[solved] sn74lvc1g3157 analog switchs keep burning. Any hint ?

[Jaunedeau]

Hi,

  I try using an sn74lvc1g3157 analog switch by TI (datasheet), but burnt two already.

  The package is SC70 (all I could find without paying way to much for shipping). So first thing I did was deadbug soldering to a pinheader :

  The photo was just taken to check marking, I cleaned the IC before using it.

  First thing : the datasheet says I should find a dot, but I did not. There is a marking saying |C5J. I could verify that this is the chip I ordered, and guessed that vertical bar is on the side of pin 1 - so if I made no mistakes, pins on the from left to right pin of the pin header are 1..6, but am not 100% sure.

  Second thing, I plugged this on a breadboard, burnt it (it started to heat then pin 5 and 6 reads 6 ohm). soldered another chip and redid everything from sratch, called an electronic professor on youtube to thow him what I did, it took time, said that should work, so I plugged power in and... the IC heated so much that one of the wire disoldered. After that I again hade continuity betzeen pin 5 and 6 (i checked before that there were no shortcuts of course)

  Here is what I did :


 Red track is 5V from USB (from a laptop).
 On the right part an LM317 is used to provide 1.86V on the yellow track.
 The bottom potentiometer is use to ensure at least 4mA are drown on the yellow track so the LM317 can regulate (I checked that it does)
 Then there is the analog switch with the pinout I think it has. B1 and B2 are tied to 0 and 1.85V so I expect to read either on A
  Finally, S is tied to 0 and I should be able to plug it to 1.86V to change the voltage on A... but the IC burns before.

  Any idea what I am doing wrong before I solder 6 0.65mm pitched wires again ^^ ?

Thx,
John.

[aliarifat794]

you can provide additional cooling or use a different package with better thermal properties.

[MasterT]

I had similar issue with another part number ts5a23157,  from TI. When control inputs are not connected to anything during start up IC is "latching". Tryed 3.3 and 5V - no difference. My "work around" was to place 100 Ohm resistor in series with power line, since switch doesn't consume much current in normal operation, resistor has no negative impact.
 Sadly I was not able to post over TI tech forum, required busines e-mail for registartion. Hobbyist are not wellcome.

[Ian.M]

Floating digital inputs are problematic for CMOS, so provide a 10K pullup or pulldown for the select (S) pin *before*  applying power.

Confirm you have correctly identified the pin one position by using a DMM on its diode check range to measure from the Vcc pin to the Gnd pin.  Due to the internal protection diodes of the analog pins, if the red lead is on the true Gnd pin and the black lead on the true Vcc pin, it should read as a diode with Vf of around 1.1V to 1.5V,  which will read as open if the red lead is on the true Vcc pin and the black lead on the true Gnd pin.  If those readings are the other way round you have misidentified pin one.

Take full ESD precautions while assembling the IC on the header, and while breadboarding with it.  Also, check your soldering iron for leakage current to ground (PE, Earth) from the tip.  If you can measure more than a couple of volts AC or DC between the tip and true ground, its not properly grounded and may be blowing your IC. This is commonly caused by plugging a soldering iron/station with a three wire mains lead into a socket that has a defective ground.  Also many two wire mains lead irons and stations have excessive leakage current due to poor design or maintenance and unless arrangements can be made to reliably ground their tip, are unsuitable for delicate CMOS parts and unprotected MOSFETs.  See https://www.eevblog.com/forum/beginners/simple-fet-circuit-for-onewire-protocol-doesn_t-work-as-expected/ reply #9 onwards.

[Jwillis]

 You have no pull up or pull down resistors. Ian beat me to it but take a look at this Implications of Slow or Floating CMOS Inputs

[shapirus]

The circuit appears to be wired correctly, as long as pin 1 is located correctly.

In addition to the ESD and mains voltage leakage considerations (and of course the pull-down resistor for S to set its default state) in the above posts, did you make sure that the LM317 outputs correct voltage?

I would add a capacitor between its input and ground, even though I doubt that it will really change anything. Do you have an oscilloscope to check what happens on the output of the 317 at power-on? What supply voltage do you use before the 317?

Did you try to power it on with *both* B1 and B2 connected to ground? There can potentially be a situation at power-on when both the switches (see simplified schematic on the first page of the datasheet) are on or partially on, which will short VCC to GND via B1 and B2, and maybe the IC will stay latched in this state if that happens.

