Low slewrate, abnormal current consumption.

K

karl.bengtsson

Guest
Hello.

I have a design targeting the (ancient, perhaps) Actel A54SX32A. Durin
testing I exposed several of the input pins to signals with extremly lo
slewrate, using a power-supply to supply the input. This was most likel
way outside the minimum slewrate requrements for the device.

The FPGA seems to be working and it passes all functional tests, but th
current to the core is way to large at about 70mA. Sadly, I did not measur
the current beforehand and the cost of a new (ceramic) FPGA prohibits m
from just burning a new one to test that. A freshly burned plastic FPG
works fine and consumes the expected current.

Does anyone have any experience with destroying FPGAs due to low slewrates
Is this even probable?

Also, Hello everyone. This is my first post here, Ive been "lurking" so t
speak, and there seems to be a lot of talanted people here.
Regards!
Kalle



---------------------------------------
Posted through http://www.FPGARelated.com
 
On Fri, 07 Jan 2011 07:24:21 -0600, "karl.bengtsson"
<karl.bengtsson@n_o_s_p_a_m.norspace.no> wrote:

Hello.

I have a design targeting the (ancient, perhaps) Actel A54SX32A. During
testing I exposed several of the input pins to signals with extremly low
slewrate, using a power-supply to supply the input. This was most likely
way outside the minimum slewrate requrements for the device.

The FPGA seems to be working and it passes all functional tests, but the
current to the core is way to large at about 70mA. Sadly, I did not measure
the current beforehand and the cost of a new (ceramic) FPGA prohibits me
from just burning a new one to test that. A freshly burned plastic FPGA
works fine and consumes the expected current.

Does anyone have any experience with destroying FPGAs due to low slewrates?
Is this even probable?
Click to expand...
CMOS structures are push-pull ie they have two switches which
"simultanenously" turn-on/off to change state; alas the switching is
never instantenous so there is overlap so dynamic current flows during
the switch. The worst case current is near the middle voltage level,
the longer you stay around that point the higher the average current.
It's possible to destroy individual IOs by leaving unconnected inputs
and letting them float. Your case is not as extreme but would affect
lifetime of the IO.
--
Muzaffer Kal

DSPIA INC.
ASIC/FPGA Design Services

http://www.dspia.com
 
On Jan 8, 5:50 am, Muzaffer Kal <k...@dspia.com> wrote:
On Fri, 07 Jan 2011 07:24:21 -0600, "karl.bengtsson"



karl.bengtsson@n_o_s_p_a_m.norspace.no> wrote:
Hello.

I have a design targeting the (ancient, perhaps) Actel A54SX32A. During
testing I exposed several of the input pins to signals with extremly low
slewrate, using a power-supply to supply the input. This was most likely
way outside the minimum slewrate requrements for the device.

The FPGA seems to be working and it passes all functional tests, but the
current to the core is way to large at about 70mA. Sadly, I did not measure
the current beforehand and the cost of a new (ceramic) FPGA prohibits me
from just burning a new one to test that. A freshly burned plastic FPGA
works fine and consumes the expected current.

Does anyone have any experience with destroying FPGAs due to low slewrates?
Is this even probable?

CMOS structures are push-pull ie they have two switches which
"simultanenously" turn-on/off to change state; alas the switching is
never instantenous so there is overlap so dynamic current flows during
the switch. The worst case current is near the middle voltage level,
the longer you stay around that point the higher the average current.
It's possible to destroy individual IOs by leaving unconnected inputs
and letting them float. Your case is not as extreme but would affect
lifetime of the IO.
Click to expand...
Which part numbers have you actually seen this occur on ?

Your explanation is correct, that there is a Icc peak (or double Icc
peak on a Schmitt pin), but that is usually in the sub mA region, and
so will struggle to generate more than 1mW of heating in each FET.

eg look at NXP 74AUP2T1326, which gives a classic schmitt Icc plot,
twin peaks at 150uA & 200uA, 1.8V testcase.

A NXP 74LVU04 shows over 200uA @ 1.2V and > 1mA @3V, but that if for
Output Drive fets, (~ 40 ohms)

Another Digital-Pin effect, is not the current of a linear buffer,
but the parasitic transition oscillations that can occur, when there
is no hysteresis, and a multiple buffers hit the 'linear' region.

I have seen the chip-wide effects of that, yield high Icc levels,
and also give 'crosstalk' effects, where supposedly logically
independent pathways, jitter as a result of this. (common mode
inductance).
This would (over?) heat a larger portion of the chip.

So the advice is don't float (non schmitt) pins, but has anyone seen
physical/permanent damage/degrade from these effects ?

-jg
 
Which part numbers have you actually seen this occur on ?
Click to expand...
My device is an A54SX32SA-208CQFP. A freshly burned A54SX32SA-208PQFP show
normal current consumption.

I understand how leaving the voltage at an incorrect level will caus
current to flow in the input, which could cause excessive heating.

But how would this translate into an permanent increase in current on th
core supply?

