Discrepancy between PCA9306 datasheet (from TI) and Ulticontroller rev2.1 schematic

[umagi 8]

Hi,

Trying to figure out why the PCA9306 chip on my Ulticontroller keeps blowing up, I checked the datasheet from Texas Instruments for it.

I found the following:

However, the Ultimaker schematic shows this:

So in the UM2 schematic, VREF1 is at 5V whereas the range for VREF1 as per the datasheet is only between 1.2V and 3.3V...

What the ... ??

Can any of the Ultimaker design engineers please help me out here?

[gr5 898]

I alerted Coen - the guy who designed this board.  He usually replies pretty quickly.  The specs *also* say you can put up to 5.5V on vref1.  Which is strange considering the overview.

[umagi 8]

Many thanks gr5!

By further looking at the Recommended Operating conditions, it seems this is not really a big problem, since both VREF1 and VREF2 can operate between 0 and 5.5V it seems (with 7V being the absolute maximum)

So things seem to be OK after all... I should have looked more closely to the datasheet info, but I guess I was too eager to find a solution to my problem...

However, the Design Requirements table from the Typical Application section also says:

So there, the max. for VREF1 is 4.4V... All this is pretty confusing after all.

Let me just hope Cohen will be kind enough to enlighten me with his expertise and give some suggestions as to what could be wrong with my Ulticontroller board (or is it a main board problem too?)

Thanks!

Edited March 11, 2016 by Guest

[Cohen 4]

It turns out you're right umagi, the chip is not connected as it should be. I never noticed, these chips are often symmetrical, so it doesn't matter which is the high side and which is the low.

It should not, however account for the chip breaking. The 5V is still below the recommended operating voltage of 5,5, and well below the absolute maximum of 7. Going over the 4,4 volts of the design requirements shouldn't break the chip. If it did, we would see this issue much more often, but we don't.

Did you check your voltages with a multimeter?

[umagi 8]

Hi Cohen,

 

Did you check your voltages with a multimeter?

Yes, I did, and the voltages at the test points are as indicated in the schematic.

As already mentioned, I replaced the chip, things worked for a while, then it blew again.

I had no other choice but to buy a new board (a new set actually consisting of main board v2.1.4 and ulticontroller 2.1), but as I wrote in another post in this forum, that one blew too... Same U3 chip... same frying "hump" on its top.

Since noone else seems to have that problem, it must be with my setup indeed, but I can't see what is wrong with it.

I have a full aluminum extrusion build frame, with aluminum bottom plates. The whole frame is AC grounded.

Mounted on theses plates are a switched power supply 24V/10A (also grounded) and the main PCB. The Ulticontroller is mounted on the extrusion frame.

I don't know if this may be a problem or not, but I use an external SPST (single pole, single throw) switch at the AC side and I leave the switch on the PCB (i.e. DC side) always ON (I did this due to mounting and accessiblity problems in my build). What do you think?

I build my machine in an environment where I regularly get static discharges. My control knob is full aluminum, but I insulated it from the rotary controller's shaft to eliminate that possible ESD path.

Is my switched power supply bad? I tried two different ones (for the first board, and another one for the second board).

Also, since the DC power output is floating, is it better to put the PCBs in their own plastic casings instead of mounting them on the aluminum panel and frame?

Experts! Please help me out here...

Edited March 11, 2016 by Guest

[umagi 8]

Cohen,

Also kindly notice that the NXP datasheet for their version of the chip states absolute maximum rating of 6V... Since those chips should be compatible, I wonder why NXP made their requirement more conservative...

Furthermore, the NXP sheet says that for optimum performance, VREF1 should be less or equal to VREF2 minus 1 V... Again, why would they state that?

Edited March 10, 2016 by Guest

[gr5 898]

I recommend you use an oscilloscope to look at vref1 and vref2 (and later vref1 with relationship to the other signals on the chip) during power up.

I'm guessing maybe one of the signals gets to full power before one of the vrefs? Or maybe vref1 powers up *before* vref2? My point is I think this may have to do with how quickly the power comes on - it may be that if the power supply comes up slower or faster than the "standard" supply you get burn out that chip.

Alternatively it could be during the power *off* sequence so look at those signals during that procedure also.

A different theory says somehow grounding your frame is causing the problem. But this seems less likely to me.

Static discharge doesn't usually cause visual damage on the *outside* of a chip. More typically it will damage a single resistor or two inside the chip but not generate enough heat to show on the packaging.

[gr5 898]

Static discharge doesn't usually cause visual damage on the *outside* of a chip. More typically it will damage a single resistor or two inside the chip but not generate enough heat to show on the packaging.

Well actually, once the chip is damaged with static the chip can *then* have a short circuit which can cause it to heat up. So that's a 3rd theory I guess. I'm kind of repeating your 3 things you mentioned (the 3 differences between your printer and standard printer) back to you.

[umagi 8]

Hi Gr5,

On the TI site, I found that someone named Patrick Olma (someone from Ultimaker, I presume?) posted the following question:

https://e2e.ti.com/support/interface/etc_interface/f/390/t/495210

Looking at the date, I also assume that his question was posted after I flagged it in this forum.

