no hits due to level shifts at the MCU, nor have I observed any impact on the RS232 circuit. Maybe I'm just lucky. The noise from a MAX232A is less significant, so I use those when I don't have appropriate supplies available. RE |
