UPGRADE YOUR BROWSER

We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

cancel
Showing results for 
Search instead for 
Did you mean: 
Visitor rkriner
Visitor
551 Views
Registered: ‎12-05-2018

Kintex-7 MGTAVTT Power-Up Leakage / Glitch Behavior

Jump to solution

We have a Kintex-7 (410T 900) and are observing an anomaly on the MGT 1.2V (MGTAVTT) rail at power-up.  The attached image shows the waveform.

We have confirmed we are following the Xilinx-recommended power-up sequencing (VCCINT => VCCAUX => VCCO; VCCINT => MGT 1V => MGT 1.2V).

There are 2 concerns here.  First, the MGT 1.2V rail appears to start coming up before it should.  We suspect this may be leakage.  Second, and perhaps of greater concern, the rail is (technically) non-monotonic because the rail charges a little from the leakage, then when the 1.2V regulator turns on, the rail dips back to ~0V very briefly, then it ramps up to the regulated value.  Xilinx is pretty clear about avoiding non-monotonic power rails.  However, since the rail only charges up to less than 200mV (sample size 1, room temperature) before it dips back to ~0V, that may not be enough to cause concern.

We also have waveforms from a Zynq design where the MGT 1.2V rail behaves similarly—i.e. it shows the leakage behavior before the regulator turns on (and to a higher amplitude).  But on the Zynq design, we didn’t see the voltage dip back to ~0V when the regulator turned on.  On the Zynq, the regulator turns on and just ramps the voltage up from where the leakage behavior had charged it.  In other words, there wasn’t non-monotonic behavior on the Zynq design.

Q1. Can Xilinx confirm the FPGA and/or 100 Ohm RCAL resistor is the path of the “leakage” from the MGT 1V rail (since the MGT 1.2V rail leakage seems to come up along with the MGT 1V rail)?

Q2. Not shown in the figure is the fact that our 1.8V VCCAUX (not the MGT 1.8V, but the fabric VCCAUX) also turns on around the same time we see the MGT 1.2V rail exhibiting the suspected leakage behavior.  Is it possible the MGT 1.2V rail is seeing leakage through the FPGA from the VCCAUX rail?

Q3. Could any other FPGA rail be leaking back onto the MGT 1.2V rail through the FPGA?

Q4. Is the small, non-monotonic behavior seen on the MGT 1.2V rail when it turns on any reason for concern from a device functionality perspective (i.e. could it put the device at risk to wake up in a non-functional state)? 

Q5. Is the small, non-monotonic behavior seen on the MGT 1.2V rail when it turns on any reason for concern from a device reliability perspective?

Q6. Is this a known anomaly?  

Please advise

Thanks 

MGT 1p2V Anomaly.jpg
0 Kudos
1 Solution

Accepted Solutions
Moderator
Moderator
459 Views
Registered: ‎09-18-2014

Re: Kintex-7 MGTAVTT Power-Up Leakage / Glitch Behavior

Jump to solution

Rkriner,

 

Hi, I am not sure if you have seen AR59919 yet but to directly address your concerns about the small biasing between MGTAVCC and MGTAVTT this has been seen on 7-series GTX/GTP before and is expected to some degree. The transceiver circuit/block spans different transceiver power domains thus it is inevitable that there exists small leakage between the two rails. The leakage here is small enough that this should not pose any reliability issues and degree of biasing small enough there should be no functionality effects. VCCAUX does not supply any transceiver circuitry. Transceivers have their own dedicated MGTVCCAUX so biasing from VCCAUX is highly unlikely within the FPGA. Only shared fabric-transceiver rail is VCCINT. 

 

Regards,

T

-------------------------------------------------------------------------
Don’t forget to reply, kudo, and accept as solution.
-------------------------------------------------------------------------
3 Replies
Highlighted
Voyager
Voyager
502 Views
Registered: ‎02-01-2013

Re: Kintex-7 MGTAVTT Power-Up Leakage / Glitch Behavior

Jump to solution

Wow. That's a really big picture. They let you upload that? Must be all that whitespace... Ever since they put a limit on picture size, I keep having to trim uploads...

Anyway... I can't and don't speak for Xilinx, but I don't see anything wrong with your power rails. Other than your consternation due to the unanticipated leakage, is the system working okay?

The leakage looks to cause the VMGTAVTT rail to climb to ~10% of its final value. I'd be willing to bet the device doesn't notice that at all. The 'monotonic' rule is meant to enforce a single power-on event, and prevent power the rail from going above, then below, then above (etc.) a threshold where things would turn on, then off, then on again (etc.).  I don't think anything is turning on at a 10% level.

The dip before the final rise is due to the turning-on of the regulator that ordinarily drives that rail. Its output wants to start at 0V, so that's where the weak-leak-filled rail gets driven--to precisely where the regulator wants it to be. If you want to avoid something like that, you'd need another regulator--one that supports pre-charging of a rail. 

Whatever is turning on at the time represented by the center graticule of the scopeshot is the likely cause of the leakage. (I think you mentioned that's a 1.8V rail.) The important thing I would look for here is: does VMGTAVCC turn on and become stable before VMGTAVTT turns on? I see that happening, so that's good.

What I might be concerned over is the dip in the yellow (3.3V?) supply rail. That looks to be a ~300 mV dip--nearly -10%! Make sure that anything else using that rail is not impacted by that.

-Joe G.

 

 

0 Kudos
Moderator
Moderator
460 Views
Registered: ‎09-18-2014

Re: Kintex-7 MGTAVTT Power-Up Leakage / Glitch Behavior

Jump to solution

Rkriner,

 

Hi, I am not sure if you have seen AR59919 yet but to directly address your concerns about the small biasing between MGTAVCC and MGTAVTT this has been seen on 7-series GTX/GTP before and is expected to some degree. The transceiver circuit/block spans different transceiver power domains thus it is inevitable that there exists small leakage between the two rails. The leakage here is small enough that this should not pose any reliability issues and degree of biasing small enough there should be no functionality effects. VCCAUX does not supply any transceiver circuitry. Transceivers have their own dedicated MGTVCCAUX so biasing from VCCAUX is highly unlikely within the FPGA. Only shared fabric-transceiver rail is VCCINT. 

 

Regards,

T

-------------------------------------------------------------------------
Don’t forget to reply, kudo, and accept as solution.
-------------------------------------------------------------------------
Visitor rkriner
Visitor
437 Views
Registered: ‎12-05-2018

Re: Kintex-7 MGTAVTT Power-Up Leakage / Glitch Behavior

Jump to solution

I was unaware of that Answer Record; thanks for pointing it out.

The Answer Record, in concert with your response confirms the behavior is expected and/or acceptable from a device reliability and functionality perspective (including the brief non-monotonic event on the MGT 1.2V rail when it turns on).

Thanks for the helpful response.

Bob K

0 Kudos