THVD1550: RS485 Signal attenuation

Hi Raghu,

I think your analysis is a very good starting point. (I say "starting point" here because it is difficult to arrive at an answer to this sort of issue without some kind of testing as well.) I think you may be overestimating the loss, though - it may make more sense to evaluate the insertion loss based on the maximum toggling rate of the data (250 kHz for a 500 kbps data stream) rather than the full signal bandwidth required in order to maintain the output transition time.

I double-checked the figures in the application note you mentioned, and I do believe that there are some typos we need to correct. Figures 23 and 24 show 400 kbps, Figures 25 and 26 show ~600 kbps, and Figures 27 and 28 show 1 Mbps. Thanks for bringing this to our attention - I'll work with our documentation team to make the corrections.

Based on these results, I would expect THVD1550 to work well across 1000 m of CAT5e at 500 kbps. Of course, the cable tested was "typical" and may not have aligned with the worst-case loss limit allowed for CAT5e, so you may want to double-check the loss of the cable you plan to use.

Note that there are some things you can do to improve performance as well. For example, if you wanted to improve the vertical eye, you might consider the THVD1450 device (which would provide a larger output differential magnitude than THVD1550 when operated at VCC = 5 V). If you wanted to improve the horizontal eye opening, you could introduce equalization. This is designed into some specialized transceivers (like SN65HVD24) but can also be implemented externally - see this reference design for an example:

www.ti.com/.../TIDA-00790

Please let me know if you have further questions or if this is unclear.

Regards,
Max

Hi Max
Thank you for the reply. Appreciate the tips on getting a better eye.
But what about attenuation? I worry about attenuation because my previous analysis that I performed was just at 20C. Our installation is going to be in the outdoors in hot climates. So a max temp of 60C is not out of the question. The cable attenuation increases 0.4/C. so at 60C, the attenuation would increase by about 0.4x40 = 16%.
I agree that testing is the best way to come to a final answer, but I would like to atleast make a ball part assessment to see if this design is worthwhile or its better to think of a different approach :)

So I guess my follow up question is... shouldn't there by some attenuation in those eye diagrams at 3000ft? How are they 2V pk-pk when the transmitting pk pk is 2.7V.... only 700mV attenuation after 3000ft? That sounds a bit unlikely no? And there is no change in pk-pk over distance or data rate?

Hi Max
I think I might have jumped the gun when I read the voltage division as 1V pk-pk. Could you let me know what the vertical division on the eye diagrams are for each lengths? That would help me with my analysis.

Hi Raghu,

You are correct that the scale for the plots is set to 1 V per division.

In most of these plots the signal is able to reach a steady-state amplitude within the bit period. This amplitude would depend not on the AC loss of the cable but on the DC resistance. A typical estimate for the resistance of 24 AWG cable would be 25 Ohms per 1000 feet. If there is 120-Ohm end-of-line termination, then, I would expect about 40% attenuation of the steady-state differential amplitude.

That expectation seems to line up pretty well with the plots. An RS-485 driver will have a minimum output amplitude of 1.5 V (e.g., 3 Vpk-pk) with a differential bus loading of 54 Ohms. In this case, a higher initial amplitude would be expected since (1) the device is likely a "typical" rather than worst-case device and is operated under nominal conditions and (2) the total differential loading will be greater than 54 Ohms. (It will be ~83 Ohms in this case, since 120 Ohms is placed on either end of a cable that adds ~75 Ohms series resistance on each half of the differential signal.)

Please let me know if you disagree with this line of thinking. (I saw you mentioned an initial amplitude of 2.7 Vpk-pk - where did this come from?)

Regards,
Max

Hi Max

The 2.7V pk-pk is from the data sheet for THVD1550 section 7.6 under differential driver output voltage magnitude. This same number is 3.5V for THVD1450 like you pointed out. 

My biggest concern I have is that we have all our RS485 nodes daisy chained for flexibility and the connectors to each of these will cause reflections. We have 128 nodes spanining across 1000m which is a lot of connectors with reflections. 

I do agree with your line of thinking but could you explain the math for how it all be 83Ohm... with 120 ohm placed on either ends of a 1000ft cable. I'm not sure I understand that. 

