Optimising RS-485 for lowest EMI and lowest power
There are many great white papers on the subject [Reference 1]. This article provides missing details, or simplifications that can enhance the performance of your design.
Figure 1. Basic RS-485 topology
EMI considerations
There are three primary tools available to the designer to manage EMI (shielding is covered separately at the end of the article).
1. Device speed
2. Transceiver operating voltage
3. Terminating resistor currents
Speed
It has been said before, but bears mentioning again. Do not use a baud rate faster than is needed for the application – and that includes the speed of the transceivers. Transceivers are available in different speed options that affect the rise/fall times of the signals. For instance, many RS-485 links run below 1 Mbps, so a device such as TI’s SN75HVD12DR is a good choice. For 128 kbps links, the slower Intersil part would suffice.
The slower rise time of these parts (e.g., 100 nsec, [Ref. 2]) is more than fast enough for these applications, and minimizes EMI radiation. It may also reduce susceptibility from nearby noise sources since it has a slower response speed. Read the specs of the transceiver, as many standard devices will run at 10 Mbps or faster which is much higher than typically needed for these links.
|
ISL8483E (Intersil) |
250kbps |
|
SN75HVD12D (TI) |
1Mbps |
|
SN75HVD11D (TI) |
10Mbps |
Table 1. Example RS-485 transceivers and speeds
Voltage
EMI is proportional to the voltage swing of any signal. Reduce the voltage swing and you can reduce the EMI radiated by the connections. Many newer devices are fully rated to operate at 3.3V while satisfying the minimum requirements of the RS-485 signalling standard. As a bonus, 3.3V is more common these days than 5V in many system designs. What do we give up by using a lower voltage? Speed capability and noise immunity are reduced at this voltage. But if the device is rated for the speed required, and shielding is used, a 3.3V RS-485 signal is usually adequate. Again, it is up to the designer to consider all relevant conditions, and check the datasheet. As a side note, resist the urge to add a capacitor across the input of a transceiver unless you compute the frequency response as being 5-10 times that of the signalling rate (1/2 the baud rate) to avoid signal degradation.
RS-485 has a wide operating voltage range, from an Rx threshold of 200 mV to a maximum differential signal of 10V. Usually 2V P-P is the minimum recommended drive level, and 3.3V devices will meet that criterion while interfacing just fine with 5V-powered receivers, providing reasonable signal-to-noise levels, especially for shorter runs. Keep in mind that if you need high speed (defined as >5 MHz), you may need 5V power, so check the datasheet.
Current
Because EMI problems can be magnetic in nature, the currents flowing through the terminating resistors may be considered a factor. Magnetic interference can be more difficult to control as copper has a relative permeability of about 1, and may induce coupling from an aggressor circuit in spite of nearby shielding. Lower transient currents reduce the magnetic signature, and aid in minimizing coupling to other nearby circuits.
How do we do that? Isn’t the terminating resistor value fixed? No it isn’t, as long as your cable is not “electrically long” relative to the edge rate of your signals. There is no rule that says you can’t raise the value for other engineering reasons. If the primary issue or concern is susceptibility rather than radiation, then the lower the termination resistance, the better. But as in all engineering designs, there are trade-offs. Comparing a 5V/120Ω system with a 3.3V/499Ω system shows a factor-of-six reduction in current.
Termination resistors
This design criterion is very susceptible to “rule of thumb” abuse. The default value most of us learn initially is 120Ω applied differentially across the (+) and (-) data terminals at the far ends of the network. But 120Ω is not always the best choice. The original terminations were selected to match the impedance of twisted pair cable commercially available. No matter your application, don’t consider running without, even for short runs, as it provides noise immunity. Termination is required for two reasons:
Cable is “electrically long” such that 2·tp ≥ tr/5 [Ref. 1], where tp is signal transit time, one-way, across a cable, and tr = rise time of the signal from a given driver (10%-90%; see below for computing signal transit time based on the velocity factor). If the cable is not electrically long, you have more flexibility in adjusting the termination value (Rt). This is another reason to use as slow a driver as will satisfy the application.
Susceptibility: Without any termination, a receiver input (single-ended) for the SN75HVD12DR is estimated at about 109 kΩ (based on maximum input current spec and 12V on pin). Input impedance this high is susceptible to crosstalk from nearby signals on a PCB or within a cable (if more than one pair is under a shield). Lowering this impedance by applying a terminating resistor in parallel will minimize crosstalk, but at the expense of power dissipation. A compromise is recommended, but never give away “free” noise immunity; always include some value of termination.
Optimizing the terminating resistor
The first question to be answered is: Is the cable electrically long? Then we can determine the need to match Rt to the cable impedance. The answer to this starts with the above equation: 2·tp ≥ tr/5.
This is a conservative, not a hard-limit. Plugging in the rise time for the previous driver, we get: 2·tp ≥ 100nsec/5; thus a maximum 10nsec time of flight, tp. That means that the signal can take no more than 10 nsec from the driver to the opposite cable end (before hitting the first terminating resistor). Next we compute how that requirement translates to cable length based on its dielectric (and thus its velocity factor). Velocity factor is the ratio of signal propagation speed along a conductor/dielectric structure, as compared to the speed of light in a vacuum.
