TPS92515: current regulation with +/-7% accuracy

Vincenzo Pizzolante

Part Number: TPS92515
Other Parts Discussed in Thread: TPS92513, TL431

hi all,

long story short: a customer of mine is designing with TPS92515 and is struggling to achieve +/-7% accuracy in current regulations.

with the TPS92513 is doable but with -515 seems mission impossible...

could you roughly confirm that accuracy of TPS92515 is a bit worst than -513?

could someone review the equations (5) & (6) are correct?

do we have an error/accuracy estimator/calculator to help customer to check what is the best he could achieve?

thanks a lot in advance

Vincenzo

over 7 years ago

0 Irwin Nederbragt over 7 years ago

TI__Mastermind 28655 points

Hello Vincenzo,

The 515 regulates peak current so the tolerance of the inductor is part of the equation. Also if the inductor isn't sized right the saturation can change with temperature getting different results also.

Are they using Iadj? If so what value are they supplying?

I've used the equations before and they seemed to work (I will run them through mathcad again, let me know the values they are using so I can see if it matches). I also create a simplified equation if the LED stack voltage is high enough. It just uses I = C*(dv/dt) without the RC part and gets fairly close.

There is a design tool for the TPS92515.

What are they seeing for accuracy and on how many units? What are they using to collect the data, their own design or something else? Is Coff used from the LED stack?

Regards,

0 Vincenzo Pizzolante over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 25856 points

hi IRwin,

thanks for the prompt reply.
we have 7 boards at the moment, with Viadj = 2,054V (still investigating on this source, that is controlled by uC and dip switch).
Calculated Iout = 695mA; IL max = 734mA; I p-p = 83mA.
Measured: Iout = 655,8mA; IL max = 693mA; I p-p = 94mA
Anyhow the values are withing the +/-7% range, just customer is puzzled by the bias toward down...

thanks again
KR
Vincenzo

0 Irwin Nederbragt over 7 years ago in reply to Vincenzo Pizzolante

TI__Mastermind 28655 points

Hello Vincenzo,

The ripple is slightly higher so part of the offset, though small, is (94 mA - 83 mA)/2 or about 5.5 mA. Can they look at the inductor current? If the inductor is heading into saturation this will shift the average current down. Also, do you know what switching frequency it is set to.

Some things in the layout can cause this as well such as the placement of the current sense resistor.

Regards,

0 Irwin Nederbragt over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 28655 points

I have not heard back from you so I believe you may have solved the issue.
The thread will close now but will reopen if you reply.

0 Vincenzo Pizzolante over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 25856 points

hi Irwin,

sorry for the delay, I was ooo...

we have another small challenge on this design.

We are driving some on-off / supply cycling tests.

We used a lab supply to remove all the effects of the AC/DC drifts.

We measure 2 currents, 671mA vs 689mA, constant.

We figured out that the output current depends on the on/off cycles duration:

4 sec ON, 2.0 sec OFF --> ILED = 671mA
4 sec ON, 2.4 sec OFF --> ILED = 689mA

Customer believes that this misbehavior is related to the CVCC vs CBOOT ratio or absolute values (working now on tuning these values, 1uF / 100n as for the EVM by the way).

I dread is a thermal drift, the most probably parameter that changes with the on/off cycles...

Did you ever experience something like this?

Any advice is more than welcome...

thanks a lot in advacne
KR
Vincenzo

0 Irwin Nederbragt over 7 years ago in reply to Vincenzo Pizzolante

TI__Mastermind 28655 points

Hello Vincenzo,

I'd like to find out more about this design. What is the input voltage, output voltage, is Viadj used or fixed, what is Roff and Coff, is the on/off controlled with the PWM pin? What is the nominal switching frequency when testing?

You can look at the Vboot value with a meter or scope (as long as it has high enough impedance). Is it going through a boot refresh cycle?

Best Regards,

0 Vincenzo Pizzolante over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 25856 points

hi Irwin,

no is the VIN that is switched on/off...
there is a diode (reverse blocking) and a 270uF // 2.2uF bulk on VIN...
the VIN is 48V nominal, coming from a flyback (the test were driven with a la supply though)

VLED = 36V nom, ILED =0.7A nom, Roff / Coff = 33K / 470pF, L = 220uH

VIADJ comes from a TL431 reference through a voltage divider (a discrete digital pot, made with small fet and GPIO to change 4 current levels from 0.35A to 0.7A), max voltage here is around 2.05V (for 0.7A)

PWM is not touched at all...

PS: I have the schematic too, but I can't send here

hope this helps a bit
thansk a lot in advance
KR
Vincenzo

0 Irwin Nederbragt over 7 years ago in reply to Vincenzo Pizzolante

TI__Mastermind 28655 points

Hello Vincenzo,

There's a lot going on here. Have they looked at the two current waveforms to see if they line up both peak current and frequency? How are they measuring the current, is it only at the time it is on? If they are using a multi-meter they sometimes have averaging circuitry that can change current measurements that are changing. It could be thermal related but I wouldn't expect such a big shift when the two operating points are not much different unless the thermals are much different.

I get about 545 ns off time, on time will be close to three times that so Fsw is close to 460 KHz? di is about 0.09A?

One thing that can cause variance is the inductor temperature if it is in partial saturation. If it is hotter the average will drop which is the correct direction for the data above.

Best Regards,

0 Vincenzo Pizzolante over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 25856 points

hi Irwin,

thanks, I will check it.
in the meantime I got an update:

Yes, current ripple is around 90mA pk2pk.

Device works in CCM all the time, also the temperature drift doesn't apply as the results are the same at room temp & 50degC oven temperature...

The Rsense is 2x 0R56 in parallel = 280mOhm for shunt, when Current drift is ≈ 19 mA.

When changing to 2x0R22 = 110mOhm, the delta Current is ≈ 49 mA instead.

Inductor is WE 7447709221 (I DC = 1,8A max; Isat = 2,2A typ.), actual value measured is 196uH.

It is not saturating, at least from the current shape, but there is no thermal characterization of the coil...

hope this helps a bit
KR
Vincenzo

0 Irwin Nederbragt over 7 years ago in reply to Vincenzo Pizzolante

TI__Mastermind 28655 points

Hello Vincenzo,

It may be difficult to measure, I would check if the peak current is the same or the ripple is the same, one of them is changing if the average output current is changing.

Regards,

0 Vincenzo Pizzolante over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 25856 points

hi Irwin,

we tested also the EVM and there is no issue there...

we solved increasing the CVcc, that bulks the VCC when the main power supply is off (the 48V input).

Anyhow it worths to verify the state machine in power cycling considtions

thanks a lot in advance

Vincenzo

0 Irwin Nederbragt over 7 years ago in reply to Vincenzo Pizzolante

TI__Mastermind 28655 points

Hello Vincenzo,

If the IC is powering down and back up with the main supply is off I don't think you can consider that repeatable. When you increase the Vcc capacitor does the IC stay on the whole time?

Regards,

0 Vincenzo Pizzolante over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 25856 points

I would say yes, but I'm not 100% sure, need to check...

0 Irwin Nederbragt over 7 years ago in reply to Irwin Nederbragt

TI__Mastermind 28655 points

I have not heard back from you so I believe you may have solved the issue.
The thread will close now but will reopen if you reply.

Power management

Power management forum

TPS92515: current regulation with +/-7% accuracy