Maxim ADC Heat Problem

[racerxdl]

Hi all, I'm new at forum

I work with Eletronics a long time ago, and I have a website about experiments (EnergyLabs Brasil http://www.energylabs.com.br), and I am playing with samples of MAX1421 (40Msps 12BIT ADC from Maxim).


But I have a BIG problem. When I power it up after a few time, I got smoke from the ADC. But now from ADC itself, but smoke from between pins (All times that I saw, it was in the middle of a VDD and GND pin), and if I stay more time, it starts to blink like if there is burning.

First, I think of that I may shorted some pin or so, so I got a new board, new ADC from the sample package and  soldered, measure ALL pins with multimeter in conduction mode to see if there was any short, after that measuring all pins to see if there was a low resistance between any pins (the datasheet says that from any input to GND the resistence is about 10k. All of it was correct. Checked with a Lens too, nothing wrong. After soldered all wires, I triple checked if there was some wiring mistake, and nothing found. So I decided to hook up on FPGA again (Xilinx Spartan 3A) and it was seens to be working (I actually dont know if it was working correct, because I didnt hook up any signal, so I was only picking up noise), but after some time, it stoped to get output and few secs later, smoke too.

Got angry with that, its free samples from Maxim but, I dont want to destroy it. But okay, got to search for problems. After that, I only hooked up the ADC to 3.3V Power Supply, and it got smoke, so I think that ADC is dead.

Asked my dad about it, see what he thinks about (he is eletronics technician too) and he said that might the wires got melt with the solder iron and nearly shorted the pins. Makes sense, so I got a new board, another ADC, soldered on the board and I make only the Power Supply connections, and this time, one wire far from the other (the max that I can do). Triple checked all things again, all correct. Got decoupling caps at the inputs as the ADC Datasheet says, and hooked up ONLY to the 3.3V Power supply. Started, and few secs later, smoke again, and light between VDD and GND Pins (random).

I really dont know what is happening, there is no short, no pins has resistance less than 10K, and the powersupply supplies 3.32V correct (also I used the same supply to a LCD Screen).

The funny thing is that, all the times I measured the current that ADC draws and all stays lower than 1mA. Only hooked at the FPGA with the clock singal (without it is the same, below 1mA) it draws near 20mA (on the datasheet this is lower than usual current, it makes sense, because I only clocking it at 16MHz not 40MHz).

I already sent a email for Maxim IC, but they didnt returned me. Someone knows what can be the problem? Here are few pics of the first board (I know, its crap, but the last one that I Didnt take pic its better, and I was carefull to not make any short or things like that):

The ADC has a PD input (Power Down, that makes the ADC got to shutdown mode). With that on 1 (shutdown state) the current is lesser than 500uA. With that in 0, the current is about 800uA to 1mA. (Of course, without the clock signal, only power supply. Also the PCB is correct, I have more 6 of them here, but only one ADC.

Thanks !

[xygor]

Maybe foreign object debris on the board such as steel wool?

[Psi]

The funny thing is that, all the times I measured the current that ADC draws and all stays lower than 1mA.

If you're only supplying 1mA at 3.3V then that isn't enough to make it smoke. There must be some other voltage coming into the circuit from somewhere else.


Here are a few things you could check..

-Check that there is no AC voltage ontop of the 3.3VDC
-Check that there is no voltages coming into your circuit from other areas. Maybe you have two GND connections at different voltages or something is floating.
-Check that any other IC's you have connected to the ADC are 3.3V i/o.  If they're 5V i/o check they are current limited with a resistor before being connected to adc.


A few other general things to check.