VF (velocity factor) = 1/√εr
(εr = relative permittivity or dielectric constant; e.g., polyethylene = 2.25)
Thus, speed = c · VF, or, c/√εr
An example: Belden specifies 66% VF for its #9841 cable [Ref. 3] (the insulator is polyethylene, which when you compute the equation above equals 66.6%). If the VF is not specified, lookup the dielectric material and compute as above.
Calculating the maximum length of cable before a terminating resistor matching the cable impedance is required (using the above 100 nsec driver and 10 nsec maximum tp):
vprop = c (3·108m/sec) · VF (0.66) = 1.98·108m/sec
Lmax = vprop · tp (10 nsec) = 1.98m
Combining and simplifying:
Lmax = (tr / 10) · c · VF
or in nanoseconds and metres:
Lmax = (tr / 33) · VF
Low power considerations
Given our example, a cable shorter than 1.98m (an embedded application cabinet for instance) does not require a terminating resistor matched to the cable. How do you select Rt then? You can go without. However, as was mentioned above, a lower value will reduce susceptibility. A good starting value is 499Ω. For 5V signalling, the static power dissipation drops from 208 mW to 50 mW per resistor when changing from 120Ω to 499Ω. So this lowers the power, and reduces susceptibility over the non-terminated case, making for a robust communication link.
Why do we care about power dissipation if it isn’t a battery driven application? SMT components have become much smaller in current designs. A 0603 resistor has a power rating of 100 mW. With a 50% derating factor, that leaves us with 50 mW. A 499Ω 0603 resistor meets the requirement, whereas the original design at 208 mW would require a 1210-size resistor. The higher value resistor helps reduce the footprint of the design while keeping the link robust. Changing to 3.3V/499Ω signalling allows a 0402 terminating resistor.
Low power also means you might be able to get away with charge-pump regulators for the 3.3V or 5V rail. This provides fewer parts and potentially lowers overall cost. As an example, Linear Technology’s LTC3255 works with an input of 4-48 VDC, but is limited to 50 mA output.
Managing common mode voltages
While RS-485 is a differential network, there are finite common mode (CM) voltage limits that must be met for proper operation. This means a ground wire is required from one node to the other if they might be floating WRT each other. Yes: a CM voltage can be derived from the data signals (using diodes/capacitors), but in general, that will not be as noise-immune. For control of noise from one module to the other, an inductor can be tried in series with the ground lead. This will allow control of the DC common mode voltages while minimizing RF return currents through either of the data lines. Typically a 1 µH inductor will provide good protection, as its impedance will be above 1 kΩ @ 200 MHz, but not so high as to interfere with communications at or below 1 Mbps. A common mode choke can also be employed for additional RF protection. Since the power source for the remote transceiver is sometimes an isolated DC-DC converter, the ground line inductor will minimize RF noise currents through the DC-DC converter’s parasitic capacitance.
Shielding
Controlling EMI doesn’t mean that all grounds are floating above earth ground. As a general rule, the shield (if used) is tied to the signal ground at the end or node which has earth ground ties, while the other end of the shield is floating. In cases where severe magnetic interference is expected, tying the shield at both ends may be required. As cable lengths exceed 10m, the shield ties at remote locations are then converted to “soft grounds” (using capacitors) to minimize low frequency disruption through the shield (connecting distant grounds). Noise modelling using a voltage source and parasitic capacitance tied between the relative grounds, conductors, and the shield, can provide the guidance needed to make an informed choice. Understand this topic well, rather than relying on rules of thumb (see Figure 2 below as an example). There is no right answer for all applications. For further reading, I recommend:
1. Grounding and Shielding Techniques in Instrumentation, Ralph Morrison, Wiley-Interscience, ISBN:0-471-83805-5
2. Electromagnetic Compatibility Engineering, Henry W. Ott, John Wiley & Sons, ISBN:978-0-470-18930-6
Figure 2. Noise modelling of basic shielding
Conclusions
RS-485 can live up to expectations for an EMI resistant simple network, but only if you understand available options, and apply sound engineering principles. Many choices are available. Don’t simply pick parts you have seen or used before; many new devices have come to market. Consider the length of the run, device speed, power requirements, and the type of cable you are planning to use. This will help ensure your design will work right the first time.
Bibliography & References
1. Interface Circuits for TIA/EIA-485 (RS-485), Literature #SLLA036D, Texas Instruments
2. Datasheet, Literature #SLLS505N (SNx5HVD1x 3.3-V RS-485 Transceivers), Texas Instruments
3. Datasheet, P/N:9841, Multi-Conductor – Low Capacitance Computer Cable for EIA RS-485 Applications, Belden, Rev 2, 05-06-2016 (metric)
Also see:
Mark Wagner (BSEE) is a 25-year industry veteran with experience in circuit and PWB engineering for instrumentation and mixed-signal design. Sensor interfacing and precision analogue are specialties.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