The 3157 requires only a very small amount of current to operate. You can add a resistor, say, 1k, in series with the input of LM317 to limit the maximum current that can flow into the 3157, which will give you much more time to realize that the circuit isn't working properly. That's assuming you don't have a current-limited power supply, and even then, sometimes, you'd want to add a resistor, because the current limiting mode isn't activated instantly and takes some time to kick in, and the worse the power supply, the longer that time will be.

[Jaunedeau]

(and of course the pull-down resistor for S to set its default state)

I used no pulldown, but directly wired S to Gnd in the last test

did you make sure that the LM317 outputs correct voltage?

I checked it outputs 1.85V with the 500R resistor before I add the switch. With the switch it droped to 1.28V but the doler on the switch was melting ^^

I would add a capacitor between its input and ground, even though I doubt that it will really change anything. Do you have an oscilloscope to check what happens on the output of the 317 at power-on? What supply voltage do you use before the 317?

I thought there might be some interaction between the LM317 and the witch, like resonance or something. I'll try adding a capacitor and plugin the oscilloscope on next attempt

That's assuming you don't have a current-limited power supply,

I don't, but since my psu is basically 5V from usb ouput of my laptop into a raspberry usb connector, where I thake VBus (5V) to feed the LM317, I could use the LM317 as a current limiter, that's a good idea.

I'll try doing other tests and post an update tomorrow, thanks!

[Jaunedeau]

I tried several magnifying glasses and pairs of glasses I could find and manage to take a clear picture :

I looked up for the marking again on TI site, and all results are variants of sn74lvc1g3157. So I'm sure I have the proper datasheet.

So I looked again at it and found something: since the components come in a tape, "Pack Materials-Page 1", page 29 of the pdf, has the answer:


Fortunately I had to order 10 parts and only burnt 2 yet so I still have some in the tape. Reading is not that easy but holding prescription glasses + my phone in one hand, a torch light in the second, I still manage to find proper angle and tap the button on the phone's screen:


So, it seems I was correct in guessing pin 1 and the problem is elsewhere. But at least I learnt how to properly find pin 1

[Ian.M]

Do the DMM diode range test *BEFORE* soldering the IC and see if its confirmation of the Vcc and Gnd pin orientation matches up with your expectation of where pin 1 is.

If it matches, odds are you are killing them during assembly as discussed above.   

[shapirus]

To elaborate on the diode range testing idea, a typical IC input overvoltage protection circuit looks like this:



These diodes are called clamping diodes, and they divert current to either VCC or GND when voltage exceeds or falls below the respective rail's potential plus or minus one diode Vf drop.

We therefore have two diodes in series between GND and VCC, so if we test diode continuity from pin VCC to pin GND, then we should see open circuit, and in the other direction we should see a voltage drop of ~1.1-1.2 V, exactly what you'd expect from two diodes connected in series.

This can be used to check if VCC and GND are not reversed and if the clamping diodes are not dead.

Of course it's also possible to test these diodes individually by connecting a DMM in diode test mode between a given input and VCC or GND.

...use your ESD protection wrist strap when handling these. It's not THAT likely to damage them just like that, but you need to rule out as many factors as you can when troubleshooting weird issues.

[Ian.M]

OP (reply#6): "...  With the switch it droped to 1.28V but the doler on the switch was melting ^^ ..."

That's right in the ballpark for the total protection diode Vf with Vcc and Gnd reversed. 

[Faranight]

Sadly I was not able to post over TI tech forum, required busines e-mail for registartion. Hobbyist are not wellcome.

I just used my work email to register. It worked.

[Jaunedeau]

unfortunately I'm self employed and my address is banned :/

of course I could contact support about this... if only I could open an account to contact support ^^

[vk6zgo]

Many years ago, I designed a device that used quite a number of 4066 analog switches, which were turned on in sequence to change the frequency of an audio oscillator, & perform several other functions.

Until the device was commanded on, all the 4066s were "off" simultaneously for long periods of time.

On test, things seemed to go well, but inadvertently touching one of the 4066s, it was hot enough to be uncomfortable for more than a second or so-------- swiftly touching the rest gave the same result.

Looking at the Veroboard, I realised I was relying upon the simple removal of a +ve on signal to turn the switch off, which it did, but it still drew current.

"Egad!" thought I, & quickly fitted "pull-down" resistors to ground, confident in them to solve the problem.

