cancel
Showing results for
Show  only  | Search instead for
Did you mean:
Highlighted
Visitor
1,240 Views
Registered: ‎08-30-2016

## GTX Transceiver Margin Analysis, Verctical voltage range

Hello,

I'm working with a Kintex-7 device and try to run a Margin Analysis with the GTX Transceiver. The scan provides a vertical offset setting, which changes the voltage threshold between -127 and 127 voltage codes.

I couldn't find a place where the value of these voltage codes are defined. I'd like to get a value in mV corresponding to a single voltage code, to estimate the transmission margin in mV.

Is there a way to find out? I found that the differential input voltage swing depends on the receiver settings, so the mV per voltage code will be different depending on the receiver setup. Is there anything else to take into account? Would it be save to assume, that if My maximum voltage swing is 2000mV (Vdpp), That a single code would match 2000/254 = 7.9 mV?

The only resource I found was an old forum entry which stated 2mV per voltage code, but with no reference to any documentation.

1 Solution

Accepted Solutions
Highlighted
Moderator
1,565 Views
Registered: ‎07-30-2007

After the input signal goes through equalization which will vary the amplitude based on the frequency and has an Automatic Gain Control stage it may not make sense to talk in terms of a voltage conversion (and how could you measure it anyway).

That said, the conversion we use is 1.89 mv/code.

We generally look for 45 codes (85mv) in the vertical to be satisfied with the margin assuming you are meeting the specs and the guidelines for noise on the power supplies and the jitter on the refclk.

Edited for completeness:  We look for .35 ui in the horizontal.  Both of these conditions are for a BER on the scan of 10 to the -12 and assume you're meeting all datasheet specs.

----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
----------------------------------------------------------------------------

5 Replies
Highlighted
Teacher
1,227 Views
Registered: ‎07-09-2009

I think,

the output voltage and the reading you have are not calibrated.

The +- 127 I think is a ratio metric measurement,

Highlighted
Moderator
1,566 Views
Registered: ‎07-30-2007

After the input signal goes through equalization which will vary the amplitude based on the frequency and has an Automatic Gain Control stage it may not make sense to talk in terms of a voltage conversion (and how could you measure it anyway).

That said, the conversion we use is 1.89 mv/code.

We generally look for 45 codes (85mv) in the vertical to be satisfied with the margin assuming you are meeting the specs and the guidelines for noise on the power supplies and the jitter on the refclk.

Edited for completeness:  We look for .35 ui in the horizontal.  Both of these conditions are for a BER on the scan of 10 to the -12 and assume you're meeting all datasheet specs.

----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
----------------------------------------------------------------------------

Highlighted
Visitor
1,164 Views
Registered: ‎08-30-2016

Thanks, that is what I was looking for. If we have a approximate reference point to check we are happy, as long as we have some guidelines.

Highlighted
Visitor
1,139 Views
Registered: ‎08-30-2016

A quick follow up: You state you look for 45 voltage codes to be satisfied with a link. Do you have similar guidelines for the horizontal range? We're looking for a pass/fail mask for a input signal, but as you already stated, The measurement is after Clock correction, so one cannot directly apply margins for a signal standard, as this is already the "improved" signal.

Thanks

Highlighted
Moderator
1,134 Views
Registered: ‎07-30-2007

We look for .35 ui in the horizontal.  Both of these conditions are for a BER on the scan of 10 to the -12.

----------------------------------------------------------------------------
Don't forget to reply, kudo, and accept as solution
----------------------------------------------------------------------------