TRS3221: Voltage Measurement @ C2+

Hi Marco,

Marco Rojas said:

Since C2+ is a node in the charge pump circuit, I assume it is a switching node and not DC.  Is this correct?  Can you give me an idea of what the waveform at this pin might looks like?

https://e2e.ti.com/support/interface-group/interface/f/interface-forum/867035/max3232-question-about-charge-pump-waveform

^not exactly the same device but I assume this is what we would likely see.

I've put in a sample request to get the TRS3221EVM, I assume it should arrive sometime next week so I can test this myself in the lab and get back to you.

Marco Rojas said:

Is the C2+ a high impedance input?  When I probe the pin with a 10M passive probe that is grounded, I get the device to fail?  When the gnd on the probe is not connected, the device continues to work.

It should just be highly capacitive with current/charge dumping from one cap into another (C2+). Using a high impedance passive probe to GND should not cause a device failure. I'm wondering if maybe they have the probe set up for 50 ohm termination instead of 10M. 

Marco Rojas said:

PCN 20210720000.1 was issued several years ago.  Any idea if the charge pump circuit changed in some way that might explain why the test passed on older devices and no longer passes?

This was the PCN related to the new die change from the old LBC3 technology (150 and 200mm wafer technology) into our newer LBC7 technology (300mm wafer tech). So a completely different die was used. This shouldn't affect their test case though (customers using the new die haven't reported issues like this).

-Bobby

Hi Bobby,

We have a bunch of older and newer boards.  The C2+ pin on the older boards have basically the same waveform as what was shown in the link you provided.  The newer boards exhibit a slightly modified waveform which appears to cause them to fail the test that is being performed at ICT.  The test is a simply looking for 3V DC at the pin and I question the validity of that test, but I would like to understand if the waveform we're seeing on the new boards is to be expected.  The first waveform below is the old boar.  The second and third are a new board at two different time bases.

Are the new waveforms something we should expect to see?

Thanks

Steve

Hi Stephen,

I did some testing in the lab with the EVM but my waveforms don't seem to look like the ones you uploaded. 

What Vcc are they using?

Can you send a schematic for me to look at? (What are the caps on C1, C2, Cv+?)

Are they just powering up the device and measuring the C2+ pin? Or are they sending data? 

Are they also probing the V+? (Does it ever droop when the test 'fails')

-Bobby

Hi Bobby,  

We're using 3.3V, please see attached schematic.  I'm not sure how the CM is testing but I took the scope shots in the thread and those were taken with no data being transmitted or received and only the C2+ pin being probed.  Does the EVM that you're using have a relatively new device on it?

Thanks

Steve 

Hi Steven,

The schematic provided looks like the caps match the EVM's set up.

Unfortunately the EVM I'm using has is for the E version of this device. I figured we may have used the same die for the new PCN die but after investigating, it seems the PCN'd die isn't the same as the E version. I'll probably need to put in a digikey order to get breakout boards since the EVM only supports the QFN package (non E version doesn't come in this package).

Can you get a scopeshot of the V+, V-, and C2+ all on the same scope? I want to verify the V+ and V- are swinging to the correct voltages without any clipping when we see C2+ hold onto ~3.6V. It may also help to see if you send data, if T1out behaves correctly. 

A top side picture of the device would also be helpful.

-Bobby

Hi Bobby,

Attached are two scope shots showing the 3 voltages.  The first one is when I have an RS232 to USB adapter connected to the port (and plugged in to my PC) and the second one is with the adapter removed.  When the adapter is installed, I am not sending any characters and am not seeing any characters at the terminals.  A pic of the part is also attached.  I had a setup issue and could get my CPU to transmit characters so I need to resolve that.  When I had looked at the RS232 levels once before they seemed find.  Even without characters being sent, the steady state waveform at C2+ looks quite different than the older devices that we have.

