TPS62410: helps on phase margin improvement

Hello Jake,

Before we go a detailed discussion, can you please provide the complete circuit diagram?

Thank you. 

Best regards,

Excel

Hi Excel,

Thanks for the response. Please see the circuit schematics.  The step response scope shot that I got  in my previous message is with R140 = 0 ohm and C51 = 82pF.

thanks

Jake

Hello Jake,

Thanks for sharing the schematic diagram, and it looks fine.

Just to clarify again. The Vout oscillation during transient load happens only on 1.5V rail. Is it correct? Are there additional output capacitors connected to the 1.5V rail, excluding the 22uF (C192)? If there are output capacitors placed on the load side, then it can impact the loop stability as well.

Can you remove the C51 and check the transient response?

By the way, please share a screenshot of the PCB layout as well.

Thank you.

Best regards,

Excel

Actually we had similar thing ( Vout oscillation during transient load) on 3.3V rail as well, but a little bit better than that in 1.5 V rail. I have added an external Cff, which is 150pF, in parallel with R135  to knock down those oscillation. Can you confirm if it Is  a right approach?

We do have additional 11 capacitors on the 1.5V rail close to our load device ( about 6.4uF in total).  They are listed below:

1.  0.1uF    X 7

2.  1nF       X 2

3.  1uF       X 1

4.  4.7uF     X 1

Please see the scope shot on the transient response with C51 removed. I had measured the oscillation frequency. it is about 54khz. Based on this frequency and R137 and R142, the calculated the Cff  is about 61pF( based on the TI application note). I have tried 68pF, Then tried other values like 33pF, 47pF, 82pF, 100pF. The transient response with C51=68pF and C51=82pF are pretty close, but C51=82pF is a little bit better.

Also please see  the PCB layout below.

Hi Jake,

Thanks for sharing the waveform and layout.

The transient response becomes worst with no Cff on 1.5V rail. Did you try to increase the Cff higher than 82pF? I'm expecting that the transient response will improve further. Once you achieve the desire transient response, please conduct bode plot measurement and the target is to have a phase margin equal to greater than 45 deg.

For 3.3V rail, please conduct bode plot measurement as well (PM => 45 deg) to determine if the control is stable.

With regards to PCB layout, please see comments below.

1. thermal pad and GND pin should be connected directly on the top layer

2. FB resistors should be connected directly to GND pin of the IC (see figure below)

3. please follow the PCB layout reference indicated in the attached file

TPS62410_slvuat3.pdf

Best regards,

Excel

Hi Jake,

Can you try a 150pF on C51 with R140 = 0ohm and check the transient response?

Best regards,

Excel

Hi Excel,

Thanks for the comments on PCB layout. I will make those changes in the coming board spin. 

In terms of the transient responses with different capacitance of C51, I had tried 47pF, 68pF, 82pF, 100pF, 120pF, 150pF, 330pF. Non of them is good enough. relatively the best one is with 82pF. Here are the two scope shots with C51=150pF, 

Here are two  the scope shots with C51= 330pF.

Here are two scope shots with C51= 100pF

Since it is getting worse when C51 is  330pF, I did not increase the value further. 

Jake

Here are two scope shots of  the transient response with C51=120pF.

I also have scope shots for C51= 120pF. 

Hello Jake,

Thanks for sharing the waveforms. I agree it looks like that increasing the Cff further does not improve the transient response.

By the way, our field team already arranged an internal discussion with us to support this case. To avoid miscommunication, I suggest we use our field team for further discussion instead of using this platform (E2E).

Let me know your thoughts.

Best regards,

Excel

Hi Excel,

I would like to continue to work with you if it is possible. At this point, I do not want to switch to another engineer to get helps  for this case.

You have done good job so far. You have pointed out our PCB layout issues, and asked me to try different values to see the changes of transient response. I think we are moving forward smoothly. We have reached the point that I can not get through and need your helps.  

I had an introduction meeting with Sunni Love and Russell Byrd last Wednesday. They are our new account manager and  field application engineer. We had talked about this power supply issue during this meeting.

thanks,

Jake

Hi Jake,

Noted. We can still discuss your issue in this thread and in parallel, I will align with the team internally so we are all in sync.

I have additional proposal for further evaluation of your board.

1. Can you conduct bode plot measurement on your board and check the phase margin? It should be =>45 deg.

2. Do you have the TPS62410 EVM? If you have, then you can replicate all the passive component values (L8, C192, R137, R142, C51) on +1.5V rail of your board to the EVM. If the EVM with updated component values does not show any voltage oscillation during transient load, then we can conclude that the issue can be attributed to PCB layout.

3. After evaluating the EVM you confirm that the issue is related to the PCB layout, re-work one of your board based on the layout proposal below. If voltage oscillation was minimized, then we can confirm that the issue is related to PCB layout.

Let me know your thoughts.

Best regards,

Excel 

We do not have equipment here to measure phase margin. Also I do not have TPS62410 EVM right now. Russ is ordering one for me, I may get it after one week.

Per our discussion during the meeting last week, I have captured the switch node waveforms of 3 during normal operation, transient, and low frequency noise appears on Vout. For the transient one, I see the turn on time gradually changes from 275.7ns ( before applying additional  330mA load) to 291.5ns ( after applying additional 330mA load). For the switch signals with noise appears on Vout, I can not really tell the details. Since the switch frequency is high, as we keep the picture of noise, it is hard to see details on switch signals. Please see scope shots. Red one is Vout with AC coupling, yellow ome is switch node signals.