-Check that the IC is soldered onto the PCB around the correct way, maybe vdd and gnd are not where they're supposed to be. The silkscreen pin numbering could be wrong. Or the IC footprint could have got set to 'mirror image' when the board was designed.
-Check you are using analog ground (AGND) and digital ground (DGND) correctly
-Check you have connected all AGND's pins together, its rare but some ICs requires you to do this yourself and don't connect them internally (ive not checked that ic)
-Check you have connected all DGND's pins together, its rare but some ICs requires you to do this yourself and don't connect them internally
-Some IC's require a clock frequency within a specific range and can burn out if there is no clock connected or the clock is oscillating too slowly.
-Check you dont have an adc output connected to VCC/GND.

[Alex]

It sounds like the smoke is coming from the part itself. Smoke from inside an IC can be generated at low power levels and vent from the plastic package.

Check the wiring, pin functions and descriptions against your schematic.
Also check your voltage levels.
From Psi's post, the minimum clock frequency for this part is 100kHz.

[racerxdl]

If you're only supplying 1mA at 3.3V then that isn't enough to make it smoke. There must be some other voltage coming into the circuit from somewhere else.

Yes, that I was think about, but there is ONLY 3.3V going to the board. The supply doesnt make any other voltage.

-Check that the IC is soldered onto the PCB around the correct way, maybe vdd and gnd are not where they're supposed to be. The silkscreen pin numbering could be wrong. Or the IC footprint could have got set to 'mirror image' when the board was designed.
-Check that there is no AC voltage ontop of the 3.3VDC
-Check that there is no voltages coming into your circuit from other areas. Maybe you have two GND connections at different voltages or something is floating.
-Check you are using analog ground (AGND) and digital ground (DGND) correctly
-Check you have connected all AGND's pins together, you shouldn't rely on the IC to do this.
-Check you have connected all DGND's pins together, you shouldn't rely on the IC to do this.

The Board is a universal LQFP48 to PIN adapter from futurlec ( http://www.futurlec.com/SMD_Adapters.shtml ), its all correct, that is the first thing I done when the boards arrived.
The VDD and GND pins are both correct, and all connected to 3.3V Power Supply. Already tested all seeing datasheet specifications
And as I said, the power supply is a switch mode power supply with good decoupling caps and a LCD Screen was tested on it and its working fine without any issues.

-Check that any other IC's you have connected to the ADC are 3.3V i/o.  If they're 5V i/o check they are current limited with a resistor before being connected to adc.

It cant be that, Xilinx Spartan 3A FPGA only works with 3.3V and less.

-Some IC's require a clock frequency within a specific range and can burn out if the clock pin is not oscillating or is oscillating too slow.

Isnt that, the ADC works from 1kHz clock to 40MHz clock, the datasheet says that even with less than 100kHz clock, it may only have output erros.

-Check you dont have an adc output connected to VCC/GND.

I triple checked, nothing is connected as not should be, I review three times, and it was three boards.

Maybe foreign object debris on the board such as steel wool?

Checked about steel wool (I didnt used that, but who knows) and other things that can be on the board, and nothing found. The board is very clean.



Its REALLY strange thing. The only thing that can be, is that maxim datasheet is wrong. My part is MAX1421 ECM and I am using the datasheet from maxim website: http://datasheets.maxim-ic.com/en/ds/MAX1421.pdf. Also I checked the MAX1420 Evaluation Kit (1420, 1421 its pin to pin compatible): http://datasheets.maxim-ic.com/en/ds/MAX1420EVKIT.pdf and all seens wired like mine, so it should work.

[Neilm]

The funny thing is that, all the times I measured the current that ADC draws and all stays lower than 1mA.

If you're only supplying 1mA at 3.3V then that isn't enough to make it smoke. There must be some other voltage coming into the circuit from somewhere else.

You can get the magic smoke from any supply voltage as long as there is enough current behind it.

Looking at the pictures I can't see if there is any decoupling capacitors beside the chip. You have put capacitors on the board but they look like film caps which are not as good as a good seramic at high frequencies. It might be that you are getting glitches on the power rails. These could exceed the 4V maximum that the ADC is allowed on the supply. It might also cause any protection circuit on the supply to latch up - this would present a short circuit to the 3.3V rail and damage the chip. This could also happen if you have sufficient resistance in the current loop from the ADC to the FPGA.

