CML output swing for V5

page 112

page 202

MGTAVTTX = ? on your board?

Like many things, the GTP is programmable, and there are external supplies to set the common mode voltages to be whatever you would like them to be (in order to meet all the possible standards more easily).

Aust> I am using V5 GTP. As per my understanding, the GTP output should swing

between 1.2V and 700mV and common mode voltage of about 950mV.

between 625mV and 75mV with common mode voltage at about 350mV.

the V5 specification.

an idea on acheiveing this then please let me know.

Do you have the transmit impedance getting calibrated properly per chapter

Austin,

I am using Xilinx ML523 board. This board uses V5 LXT device. As per the schematics AVTTTX and AVTTRX are connected to 1.2 Volts. Does this mean that the common mode voltage is 1.2V? If so why am I seeing common mode voltage of about

Thanks.

Eddie

John,

I am using Xilinx ML523 board to test the GTP outputs. Shoud it not be claibrated already? Why is it drastically different?

Eddie

claibrated already? Why is it drastically different?

I don't know CML very well, but I know it needs the proper terminations for the right swing and common-mode range. Glancing through UG196, it seems the only place for the termination is on the transmitter. Are your signals open into 0.5 pF FET probes on an oscilloscope? Or are you driving them into 50-ohm probes, terminated to ground?

Often the problem comes down to the method of observation, not the signal itself.

- John_H

John,

Currently I have connected SMAs from ML523 to the Tektronix probe P7313SMA. This is a point to point connection and there is no other load.

I am using Tektronix P7313SMA probes. It says that the probe input has 50 Ohm termination. 13 GHz Typical.

Also I am using Tektronix DSA71254 scope. It is 12.5 GHz, 50 GS/sec scope.

Please let me know if I need to change my setup.

Thanks.

Eddie

This is a point to point connection and there is no other load.

termination. 13 GHz Typical.

Super!

Yes, the setup will cause problems. The current mode singals are current sinks if I understand CML properly. THe pull=up terminations in the transmitter will be pulled from 1.2V by that switchable current sink. If you have 50 ohms to ground, you now see an equivalent impedance of less than 50 ohms to an equivalent termination far below 1.2V.

Look at the receiver arrangement in the RocketIO user guide we pointed out yesterday. If you can configure your setup to mimic that receiver, the CML voltage will look proper again.

Personally, for looking at common mode voltage, I'd connect the signal to the normal CML receiver and use the high-impedance FET probe to tap onto the signal. The signal fidelity may go from good to miserable with the extra impedance even if it is a "small" impedance; nothing's really "small" at Rocket IO rates. Even though your receiver may have troubles getting good data when you're probing, the common mode voltage should be accurate and you would have a "flavor" of the voltage swing.

Thanks for sharing the setup. It should be a pretty quick reconfiguration if you have or can arrange a proper receiver.

- John_H

John,

Thanks for the tip. This is very helpful. I will need to get a receiver with 100 Ohm termination.

I am now assuming that the v5 GTP has 950mV of common mode voltage. In that case my next issue will be how to lower the common mode voltage from 950mv to a voltage between 370mV and 790mV.

Austin mentioned that this can be be controlled using MGTVTTTX plane. This plane is 1.2V. If I lower this voltage then will it not lower the voltage swing? I am not sure that by lowering MGTVTTTX, we can ONLY lower the common mode voltage.

I am still unable to find a document that shows how to lower the common mode voltage.

Thanks.

Eddie

The CML signal is most often used with coding like 8B10B that has a DC-balance to the signals if they were AC. I'd suggest looking at an AC decouple. Signal integrity tools *might* give an idea if using capacitor pass elements or two terminations to your common mode voltage rather than one differential termination will provide good results. Inductors can also be used on the receive side of the capacitors to establish a DC level without splitting the resistor in two. If your swing can be reduced, just using a resistor divider to ground on the receiver may give you the voltage swing you want without compromising the transmitter.

John,

I would love to use AC coupling but we have a training sequence where the transmitter needs to drive a static pattern. The protocol does not seem to support the 8b/10b. The data rate is not that high yet. Austin suggested that the V5 GTPs are flexible in this regard but I do not know how to adjust the common mode voltage only.

Thanks.

Eddie.

So what about the other comment? Simple resistor divider at the receiver?

Using resistor value should work but I was looking if this can be easily done by changing something in the FPGA.

Eddie

Although not of help to the original poster with a non-balanced code, a handy lab parlo{u}r trick from my long-gone-by OC-192 days is to run the serializer output _backwards_ through a broadband bias-T.

Then just hang your scope or multimeter off the bias-T DC input to look at the {low frequency} common mode levels.

( The notion of using a bias-T backwards to AC couple while measuring DC levels is, for me, one of those obvious-after-the-fact things that has come in handy on a number of occasions. )

Brian