Thanks

Steve

Steven Santangelo said:

The first one is when I have an RS232 to USB adapter connected to the port (and plugged in to my PC)

The V+ and V- looks pretty stable at +/-5.5V, This looks close to what I expect. (With my E version, I saw similar results)

Steven Santangelo said:

the second one is with the adapter removed

This looks like the device is in it's powered down state. During my tests with the E version, I did not see this. I would expect V+ and V- to be +/-5.5V with 10% tolerance. My guess here is that the device is in it's automatic power-down state since RIN is LOW for too long. 

From the picture, I don't see the caps (C148-C152) for the RS232 device. Are they on the bottom side of the board? You usually want to have these caps close to the device, the further away they get the less effective they are. Any changes in the layout/PCB from the passing and failing tests? 

EDIT1: I'm working on getting some breakout boards and units to retest using the non E version that your using. Can you verify the package you are using? The picture shows the device marking as 'RS21EC' but the only orderable part is the PW package which has a device marking of 'RS21I'

EDIT2: Please verify the current device's orderable name, the non-E version does not have a DB package and we did not put a new die in this package. Is customer using TRS3221E in the DB package?

-Bobby

Hi Bobby,

Yes, I agree the caps probably should have been on top.  They are located on the backside directly under the device and connected with pretty short etch so we really just have the additional inductance of the vias.  The boards with the older and newer device are identical layout wise.  The PN is our system is: TRS3221ECDBR, so yes it's the DB package.

Thanks

Steve

Thanks Steven,

Steven Santangelo said:

They are located on the backside directly under the device and connected with pretty short etch so we really just have the additional inductance of the vias. 

This should be fine. Just wanted to make sure they weren't somewhere really far away from the device.

Steven Santangelo said:

The boards with the older and newer device are identical layout wise.  The PN is our system is: TRS3221ECDBR, so yes it's the DB package.

Gotcha, I was thinking that you had the non-E version in the DB package and spent 2 hours digging through our system and asking my colleagues if we still had the non-E version in the DB package and if we placed a new die in that package.... Long story short, it didn't exist and your feedback here helped confirm my initial understanding was off.

I placed a digikey order for breakout boards for the DB package, hopefully I can get them by end of this week. My initial testing with the other E version EVM should be a fair comparison now though. Those scopeshots didn't show any clipping (around 3.8V) on C2+ like what you are seeing though. Would you be able to send 2 or 3 units (that you see fail) to me so I can see if they also have the same behavior on the breakouts? 

-Bobby

Hi Bobby, 

I can see about getting a few parts from out CM.   I neglected to mention that while I was recently asked to look in to this, the test issue has been going on at the CM for a little while.   They were concerned that it was an issue with bad devices so they initiated a failure analysis with TI with the result being no issues found with the 3 devices that were returned.  Below is the report.  

By units do you mean just the transceiver or our boards?   If it's the transceiver, it may be easier to just remove the part from a couple boards than to get the parts from our CM in Singapore.   I'm also going to order a few parts from digikey an try those as well. 

Thanks

Steve

QEM-CCR-2310-01399 Report.pdf

Hi Steve,

Steven Santangelo said:

By units do you mean just the transceiver or our boards? 

Just the transceiver (TRS3221E).

I've got confirmation from my admin that she was able to place the breakout board order and I also placed orders for new samples from TI so we can evaluate both to see if the charge pump circuit is the same between the your suspected failed units and the new units. I suspect I may receive everything by Friday or early next week. For a mailing address for the transceiver devices, you can ping me at duynguyen@ti.com or coordinate with your TI representative (Marco who created this thread). 

I'll update you on when I receive the breakout boards and the new TI units.

-Bobby

Hi Bobby

I tried sending you an email at that address but for some reason it bounced back.  Can you please send me an email at steve_santangelo@waters.com and I will respond?

Thanks

Steve

Hi Steven,

Looks like Marco put us in touch offline. I'll let you know when I receive the breakout boards for testing, for bookkeeping purposes please allow me to have the last response on this thread until any new updates occur. 

-Bobby

I'm going to close this thread since the discussion has moved offline. 

-Bobby