Try getting some ceramic caps and putting them closer to the ADC. Otherwise slow the clock down very slow and see if it survies. Monitor the pins as close to the chip as you can. If it does survive, gradually increase the speed and see if you see any glitches starting to appear. If they do then start to investigate improving the return connection from the FPGA to the ADC.

Neil

[racerxdl]

I am using the power supply from the FPGA, that is very precise (0,01% of noise), but the chip smokes in the shutdown mode, with no clock. Only with the power supply aplied (decoupled all inputs). THAT is strange, the shutdown state is for being out of work, and should not do anything. I am starting to think that is factory problem as all the 4 ADC's that I have is from the same production  (they are numered samples from Maxim). And also, they was stocked (with the original conductive foam and box from Maxim) for about 2 months.

I never had any problem with any chips (even sensitive mosfets), but that ADC is making me have headeach :S

The Maxim answered my first mail and said that I shouldn't have any problems, even without any decoupling caps, Big noise supply, unstable clock, the ADC shouldnt smoke or so. The worsed thing that I might get, is much noise at output or wrong data, not more.

And I said that is strange to smoke at 1mA not about 3.3V.

[allanw]

Maybe be more gentle with the iron. Use leaded solder at 650F with lots of flux so the soldering is basically instant. That way you can rule out heat damage to the pins.

How about only solder a few pins at a time, and apply power. See which pins cause the problem. Try decoupling caps - sometimes they are mandatory.

Like someone else said, it's impossible for the chip to heat up from 3.3V at 1mA. There's something else at play here, like the measurement being faulty. That's the only explanation. Try putting a 100 ohm resistor in series with the voltage rail. Make sure this is in series with all of the power pins, not just one set. Then measure the voltage across that, with a scope. The resistor will also limit the maximum current provided to the device, so you won't break it while trying to figure it out. If it tries to draw too much current, it'll receive less voltage and limit itself. If it needs more current to actually power up in actual operation, start putting smaller resistors instead. Oh and while you're at it, put in some resistors in between every IO, like from your FPGA clock to the ADC clock in. This will limit the current just in case there's some weird voltage level thing going on.

I know what it's like, trying to get a new part to work, especially when prototyping it on some breakout board. It's come to the point where I'm very surprised when something actually does work right after hooking up power for the first time.

Oh and one more thing - never leave any CMOS digital inputs open circuit. The input buffers can draw a damaging amount of current if you put in a voltage around halfway between logic levels. This can happen when you power up the device for the first time, before you even read off the current. The buffer could be destroyed and the chip broken and it'll just draw 1mA of current while it's dead. You'll want to use a digital scope in single shot mode on the resistor I mentioned above to make sure there's no huge current spikes on power-up of a brand new device. I think this is the most likely explanation.

[racerxdl]

Okay, what I did make now:

- Soldered with much solder flux and the solder iron at 650ºF
- Carefully checked each pin
- Decoupling ceramic caps (the best that I have here, some yellow ones) at each VDD to GND (the near GND as possible for VDD)
- Resistor at Differencial Clock Input (as it is LVDS that I am sending, all manufactures recomend a 100R resistor on receiver, so I did that)
- All Outputs with 100R Resistor in series
- Measured the power supply voltage
- Turn on with different resistor sizes in series with the ADC Supply.

Here are what did I built:




The voltage measured at the power supply:


The setup: (Litte wire mess , but I checked and double checked all wire connections to see if its correct)


Also, I saw that I was measuring current with 10A point probe, so the current was unprecise, now I measured the current with the 200mA input (10uA precision, but DMM display cant show that)

All in mA scale:

Current for 100R Series resistor:


Current for 47R Series resistor:


Current for 22R Series resistor:



The circuit didnt smoke, but it didnt work. I am testing with a not good way, but I can know if its working or not. I am getting the output from LCD and putting at the pixels of the LCD screen. As the LCD clock is higher than ADC Clock, I will get more than 1 pixel with the same data.


