Virtex-4 inrush power-on current

I am replacing a V4FX20 chip with a V4FX40 on a board. I am a little bit concerned with the power-on requirements of the bigger chip. The existing core voltage power supply is designed for 1A. The datasheet says that Iccintmin is 244 mA typically, but can be as big as 1.65A under worst process, voltage and temperature conditions. It also says that the actual current consumed depends on the power-on ramp rate of the power supply. I wonder if there is some more precise information on how it actually depends on it?... Also, the board is never going to see worst ambient temperature or voltage conditions. Am I safe with 1A of continious current? Unfortunately, replacing the regulator is not easy due to very limited space and some other issues...

Thanks, /Mikhail

Reply to
MM
Loading thread data ...

Mikhail,

Per the data sheet, the worst case possible Iccint is 1.65 amperes, yes.

If you do not have this available, you risk not powering on, and not configuring.

Now, that said, that specification if for a 85C or 100C junction temperature (C or I grade), highest voltage, fastest silicon.

It is up to you to gage the risk. If the ambient is never that hot, then the max current is a lot less. If the Vccint is not 5% high, the max current is less. If the process is closer to typical, the current is a lot less.

The FX20 had a max of 1.1 amperes, so you were violating that spec (with only 1.0 amperes) before, and you seemed to be OK with that...

Perhaps if you test every pcb prior to ship at hot, you could by testing, guarantee the power on?

Aust> I am replacing a V4FX20 chip with a V4FX40 on a board. I am a little bit

Reply to
austin

How much is a 'lot'?

Current is measured in units of 'less'?

Another vacant 'lot'?

I think you failed in your attempt to answer the OP's basic question which was

Vague generalities as a response to a question about needing more specifics to verify the design aren't usually too helpful.

Maybe you'd like another crack at it?

KJ

Reply to
KJ

K,

I would imagine that Xilinx, or any other company, will not get any more specific on specs that are being violated, because of liability issues. Austin has given "rules of thumb" that may make one sleep more easily with this under-designed product. If it were my product, I would (re)design it to meet the specification.

Bob

--
== NOTE: I automatically delete all Google Group posts due to uncontrolled 
SPAM ==
Reply to
BobW

KJ,

I am trying to politely say, he is on his on, he has exceeded the spec.

I am also saying that there are those who might test their systems in a burn-in oven, and reject the units that won't power on when hot.

If he has 10 units he has to ship next week, this might be a perfectly good temporary solution, until he re-designs his power supply.

Austin

Reply to
austin

Bob, Austin,

The request was "if there is some more precise information on how it actually depends". If Xilinx doesn't have (or doesn't want to disclose even under NDA) then the answer to the question is quite simply 'no'.

Whether Austin is representing Xilinx or not on this newsgroup he certainly is privy to more detailed info on the parts as are all of the Xilinx engineering reps that support Xilinx products. As such, those people are in a position to provide some form of engineering assisstance to customers (since that is pretty much the job description of such a rep). Whether Xilinx is willing to share such info (if it exists) is a business decision that Austin and other reps would also know....if not, the answer to the request can simply be 'no'.

The answer that Austin gave was the type of lame response I might expect from someone who doesn't know much about the particular parts. Since Austin is clearly not in that class, his answer came across as pretty lame, so I thought pointing out that he might want to take a second go at it or simply say that no more detailed info can be provided would clarify the issue a bit. If more useful and precise info could be provided in an open forum, then he could have provided it. Since it likely can not, the reply could have been taken off group or Mikhail could have been referred to a local engineering rep for assisstance or simply told 'no' and maybe a statement it's possible at any time and temp that a particular part might draw the max inrush current therefore you have to redesign your supply.

In any case, don't interpret this as a slam on Austin or Xilinx, just an attempt to elicit a clearer response to whether the info that Mikhail requested is available or not.

Kevin Jennings

Reply to
KJ

Austin,

Just curious: why even spec "Iccintmin is 244 mA typically" if Xilinx really wants their users to design for 1.65A on power-up/configuration?

I haven't read the data sheet on this part but the OP makes it sound like the data sheet suggests a lower power-up requirement if the design is within the nominal limits on voltage, temperature, etc...