Some additional findings. I desoldered (well, someone did it for me) th
VCCA pins of the FPGA to see if there where any difference between them.
then connected them one-by-one and looked at the current delivered. A fe
of the pins showed a stable, high, current consumption. While a couple o
others varied wildly for a few seconds before settling to the same, high
current consumption.
I have no idea if this tells anyone anything, but I felt it was wort
noting.

Anyhow, thanks everyone for the replies so far! :)




---------------------------------------
Posted through http://www.FPGARelated.com
 
On Jan 10, 8:53 pm, "karl.bengtsson"
But how would this translate into an permanent increase in current on the
core supply?

Some additional findings. I desoldered (well, someone did it for me) the
VCCA pins of the FPGA to see if there where any difference between them. I
then connected them one-by-one and looked at the current delivered. A few
of the pins showed a stable, high, current consumption. While a couple of
others varied wildly for a few seconds before settling to the same, high,
current consumption.
Click to expand...
So you are saying the device(s?) work ok, but just have higher
currents ?
Are you also saying that higher Icc, is evenly spread over Banks, even
with no floating pins and a static clock ?

-jg
 
On Jan 10, 8:53 am, "karl.bengtsson"
<karl.bengtsson@n_o_s_p_a_m.n_o_s_p_a_m.norspace.no> wrote:
Which part numbers have you actually seen this occur on ?

My device is an A54SX32SA-208CQFP. A freshly burned A54SX32SA-208PQFP shows
normal current consumption.

I understand how leaving the voltage at an incorrect level will cause
current to flow in the input, which could cause excessive heating.

But how would this translate into an permanent increase in current on the
core supply?

Some additional findings. I desoldered (well, someone did it for me) the
VCCA pins of the FPGA to see if there where any difference between them. I
then connected them one-by-one and looked at the current delivered. A few
of the pins showed a stable, high, current consumption. While a couple of
others varied wildly for a few seconds before settling to the same, high,
current consumption.
I have no idea if this tells anyone anything, but I felt it was worth
noting.

Anyhow, thanks everyone for the replies so far! :)

---------------------------------------        
Posted throughhttp://www.FPGARelated.com
Click to expand...
It looks like that part was irradiated by particles during a test
beam.
This is a OTP FPGA. That Actel's family is used in the satellites/
rockets/airplane as also in particle physics experiments and usually
are submitted to radiation test.
After radiation, one major TID effect is an augment of leakage current
in CMOS that is what you observe.

So the question are:
where this Ic come from ?
Could it been irradiated ?

It is only a hypothesis but it is better to have care in managing that
part.

Fabio
 
On 10 Jan., 11:47, F M <fmon...@gmail.com> wrote:
On Jan 10, 8:53 am, "karl.bengtsson"



karl.bengtsson@n_o_s_p_a_m.n_o_s_p_a_m.norspace.no> wrote:
Which part numbers have you actually seen this occur on ?

My device is an A54SX32SA-208CQFP. A freshly burned A54SX32SA-208PQFP shows
normal current consumption.

I understand how leaving the voltage at an incorrect level will cause
current to flow in the input, which could cause excessive heating.

But how would this translate into an permanent increase in current on the
core supply?

Some additional findings. I desoldered (well, someone did it for me) the
VCCA pins of the FPGA to see if there where any difference between them.. I
then connected them one-by-one and looked at the current delivered. A few
of the pins showed a stable, high, current consumption. While a couple of
others varied wildly for a few seconds before settling to the same, high,
current consumption.
I have no idea if this tells anyone anything, but I felt it was worth
noting.

Anyhow, thanks everyone for the replies so far! :)

---------------------------------------        
Posted throughhttp://www.FPGARelated.com

It looks like that part was irradiated by particles during a test
beam.
This is a OTP FPGA. That Actel's family is used in the satellites/
rockets/airplane as also in particle physics experiments  and usually
are submitted to radiation test.
After radiation, one major TID effect is an augment of leakage current
in CMOS that is what you observe.

So the question are:
where this Ic come from ?
Could it been irradiated ?

It is only a hypothesis but it is better to have care in managing that
part.
Click to expand...
The A54xxx is the commercial device without radiation hardening. Would
not expect any irradiation of this device expecially not in device
screening process, as rad tests are not part of device screening only
qualification (means the design is irradiated, not the delivered
device). Screening should only cover a burn-in for each device (normal
for a RT54 device, don't know if some mil flow covers burn-in).

For the RT54SXxxS devices from MEC there was an issue with inrush
during power cycling. Don't know if commercial device could have the
same issue as it is different silicon with AFAIK different process
than RT device.

bye Thomas
 
On Jan 10, 5:39 pm, Thomas Stanka <usenet_nospam_va...@stanka-web.de>
wrote:
On 10 Jan., 11:47, F M <fmon...@gmail.com> wrote:









On Jan 10, 8:53 am, "karl.bengtsson"

karl.bengtsson@n_o_s_p_a_m.n_o_s_p_a_m.norspace.no> wrote:
Which part numbers have you actually seen this occur on ?

My device is an A54SX32SA-208CQFP. A freshly burned A54SX32SA-208PQFP shows
normal current consumption.

