THVD1552: Radiated emission with THVD1552

Son Nguyen said:

Common mode filter on the output

Was this placed close to the differential output pins? From the schematic I see, placing it after the caps would have minimal effect.

Son Nguyen said:

Filtering/slowing the output signal.

What are you doing to filter/slow the output?

Son Nguyen said:

there is another THVD1552 in the other end of the Cat6 cable.

How long is the cable?

Son Nguyen said:

Your communication looks like it's full duplex. You likely do not need the 220 ohm resistors present on the driver side.'

Son Nguyen said:

the other end of the Cat6 cable

Cat6 cable has a standard characteristic impedance of 100 ohms. You typically want to terminate your receiver node with the same characteristic impedance of your cable. So in this instance a 100 ohm resistor on the receiver would likely perform better in minimizing reflections.

Generally you will also want to choose a device that has a slower rise/fall time. Faster can be bad for EMI, usually customers who want better EMI will choose a transceiver whose max data rate just barely meets their required system.

For further EMI protection, you typically want to use split termination with a cap between the termination.

Other sources of imbalances could come from:

• Layout: specifically trace length mismatch of diff. traces and trace impedance control (100 ohm characteristic impedance for RS422 differential traces are recommended)
• Cabling: impedance controlled twisted pair cables with shielding would provide the best results (more twists the better)
• Mismatches in passives: (example you have 10uF caps, but what is the tolerance, the bigger the difference the larger the potential reflection will be)
• Network topologies: (Daisy chained connections provide the smallest stubs and will minimize reflections)
• Stub length: If your RS485 transceiver is fast (THVD1552 has a 2ns typical rise/fall time) then the stub needs to be very short to minimize the reflection

Son Nguyen said:

Shouldn't the chip be able to pass the emission test? I see the RS485 interface as an old and proven technology

You can still fail EMI with a differential hardware layer communication standard. The more unideal the system, the more imbalances will occur. 

I also see you're using an AC coupled set up, this usually ends up with more EMI because it requires more toggling of the bits to break through the capacitive barrier typically done with Manchester encoding where the clock frequency being encoded ends up generating a larger frequency response. 

RS485 is also more power hungry (more power means you will likely have more EMI), If your distance is relatively short, you could look at using LVDS which is a lower power differential communication standard that would likely perform better at reducing EMI.

-Bobby

Hi Bobby

Thx for the fast response

The common mode filter was placed just after the output of the driver. 

for the filter we tried different configurations, both with RC filter, capacitor directly on the output, and a snubber like configuration. 

We tried 3m and 1m cables and even a short loop back of a few centimeters.

Yes we know that the impedance of a cat6 cable is 100ohm, the driver is able to drive 54ohm so we used 110ohm in both ends, we did also try with 100 ohm resistors. 

The differential signal looks really nice and clean with no reflections and ringing, so I believe that the termination is good enough.

We had a lot of focus on the impedance and the stubs during the layout phase, but i guess that is something that you always can improve a little. 

We shorted/removed the caps to take the capacitors out of the equation.

there is only 2 devices in our system, that makes the network topology easy

Hi Son,

Son Nguyen said:

The differential signal looks really nice and clean with no reflections and ringing, so I believe that the termination is good enough.

Do you have a scopeshot of the differential pins on both the driver? (without the cable)

And do you have a scopeshot of the differential pins on both driver and receiver? (with the cable)

Are you able to use split termination?

Have you measured the common mode drift/delta? 

-Bobby

Hi Bobby

Below is screenshot of Tx & Rx with 2m cable I don't have without cable. 

 We did try split termination similar to this with different capacitor values.

Yes the common mode delta is within the spec of the driver of +-200mV (I believe that my colleague measured it to 150mV), with a filter we manage to lower this to around 20mV. 

Hi Son,

The waveforms look really good from what I can see. There's a small amount of ringing after the transitions but its pretty minimal. The only other thing that would minimize the transitions further would be series resistors though I typically don't recommend this since it affects max distance and max data rate. (You could try adding ~110ohm series resistors on the differential outputs (Y/Z). 

Other idea....

Are you able to get samples of the THVD1452? It should be pin to pin with the THVD1552. 

EMI emitted is affected by the amount of power that occurs during transitions. I'm thinking you could try to run the THVD1452 at 3.3V. I think this would probably help. 

-Bobby

Hi Bobby, we made an PDN analysis to optimize decoupling capacitors and the Power plane. I added an additional 3V3 supply voltage so we can try out THVD1452 as an option.

I will update you in a couple of weeks when we have evaluated the new PCB. 

We tried the series resistors already, that didn't help us much.

Sure thing, let me know how it goes.

-Bobby

Hi Bobby

The new power planes and improved decoupling (from PDN analysis) didn't help at all. We made an option with THVD1452 and supplied it from 3.3V that made a little improvement but still far from enough.

Do you have any evaluation boards that we can buy and try out? 

We plan to make a new PCB that includes a POE transformer. 

Son Nguyen said:

Do you have any evaluation boards that we can buy and try out? 

From our half duplex RS485 line up:

https://www.ti.com/tool/RS485-HF-DPLX-EVM

Full duplex:

https://www.ti.com/tool/RS485-FL-DPLX-EVM

These probably aren't 100% optimized for what you're doing though.

You may need to look for a device with a slower edge rate if switching to 3.3V didn't work. Problem is I am not sure if we have a 10Mbps device in that DGS package that you're currently using. You could try SN65HVD1473 (20Mbps). I believe there should be a pin to pin one that you can use to fit on your existing set up. 

-Bobby

Hi Bobby

The conclusion from us is that it is simply not possible to pass the radiated emission test with a unshielded cat6 cable.

Product management had a big wish to use unshielded cable as the shielded cable are too stiff and thick. 

That forced us to make another architecture. 

Thanks for your support anyway. 

Hi Son,

Thanks for following up. Just to be clear, did you try to use the SN65HVD1473? I figured this would probably get you the best result. 

-Bobby