TPS63700: startp issue

Do you want to discuss a TPS63700 issue or do you need help logging in?

I’m having startup-issues with the TPS63700 too (welcome to the club…)
Initially, Enable was tied to VIN. After having startup issues, I decided to drive the Pin from my application. This made things worse: Re-Enabling the converter, when the output has now completely fallen to zero, likely makes it switch off. The bias-problem was already addressed by other users.

But even reducing the shutdown-time to several seconds doesn’t completely resolve the issue.

So, if it doesn’t start the 1st time, it reliably runs into the problem next time, when the enable-function is used. Unless you wait for an unacceptable long time…

The problem seems to depend on the input voltage: on some, it only happens with high input voltage (5V), on some only with low input voltage (3,6V).

The resistor divider was on the datasheet limit of 10µA, so I increased its value. This improved the startup behavior somehow. But it still happens.

The negative output is being filtered, the load is only about 4mA (it should be able to deliver >300mA).
Both inductor and rectifying diode are designed for 1A peak current.
From the datasheet, there is no condition, that would explain the behavior:
- No undervoltage lockout (input voltage is stable as can be)
- No overtemperature (overtemperature would make the revice restart)
- No overvoltage lockout: I sometimes observe a light overshot in the voltage but far under the -19V-limit from the datasheet.
Is the regulation loop wrong somehow? If so, why?
Some forum users have resolved the problem increasing the inductor value, some decreased it.
Here's somecharts form average, bad and good startup:
All prerequisites seem to be met:
- Input voltage within range, filtered and stable
- Reference voltage stable
- Reference current below recommended value
- Output filtered, current consumption low.
Why can the enable not just fully enable the device and makes it do his job?
Why does it go off without a proper reason?
And if so, why can’t it try to come up to life again?
Instead, it’s a mortally offended diva sitting in the corner and ignoring the world (sorry for the untechnical metaphor)…
@TI: Any hints…?

It looks like the waveforms did not come through the post.

He can always order the EVM and test that with its known good layout and components. But I would likely recommend a different device ultimately. What Vout and Iout does he need?

Hopefully, the pictures will get through this time, otherwise, I have to rely on Katarina doing it in my place.

Schematics:

 

Good startup

  

Bad startup caused from bias ( I can avoid that now...):

Bad startup, even though no bias present. Reason unclear...

 

>But I would likely recommend a different device ultimately.

Is that a hint, that the TPS63700 contains a design flaw somehow...? ;-)

The TPS63700 is already designed in and produced in the 1st delivery batch of the PCB. So a solution would be very welcome.
Is there any known workaround for the startup problem?

The input voltage ranges between 3,3V and 5,5V. Output voltage is -4V, current is approx. 4mA only.
Is there a pin-compatible replacement?

Best regards,

Phil

Sorry, I mixed the last 2 pictures:
- The last image is the biased case. The red curve is the reference voltage. (just to make sure, it's stable: it is).
- The 2nd image from below is the bad startup without a bias. The red curve here is the feedback.

The interesting point about the bad startup without a bias, is that the output voltage is going significantly below -4V (like in the good case). It seems, like the regulator is going too far and noticing its fault, stops working completely (?)

Best regards,

Phil

Yes, it seems you have an overshoot which then causes the IC to turn off.

Could you post your PCB layout and give the part number of your inductor? You should also try with the recommended 4.7uH.

In any case, this IC is quite overkill for a 4mA output. Most customers that I work with would much rather use the TPS60400 family to get a simple and small inversion of Vin.

If you need a regulated Vout, the TPS62120 can do this in a much more efficient way with this app note: www.ti.com/.../slva478b.pdf

Hello Chris,

 thanks for your prompt answer. I'm using an inductor from Wuerth from the TPC-TH-line: 744029003. I have already ordered samples of the 4,7µ type.

According to figure 17 of the datasheet of the TPS62120, the converter would reduce the switching frequency around 100kHz at light loads. This is likely to sample down with the sampling frequency of my A/D converter into the audio spectrum.

At the moment, I still hope to save the current design with TPS63700 somehow...

It would help to understand, in what conditions the converter is going out of regulation. Shouldn't the FB pin sense the rising voltage and react before the overshot is taking place? Even, if a light overshot occurs: why doesn't it restart working, after dropping again? Is that an additional safety circuit? (the datasheet doesn't mention that behaviour). How can I avoid the overshut?

What possible workarounds can you imagine?

- An additional series resistance with the inductor to slow the starting ramp (at the cost of efficiency)?
- Increasing (or decreasing) capacity on the output?
- Does the converter need or benefit from a minimum load?

Best regards,

Phil

I recommend ordering the EVM and modifying it to your circuit to see if the same behavior occurs. It may be layout related with noise, etc.

Yes, the TPS62120 has a power save mode which will reduce the frequency. It will likely be different than Figure 17 as you will have an inverter circuit.

I recommend increasing the inductance and making your layout match the EVM.

From your latest experiment, it sounds like the device does start reliably but the output is not well regulated. This likely comes from PCB layout. Testing the EVM with your components is one way to identify a layout issue. This device, like all SMPS, is sensitive to board layout.

Hello Chris,

  If you take a look at the charts, you will notice, that the overshot is not a high frequency ringing, as you would expect from poor PCB layout. I checked this using a probe with short traces: It's rather a slow, "clean" overshot in 100µs time-domain.

Giving the status of my project, experimenting with the EMV is not a practical option:

- Not all of my devices show the effect. Some boards start reliably every time. Finding the startup problem would mean, I would have to buy several EVM boards, in order to find a board, showing the same behaviour.

- We have a  delivery batch of fully fitted boards. Going back to experimenting with EVM, redesigning the board, producing and testing it would take several weeks if not months. Please understand that we are close to production already.

It would help me to understand the reasons behind the overshot:

1) Why is the regulator not reducing the output voltage, if the voltage becomes too high?

2) Why is the regulator not starting again, if the voltage drops?

3) Is there a mechanism, that restarts the regulations, apart form reenabling?

4) Is there a known bug in the regulation, apart form the bias issue?

5) Does the regulator need a minimum load? If yes, what is the minimum load?

6) Would a series resistance help to reduce the overshot at the cost of efficiency?

Best regards,

Phil

Ah, it is new information that not all ICs show this issue. In this case, order 1 EVM and take an IC that did show the issue and put it on the EVM. In this way, a failing IC can be tested.

As this app note explains, layout is very important: www.ti.com/.../slva219a.pdf

Excessive noise can cause any IC to malfunction.

There is no minimum load requirement for this device.

After it turns off, it is latching off which requires a re-enable to start again.

A series resistance with the inductor should not have an effect but you could test it and see. This would be the same as using an inductor with higher DCR.