Rob

aust> KJ,

Reply to
Rob

Reply to
Peter Alfke

Give us a break! This particular parameter has wide variations. It is higher at high temperature and it varies widely (much more widely than most other parameters) with processing tolerances. We sell hundreds of thousands of parts every day, to thousands of customers with thousands of different board designs. We want to make sure that every user-design works all the time. That's why we test exhaustively, and we guardband the test results. That makes some of them look excessive, and they often are. But that is the price for guaranteed performance over a wide range of ambient conditions and the expected processing variations.

You can probably violate the max inrush current spec, if the part is always cool when it is being turned on. But one day one user might turn power back on when the part is still hot, and the configuration might then not work, and we get ugly complaints... There is no way for us to predict what hundred thousand users might occasionally do with our parts. That's why we do worst-case testing and guardband our specifications. We prefer satisfied customers...

Peter Alfke

Reply to
Peter Alfke

Peter,

I wasn't trying to give Xilinx a hard time. If I was, I would have been more pointed in my remarks. I think I started off my email by saying, "Just curious"; which I thought was a polite and non-aggressive choice of words. Remember, Austin said the OP is on his own since he violated the spec. My curiosity was peaked because the OP's interpretation of the data sheet suggested that the spec was between 244mA and 1.65A; and depending on other operating conditions his 1A might be OK. I believe this is what the OP was inquiring about inf the first place.

I don't need to be patronized. I am very well aware of how successful Xilinx is, so you don't have to tell me that they sell 100's of 1000's of parts a day.

So the answer is, since this parameter varies widely Xilinx doesn't feel comfortable recommending anything lower than the 1.65A. This is a simple answer that would have appeased everyone from the beginning.

Take care, Rob

Peter Alfke wrote:

Reply to
Rob

Rob,

Line 1 of my first reply.

One last comment, there is no 'in-rush current', the current is what is needed, and does not vary while powering on. 1.65 amperes is primarily static leakage at the highest junction temperature, fastest silicon process, and highest Vccint. It is NOT something special, or unusual, required ONLY during power on. After configuration, if anything, the current will be more as the IOs and clock trees and the fabric begins to operate.

Austin

Reply to
austin

Austin,

I think there will be a flame of responses to what you have just said! If that's the case, this current has to be clearly specified as maximum static leakage and NOT as a power-up requirement. I think there are lots of boards out there, which violate this spec, because they were designed for this as the maximum total current at best, based on other data sheet values, power calculator reports, and actual measurements. What's the point of using your power tools if they report static current value, which is 2-5 times less than the spec?! I've just run the XPower Estimator off of your web site for XC4VFX40, FF1152, commercial, maximum process, 1.26V core voltage, 78C ambient temperature with no air flow (which gives 84.6 C junction temperature). Under these conditions the reported Iccq is 0.723 A. Even at

100 C junction for the industrial part the projected current is only 0.895 A

/Mikhail

Reply to
MM

MM,

Sorry, those are the facts.

The 'estimator' is just that: an estimate.

The minimum power on current required is just that, the minimum required so that 100% of the parts power on under all possible conditions.

Is there a "surge"? According to common definitions:

formatting link

No, there is no "surge."

The data sheet is clear. I am trying to be just as clear, and just as precise.

You bring up a good point: the estimator should never return a value that is less than the minimum required for power on (at a given temperature and voltage and process). Not sure how to do this, but I will put in on the list things to improve.

Austin

Reply to
austin

Sorry Austin, but it is ABSOLUTELY not. The table 4 of the datasheet says that Iccintq is typically 139 mA for the particular part we've been discussing. The note 6 to that table is saying to "Use the XPower Estimator (XPE) tool to calculate maximum static power for specific process, voltage, and temperature conditions." Now, the Power-On Power Supply Requirements chapter says that "Xilinx® FPGAs require a certain amount of supply current during power-on to insure proper device initialization. The actual current consumed depends on the power-on ramp rate of the power supply." It doesn't say anything about how it depends on the power-on ramp however. Now, on the top of all this you are saying that the Iccintmin number from table 5 is the real Iccintq, and not the one given in table 4 or reported by the power estimator.

I can't agree that this is clear by any means.

/Mikhail

Reply to
MM

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.