I understand how leaving the voltage at an incorrect level will cause
current to flow in the input, which could cause excessive heating.

But how would this translate into an permanent increase in current on the
core supply?

Some additional findings. I desoldered (well, someone did it for me) the
VCCA pins of the FPGA to see if there where any difference between them. I
then connected them one-by-one and looked at the current delivered. A few
of the pins showed a stable, high, current consumption. While a couple of
others varied wildly for a few seconds before settling to the same, high,
current consumption.
I have no idea if this tells anyone anything, but I felt it was worth
noting.

Anyhow, thanks everyone for the replies so far! :)

---------------------------------------        
Posted throughhttp://www.FPGARelated.com

It looks like that part was irradiated by particles during a test
beam.
This is a OTP FPGA. That Actel's family is used in the satellites/
rockets/airplane as also in particle physics experiments  and usually
are submitted to radiation test.
After radiation, one major TID effect is an augment of leakage current
in CMOS that is what you observe.

So the question are:
where this Ic come from ?
Could it been irradiated ?

It is only a hypothesis but it is better to have care in managing that
part.

The A54xxx is the commercial device without radiation hardening. Would
not expect any irradiation of this device expecially not in device
screening process, as rad tests are not part of device screening only
qualification (means the design is irradiated, not the delivered
device). Screening should only cover a burn-in for each device (normal
for a RT54 device, don't know if some mil flow covers burn-in).

For the RT54SXxxS devices from MEC there was an issue with inrush
during power cycling. Don't know if commercial device could have the
same issue as it is different silicon with AFAIK different process
than RT device.

bye Thomas
Click to expand...
I think Karl knows very well that A54SX32A is commercial - not RT -
part.
In any case, A54SX are currently used in many radiation environments.
Just try to google with "A54SX32A radiation".
They are a compromise between robustness against radiation and costs.
When you need thousand of parts it is really costly to purchase RT
version. This is common in application like electronics for particle
physics. I'm used to follow that way with flash A3P commercial (not
RT) family. In this case, the RTA3P's silicon is exactly the same as
the commercial A3P ones. Anyway I don't thinks this applies also to
A54SX vs. RT54SX. Ususally RT OTP flash are tripled (TRM).

Fully agree with the sentence as A54SX32A are OTP part.

as rad tests are not part of device screening only
qualification (means the design is irradiated, not the delivered
device).
Click to expand...
but pls be noticed that Karl does not explain the application on
A54SX32A nor it is fully clear if he was really the engineer which
programmed the part.
Could have got it already programmed from someone else ?
In this case, could a radiation test be done ?

I currently work with A3P commercial parts which was irradiated time
ago and now I'm making measurements on the leakage currents in order
to understand 'end of life' behavior. This is fine with me as I know
what I have in my hands.

As I wrote, mine was only an hypothesis. The suggestion is to trace
back from where that part comes.
I'm convinced that this is really a far - not probable - possibility
but errors and mistakes could have happened.

It is up to Karl understand if this could be possible or not.
It depends from where he works, who is his supplier and so on ...

Anyway I'm sure that many other reasons could explain the part's
behavior.

Cheers, Fabio
 
Hello again.
The device has not gone trough any in-house irradiation. I do not believ
radiation caused this.

To me, it seems that I ruined the device by providing input way out o
spec. Since I see no problems in newly programmed parts (which differ i
encapsulation only), I don´t think I have a problem with my VHDL design.

I´ll post an update when this design is burned to an RT-part.

Thank you all for your insight.
/Karl

I think Karl knows very well that A54SX32A is commercial - not RT -
part.
In any case, A54SX are currently used in many radiation environments.
Just try to google with "A54SX32A radiation".
They are a compromise between robustness against radiation and costs.
When you need thousand of parts it is really costly to purchase RT
version. This is common in application like electronics for particle
physics. I'm used to follow that way with flash A3P commercial (not
RT) family. In this case, the RTA3P's silicon is exactly the same as
the commercial A3P ones. Anyway I don't thinks this applies also to
A54SX vs. RT54SX. Ususally RT OTP flash are tripled (TRM).

Fully agree with the sentence as A54SX32A are OTP part.

as rad tests are not part of device screening only
qualification (means the design is irradiated, not the delivered
device).

but pls be noticed that Karl does not explain the application on
A54SX32A nor it is fully clear if he was really the engineer which
programmed the part.
Could have got it already programmed from someone else ?
In this case, could a radiation test be done ?

I currently work with A3P commercial parts which was irradiated time
ago and now I'm making measurements on the leakage currents in order
to understand 'end of life' behavior. This is fine with me as I know
what I have in my hands.

As I wrote, mine was only an hypothesis. The suggestion is to trace
back from where that part comes.
I'm convinced that this is really a far - not probable - possibility
but errors and mistakes could have happened.

It is up to Karl understand if this could be possible or not.
It depends from where he works, who is his supplier and so on ...

Anyway I'm sure that many other reasons could explain the part's
behavior.

Cheers, Fabio
Click to expand...
---------------------------------------
Posted through http://www.FPGARelated.com
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top