It wont show anything usefull, but I can know if its a noise or if it isnt working. Other times that I tested the ADC, I got noise for a few time, and after that time, the screen got with lesser noise, and vanished. Only black screen. My LCD Controller is working, as is the same of my Virtual Terminal Project.

It didnt smoke, but it didnt show anything, only black screen, but I am afraid that the ADC is consuming more than he was built for. The datasheet says about 65mA of Analog Supply and 5mA from Digital Supply, that running at 40MHz (I am running at 16), so the current must be arround 70mA, but with a 22R resistor I got 100mA. The current increases a little when clock is supplied (after programing FPGA), but only 2mA

Any ideas?

[Psi]

This is ridiculously unlikely but possible, when you ordered those chips you didn't get them posted to a US government department or diplomatic office did you? (or any location to do with the government)

I ask because the united states postal service put all mail to government departments through a process which irradiated the mail to kill any biological contents, eg Anthrax. I assume they have a separate mail system in place for items they don't want irradiated but the bulk of the printed material gets irradiated.

The levels of radiation used will also kill seeds, photographic film, biological samples, food, medicines, and electronic equipment.
Unlike an modern airport xray machine ,which is safe for such things, the mail irradiation is quite a few orders of magnitude more powerful.

If you just so happened to get those IC's delivered to a building which gets their mail irradiated they would be DOA.

[allanw]

It didnt smoke, but it didnt show anything, only black screen, but I am afraid that the ADC is consuming more than he was built for. The datasheet says about 65mA of Analog Supply and 5mA from Digital Supply, that running at 40MHz (I am running at 16), so the current must be arround 70mA, but with a 22R resistor I got 100mA. The current increases a little when clock is supplied (after programing FPGA), but only 2mA

Good, at least now you can debug without frying the chip everytime. Measure the voltage at the other end of the resistor (what the ADC is receiving), because it'll be much lower than 3.3V. 100mA dropped over 22R = 2.2V drop, so you should be measuring 1.1V at the ADC power inputs. That means the device is drawing 100mA at only supply voltage of 1.1V. So there is definitely something wrong because most of the chip shouldn't power up at that voltage. There's some kind of low resistance path. Have you checked with an ohmmeter between VDD and GND?

edit: This is a good reason why you should power up new untested circuits with a current-limited power supply, instead of hooking it up to something that can blast a few amps through it and destroy it. If you had a power supply with a current limit at 100mA, it'd basically be the same as having that 22R resistor there - to protect the device.

[Psi]

Remove any data lines and try to power up the chip alone with the bare minimum of connections ,  eg power, gnd and clk  Keep the series resistor in place and see if the current changes to something normal. If so, then one of your other connections is causing the problem.

Also check your differential clock, it may have a 3.3v clk on it but, being differential the potential voltage to gnd might be much higher which would force both CLK and not CLK to high and might draw lots of current.

Also might be a good idea to try a different clock in single ended mode if you have one.

[racerxdl]

Yeah I did tryed powering up only with supply, but the current is same.

Now I see what happens, when I turn on, it starts with the normal current, like 56mA (or 15mA without clock) after sometime the current rises for 100mA (not more because the series resistor). That happens independently of what I do, with data lines connected/disconnected or/and clock disconnected/connected or something so.

I think I maybe out of luck and my ADC's got problems. I asked that for maxim, they said that can happen. I will try to ask for more samples to test it, Maxim already authorized that.

These samples was stocked here for a long time, and sometimes it has opened by some people, maybe someone has discharged static in there or so, improbally (because the conductive foam), but who knows.

Let we see when new samples arrive.

Thanks all for the help!

[bilko]

Have you checked that your soldering iron is earthed ?