DAC8775: Buck-Boost converters not starting at elevated temperatures

HI Henrik, 

Joseph will review your questions and get back to you soon. There may be a delay in response due to the thanksgiving holiday. 

Best,

Katlynne Jones

Happy thanksgiving to your all.

Additional observations are that it appears that the Alarm pin is pulled low only when more that one channel or BB-converter fails. Either that or only when a current output is failing (HSCLMP fail??)

Looking forward to hear from you.

BR Henrik

Henrik,

Sorry, but I've been out of the office for a few days because of the holiday in the US.

I don't know of any past startup problems for the buck boost. The DAC8775 is rated to go to 125°C, and I've never had any issues operating over temperature. Additionally, I can't think of any reason any channel would operate differently unless they were set up in different modes. And while you're seeing the problem most on channels C and D, that would mean that it's not aligned in that way (in your setup, C and D would be different setups).

Out of curiosity, have you tried changing the mode of operation of a failed channel to see if it comes back? Have you looked at the voltages at VPOS_IN_x or VNEG_IN_X.

I think that the PGx indicates if the voltage output fails to reach an operational level (which basically is VPOS_IN_x > 4V and VNEG_IN_x < 3V as described on page 41 of the data sheet). I'll need to check, but I think that the converter still tries to operate if it does not reach the correct voltage.

Are you able to post a schematic? I think one of the things I'd most like to see is what inductor you're using in the buck-boost.

For the last post, you may not get an alarm indication just for not getting the PGx of a channel. For the alarm, there are a set of conditions outlined on page 40 of the datasheet. You'll get the alarm when the current output is past HSCLMP and you may get the alarm when the voltage at IOUT isn't high enough to drive  a load. Other than that, the alarm might not get triggered.

Joseph Wu

EATON_SDCH1V5040-101M-R.pdfJoseph

Thanks for your response.

The inductor used on this is Eaton SDCH1V5040-101M-R (datasheet attached). It shares footprint with the Würth type suggested in the
reference design and is our suggestion as alternative. Looking at the data the differences are small, but I think the Eaton type
is slightly better. But again differences are small. I have ordered the Würth type to test with, but I really don't think it changes anything.

It kind of makes sense that the alarm pin is only triggered when BB-converter fails on current outputs. I guess the DAC looks at voltage differences
between the output mosfet gate and the drain voltage and not the actual voltage on the gate (?)

Due to customer confidentiality I am not able to share full schematic. If needed it requires we take it offline. The output
channel is shared here. They are the same for all channels.

Measuring the voltages at VNEG and VPOS, reveals that only the positive arm is failing. Regardless of the output mode (+Vout, +Iout, +/-Vout, +/-Iout)
Reconfiguring the output from Iout to Vout or Vout to Iout has no effect. The failing channels comes back only after a power cycle when temperature has dropped.

VNEG / VPOS measurements. Negative currents seem to work on the failing output (at least -20mA), but not voltages.

Current output load = 15.27R, Voltage output load = 28kohm

Schematic of one channel.

NNOTE: the 15V TVS (D1405) will need to be changed to a higher rating in production - but that is another case that we sorted together with our surge problems. The naming convention: Vxxx_AB means channel A on DAC B

BR Henrik

Update:

A: I added 47uF electrolytic cap parallel to the equivalent to C1404, to rule out the effects of the DC bias on C1404. This had no effect.

B: I replaced the inductor on one channel, to the one suggested in the reference design. The Würth type. This had no effect, the DAC behaves as described above.

Best Henrik

Thanks Henrik,

I was concerned that the inductor had too low of a saturation current, but the inductor looked like it was a good substitute for the Wurth inductors we use in the EVM. I still don't have too many ideas about how this problem is coming up. I did have two suggestions that may or may not help.

First, we do have an EVM and I know we've run it over temperature in the past. The only difference that I see in the buck boost is that we do use some ferrites and a snubber in the connection from the diodes back to VPOS_IN_x and VNEG_IN_x. You can see it here:

I might consider duplicating adding these passives from the diodes going back to VPOS_IN_x and from VNEG_IN_x.

Another thing is that each channel has the TLV9362 connected to VSENSEP_x and VSENSEN_x. If the VPOS_IN_x and the VNEG_IN_x are used as power supplies for these opamps, I'd consider testing the circuit removing these op-amps, just using a short to see if the op-amp startup is causing a problem pulling current from the supplies. The TLV9362 does use a moderate amount of current, and maybe there's extra current at startup that the buck-boost isn't able to supply. You could a board where a channel has not powered up, remove the op-amps and recheck the operation.

Joseph Wu

Josheph

Thanks.
I don't think the snubbers and filters will help, but it might be worth considering if we encounter too much over/undershoot, or noise problems.
We haven't seen that. Yet... fingers crossed.

Actually, removing the TLV9362s was next on my list of things to try (eays for me to say ;-) ).
And actually it seems to help.

At start up, the local MCU is setting the outputs to voltage mode by default, before getting the configuration from the master controller.
Setting to voltage mode, also sets the VPOS/VNEG rails to +/-15V. The current draw in the opamps might be just enough to the FET inside the DAC past what they can handle at startup.
So now we will try setting the outputs to current mode by default. This should set the rails to +4.5V/-5V, and consequently currents should be lower.
Then hopefully we will be able to reconfigure once the buck-boosts has started running.
I am waiting for the sw guy to do an update. I will get back, once this has been tested.

thanks Henrik

Henrik,

Adding the snubbers was a bit of a stretch, but I thought I'd suggest it. The removal of the TLV9362s was the big suggestion. It's an added load on the buck boost, and I thought there was a chance they might require extra current at the initial power up. I'm not entirely certain that initially setting the rails to +4.5V/-5V, will be the best way to go. At startup, the current may be very different than when the device has reached some steady state. Regardless, you may try it. 

You could also use a different op-amp with lower current. I think the TLV has a very low input bias current, which is good because of the high input resistor you've placed on it. However, the offset is still high enough to think about and the bandwidth is probably higher than you need. You could change it out with an OPA202 with some series resistance or an OPA206 which does have some extra over-voltage protection.

Joseph Wu

Joseph.,

We tried with current mode as default setup, but it had no effect. I know it would not be a long term fix, but could have satisfied the immediate need and calmed the fire - if it worked.

I see your suggested opamps draw much less current, and will probably work on the buck-boost rails. The 27k input resistor is there for surge, eft and esd concerns. 27k requires low input bias. It works with 27k, but we have not tested other values. With higher input bias currents we may need to lower the input resistor value, to achieve required accuracy. But will the opamp survive surge then? that is a big question.

Due to space constraints the TLV we are using is in the 8-pin SOT-23 package (DDF). Can you suggest a pin compatible alternative?

Thanks in advance.

Best Henrik 

Henrik,

Just as a test, have you just completely removed the op-amp? Just by reconnecting the VSENSEN and VSENSEP to the ground and output, the board would be functional, and you could just check the buck-boost startup in those conditions. 

I don't know about a pin-compatible alternative to the TLV - I'll ask around for that.

Joseph Wu