Thanks 

Hi Raghu,

The concern about reflections is valid but may be hard to estimate. As long as the connectors themselves don't substantially degrade the signal (e.g., by adding high capacitive loading or dramatic impedance discontinuities) and the cable connections to them are short with respect to the transition times of the signaling then generally a network of this size is possible.

My thinking for the 83-Ohm estimate was as follows:

- directly at the transmitter output for a terminating node there will be 120 Ohms
- in parallel with this will be the series resistance of the cable (75 Ohms) plus the far-end termination (120 Ohms) plus the cable resistance of the return connection (75 Ohms). In total this is 270 Ohms.
- 120 Ohms || 270 Ohms = 83 Ohms.

Please let me know if this is still unclear.

Regards,
Max

Hi Max

Absolutely. Yes. This makes total sense.  Thank you so much. 

I would like to ask a follow up question about isolation if that is ok with you in this same thread. Our nodes are spaced quite apart as you well know now. I was initially thinking about using isolators for both the supply and the RS 485 transceiver because the ground potential would definitely vary between these nodes. During lighting strikes or maintenance they could vary the Vcm to beyond the RS 485 limit (-7V to 12V).

But supply Isolation would cause a drop in efficiency to 60% and would also increase our BOM, so this is a problem from a cost perspective for our application. 

Now if we have error checking in our FW for every packet, we could always re-transmit our packet assuming this varying of ground potential during lightning strikes causes only a temporary disturbance in the communication. After the potentials equalize back to normal value, it should be back to working. What is your opinion on this thought? Is there something else we could be missing in our analysis for the need for isolation?

1. Is our assumption that a lightning strike or maintenance work would cause only a temporary disturbance in our communication correct? 

2. Could there be something else that would cause permanent failure in our system because we did not isolate the nodes supply and communication. 

Thanks

Raghu

Hi Raghu,

Sorry for the delay - I was out of the office yesterday.

Your assumption is correct. You should be careful not to exceed the "absolute maximum" pin voltage ratings of your RS-485 transceiver in order to avoid permanent damage, but within this range a common mode deviation that exceeds the recommended value should only result in a temporary loss of functionality. After the common mode normalizes, data communication could resume as normal. It doesn't sound like isolation is needed here to me. If you were concerned about application conditions stressing the RS-485 transceiver (e.g., by forcing voltages beyond the maximum ratings, whether DC or transient) then a simpler solution may be to just include some additional protection components on the bus (e.g., transient voltage suppression diodes to clamp ESD events, etc.). I would introduce isolation only if you need stable communication in the presence of large common mode shifts or across independent power/ground domains.

Max

Hi Max
One last question. We are thinking of operating our driver with a supply of 3.3V so that we do not require voltage translators. All the specs in the data sheet are given for Vcc of 5V.
Can the output voltage differential 'Vod' be greater than Vcc? RS485 are current drives, so I assume they should be but wanted to confirm with you. Maybe there is a headroom problem if Vod > Vcc?
We are using an 18 AWG wire. Over 3000 ft the load resistance comes to be 68 ohm. So extrapolating from the data sheet my Vod would be between 3.5V and 4V for typical performance. Is this correct?

Raghu,

RS-485 transceivers (THVD1450 included) generally are not able to drive voltages on the bus that are beyond their power supply voltages. So, VCC would end up being the upper limit on the VOD (peak) value, and in most cases VOD would be less due to some voltage drop in the output stage (due to the integrated reverse-polarity protection as well as the interaction between load current and the driver output resistance). For more information on the driver output structure and what influences VOD, you may want to reference this blog:

e2e.ti.com/.../rs-485-basics-the-rs-485-driver

Note that while datasheet typical values are based on VCC = 5 V, the minimum and maximum values reflect performance across the full range (3.0 V to 5.5 V) unless otherwise noted. Figures 2 and 5 may be useful in understanding the typical differential output levels for different VCC levels and load currents.

Referencing Figure 5 (for VCC = 3.3 V), it looks like the expected VOD with 68-Ohm loading would be between 2.4 V and 2.5 V. (I'm dividing the VOD on the y-axis by the IO on the x-axis in order to determine the equivalent load resistance.)

Regards,
Max