Unfortunately they didn't, & I ended up having to bias the control pins negative.
The + and - supplies were a bit unusual, at 6.8 v due to being out in the middle of the Western Australian Wheatbelt, with a well stocked, but a bit old-fashioned parts store,where there were a bunch of 6.8v zeners.
The -6.8v worked well & the 4066's were cool to touch.

I was always going to attempt to reproduce the conditions, but life went on, as it does
I'm still a bit wary of analog switches, however.

[shapirus]

I've actually ordered, unrelated, some of these too. Will maybe play with them and see if I have the same issue.

[Jaunedeau]

I found a TSOP to DIP PCB, on the backside it has 0.5mm tracks and it worked for the 2x3 0.65mm package, which eliminates lots of error sources.

I measured between each pair of pins in both direction with my DMM on diode position, without any connection to anything (so no power supply) here are the results :


The cells in pink (e.g. from B1 to A) do not stabilize and swing for about +-0.15V or +-0.3V around the average position.

I'm not completely sure I manage to match these numbers and what I understand from @shapirus drawing.

Should I try to put a 200R resistor on Gnd, Vcc, a 1kR between S and Gnd and another between A and Gnd, supply 5V and test resistance between B1 and Gnd and B2 and Gnd ? Or is there already something wrong in these numbers ?

[shapirus]

0.5V GND to VCC seems low -- as if it were a single diode instead of two. It also doesn't compute: you sure there's no mistake? You have 0.6V GND->B1 and 0.7V GND->VCC, but 0.5V GND->VCC? I'd expect 0.6V+0.7V=1.3V GND->VCC instead.

But it may be fine, who knows, there may very well be something that makes one p-n threshold between GND and VCC (in parallel with the two clamping diodes). Outputs look good, you see your clamping diodes connected to both rails for all inputs, except S->VCC, for some reason. The pink (and 2V+) values are probably good as well, since the IC is in an undefined state when not powered, and who knows what charge is there on the gates of the switches in this state -- so the switches can be in a partially conducting state too.

My ICs should arrive today, I'll check how their pins measure in diode mode and maybe try passing a signal back and forth.

[shapirus]

My ICs should arrive today, I'll check how their pins measure in diode mode and maybe try passing a signal back and forth.

So it works for me.

It tests exactly the same as yours in diode mode.

Here's the schematic:



And here's how I wired it on the breadboard:




Works just fine, including trouble-free power-on. The only apparent difference is that I'm powering it with 1.86V from a lab PSU without a linear regulator. And I'm pulling S down to GND with a 10k resistor here.

Maybe there is something which happens on power-on in your case, but I struggle to imagine what it could be.

For debugging, add current limiting to LM317 and don't tie any of the inputs/outputs (B1, B2, A) directly to VCC, instead use a resistor, say 10k or larger, to avoid shorting VCC to ground via B1 and B2, in case there is some transitional condition that makes it possible (and the IC latches to that state).

[Jaunedeau]

thank you so much ^^ I will be less reluctant to do other tests tomorrow.

I'll add the current limiting resistor, and some filtering caps befor and after the lm317 in case something weird happens there and post the results !

[shapirus]

and some filtering caps befor and after the lm317

When using caps, make sure you read and understand the typical applications section in the LM317 datasheet (TI's is a good one) where they talk about protection diodes.

[Jaunedeau]

Ok, I redid everything from scratch, and I found a TSOP to DIP pcb that happened to have a 4x2 0.5mm footprint on the back... just good enough for the 3x2 0.65mm and... it worked !

I could remove the current limiting resistor and add the LM317 and did not add capacitor and diodes. I don't know exactly what corrected the problem, but the differences I know between the two failed attempts and the working one are:
- use of a pcb instead of deadbuging (though I checked the first two several times)
- addition of a 10k pulldown on S pin at tome time the breadboard fall from the table and I did not notice I lost the 10k resistor. The circuit behaves the same (I did not measure if switching is different, but nothing burnt when I disconnected S from 3.3V / Gnd and connected it to the other one several times)
- the switch is powered by the 3.3V output of the raspberry pico clone that I use to get the 5V from USB (and will later talk to the switch) (5V also worked, I did not try with 1.85V from the LM317)
- the LM317 1.85V output is only used on B1

Here is the circuit for reference:

- top red rail is +3.3V
- bottom red rail is +5V
- both blue rails are Gnd (connected together)
- yellow is 1.85V
- green is Adjust for the LM317
- orange is S (select)
- top potentiometer is a load to ensure the LM317 sources at least 4mA and can regulate
- the top left black wire was meant to connect the resistor that I did not notice fall from the board

thank you all