This was the message:

Patrick Olma, Mar 3, 2016 3:49 PM:

Hello,

We are using the device PCA9306 as i2c level shifter to communicate between a 3,3V Master and a 5V Slave. Both sides of the SCL and SDA Pins were pulled up to VREF1 / VREF2 by using 1,8k resistors.

Unfortunately we mixed the VREF pins - so the 3,3V Master is connected to VREF2 and the 5V Slave is connected to VREF1.

Most of the time the device as well as the i2c communication works well, but under some circumstances, the device gets damaged.

Could the wrong connection of the VREF pins be the reason for this?

Best Regards,

The response from the TI employee is rather interesting:

Response from Rajan Arora, TI Employee Mar 3, 2016 4:25 PM

If VREF1 (5V) is larger than VREF2 (3.3V) then the VREF1 pin can be tied to the Enable pin, and pulled up to 5 V using 200kohm resistor. The EN pin should be tied to the larger voltage supply. Is this a possibility?

You are seeing damages on the device in the configuration because the terminal voltages on the transistor are exceeding 5.5V which is the recommended max on the terminals. This can result in extra leakage current. Can you list under what circumstances you see the damage (high temperature)?

So to summarize the damage is due to the incorrect connection. if possible I recommend changing the connection on the board to match the datasheet, or tie the EN pin to the VREF1 pin.

I guess I will try to jumper EN to VREF1 then, if possible at all. I am expecting some clarification from Ultimaker though...

[Cohen 4]

I don't know any Patrick Olma, the question did not come from Ultimaker (weird coincidence timing wise though).

They are still saying the voltage needs to exceed 5,5 volts to damage the chip, which it doesn't during normal use. Tying the enable pin to 5V might be a good idea though.

Another possible reason for the chip breaking could indeed be esd. The best advise I can give you is to use the printer in a more traditional setup (original psu, not grounding the casing), and see if that fixes it. EMC can be pretty hard to grasp, and is hard to troubleshoot when not on site. I would however like to see a simple diagram of how your printer, psu, grounding etc. is connected together. Maybe I can give you some better grounded (Pun intended) advice. :-)

I will fix the chip pin-out on an upcoming version of the Ulticontroller, but it takes some time before it will be ready for production especially with al the current stock we have.

[umagi 8]

Hi Cohen,

 

I don't know any Patrick Olma, the question did not come from Ultimaker (weird coincidence timing wise though).

Amazing coincidence indeed! Some would call this synchronicity....

As to your request for more info about my setup, I attach the following pictures. Hopefully they will turn out clear enough:

I like to refer to it as "full metal jacket", because I used aluminum everywhere (support frame from extruded alu 2020 and 2040 profiles, alu bottom plate and middle cover plate, alu shaft support plates, etc.). The cover panels themselves are composite alu ("Dibond"-like), 4mm thick. The control knob is massive alu as well, and I took the - late - precaution to insulate the rotary encoder's shaft with a thick insulating crimp sleeve, after my first Ulticontroller blew up.

As mentioned earlier, I connected the ground lead from the AC mains to the alu bottom plate, near the AC power input connector. Also, I installed an external SPST switch (with built-in lamp) at the AC side to switch the system ON and OFF. This implies that I leave the power switch on the Ultimaker main board ALWAYS on the ON position... I don't know if this may be a problem or not.

I tried two different SMPS supplies, first suspecting that the first one may have been of bad quality. The 2nd one though seemed to be of high quality but I also had the Ulticontroller failure with that one. However, I'm not experienced enough to determine the quality of a switched power supply...

As to the grounding thing... this is still a shady area to me. I thought it was good practice to AC ground metal frames of machines and electronic equipment wherever possible. But maybe not when using supplies with floating DC outputs mounted the way I did? Should I put the PCBs in plastic casings instead of directly mounting them to the alu plate (main board) and alu frame (ulticontroller)?

Edited March 12, 2016 by Guest

[gr5 898]

What about scoping vref1 and vref2 on power up? I still want to know if they are coming up in a weird order due to your power supply or if the 24V overshoots to say 50V briefly?

[umagi 8]

What about scoping vref1 and vref2 on power up?  I still want to know if they are coming up in a weird order due to your power supply or if the 24V overshoots to say 50V briefly?

 

Hi gr5.

I will try to do this first thing tomorrow and post the results.

Thanks!

[umagi 8]

Hi...

I did some measurement with the scope, and found some problems indeed with my setup.

As stated above, I use a self-sourced switched power supply and an external AC switch whilst leaving the PCB switch permanently ON. This may be the cause of my problem.

When using this setup, the VREF1 and VREF2 measurements look as follows:

CH1 (blue) is VREF1, CH2 (yellow) is VREF2.

CH3 (green) is measured directly at the DC output of the supply.

Thus, a spike of almost 1V appears on VREF1 and VREF2, and the timing seems to correspond with the time when the 24V rail becomes live (i.e. the relay K1 closes), one of the figures below.

When I use the PCB switch instead, the spikes do not occur:

The 24V rail comes live as follows:

The 24V rail activation timing seem to correspond to the time when the spikes appear when I use the AC switch instead of the PCB switch.

I'm not quite sure why this is happening (is power supply coming up too slowly?), but I will immediately change my configuration and eliminate this AC switch and only use the PCB switch (or use an external switch, but at the DC side this time). I will also try to source a power adapter closer to the "official" one.

Edited March 13, 2016 by Guest