I also have tested this 3.3V power supply ( with Vcout = 10uF) with a pure resistor load.  I disconnected all the load on the board and provide a 220mA steady load via a external 15 Ohm resistor, the use an additional  10 Ohm resistor in parallel with it to provide a 330mA load step change.

I have noticed that the low frequency noises now are different from those with our board loads, which have 26uF capacitance as a part of load. It appears it has less noises, and the noise frequency is lower. I have tested  cases with Cff=150pF and without Cff . It looks like it has slightly bigger phase margin with Cff=150pF. please see attached scope shots on 3.3V supply with  configurations of   Cff=150pF, Vcout =10uF, external steady current load of 220mA .

also please the scope shots on the noise I see and step response on  the 3.3V supply with  configurations of   no Cff, Vcout =10uF, external steady current load of 220mA .

Hello Jake,

Thank you for sharing the waveforms.

Both the ripple voltage and transient response improves when you disconnect the 3.3V rail to the system load with 10uF output capacitor. Please see suggestion below.

1. With Cout = 10uF; Cff = 150pF @ 3.3V rail, can you connect R136 directly to pin 8 of the TPS624210 instead of the GND pad? Check the ripple and transient load waveforms. The intention of this experiment is to determine if the issue is related to FB resistor layout.

2. Increase the Cout from 10uF to 26uF (10uF + 16uF) to simulate the actual capacitance of the system seen by the converter. Check the ripple voltage and transient load waveforms again.

If the everything works fine with the configuration above, then we are confident that the device can support your application.

By the way, we cannot conduct the EVM evaluation yet because there is no available boards in the lab. We are expecting to receive it either end of the week or early next week. Keep you posted for any update.

Best regards,

Excel

Hi Jake,

Can you re-measure the waveforms and include the VIN?

Thank you.

Best regards,

Excel

Hi Excel,

With Cout = 10uF, Cff= 150pF at 3.3V rail, I have done the reworks to directly connect R136 to pin 8 of the TPS62410, I do not see improvement on the low frequency noise ripples and step load transient response.  please see the noise waveforms and transient response waveforms below.

with external resistance load, Cff=150pF, and reworks on R136 ( directly connect R136 to pin 8 of TPS62410) , I also have increased Cout  with the following values:

1. Cout = 26uF , as you suggested. I do not see noticeable change

2. Cout = 48uF, simulate with our original conditions, which are Cout =22uF plus  26uF decoupling caps on the loads. I was expecting worse performances, but do not see noticeable changes as well. 

please see the low frequency ripples and transient responses with Cout=26uF:

also please see the low frequency ripples and transient responses with Cout=48uF:

Based on what I see with those signal waveforms, the power supply has better performance with external resistor load (even with Cout =48uF.). 

For comparison, here are step responses when power supply (with Cout-22uF and Cff=150pF) is with system load on the board. Those waveforms were captured before we had reworks on R136. I think it will be the same even we connect R136 to pin8 of TPS62410 based on the tests with external resistor loads.

Hi Jake,

Thanks for sharing the waveforms.

I looks like that the Vout waveform during transient load does not have the voltage oscillation with the new setup while high frequency oscillation on the ripple voltage is still present. Anyway, can re-measure the transient load and ripple voltage and include the VIN waveforms (place the oscilloscope probe across VIN and GND pins of the IC)? I wanted to know if the voltage oscillation can be attributed with IC supply voltage.

Best regards,

Excel

Hi Excel,

As you suggested, I have looked at Vin with/without  load step changes. The input does some small noises, some of them sync to those output ripple noises, but some of them do not. The test setup is: 3.3V rail with external resistor load, Cout= 48uF, Cff=150pF, R136 is directly connected to pin 8 of TPS 62410. please see the scope shots. The red one is Vout, and the yellow one is Vin.

Hi Jake,

Thanks for sharing the waveforms.

Can you identify which of the waveforms were during transient load and stead-state conditions? I will ask Russell to arrange a concall to speed-up our discussion and close your issues.

Best regards,

Excel

Hi Excel,

Among the four scope shots in my previous post, the first three scope shots are with load step changes and the last one has no load step change. 

Also in the first scope shot, the yellow channel ( Vin ) is with DC coupling, and in rest of three scope shots, both channels are with AC coupling.

Thanks,

Jake

Hello Jake,

Let's discuss in the details tomorrow the progress of your evaluation during our concall. 

Thank you.

Best regards,

Excel

Hi Excel,

per our discussion this morning, I have done the reworks on 1.5V rail power supply to directly connect R142 to pin 8 of TPS62410 and checked the transient response. I do not see improvement.

Please see attached scope shots. circuits configurations are: C51 = 82pF, C192 = 10uF, system loads are connected to 1.5V rail.

Thanks,

Jake

Hi Jake,

Thanks for sharing the waveforms. It looks like there is no improvement in changing the R142 connection.

Can you capture the waveform with test conditions with VIN, ILOAD, SW pin, and VOUT signals? Please zoom-in on the waveform so I can see the SW pin behavior (refer to the figure below). Lastly, add a Rload on the output to make the device operates in PWM mode, then do a transient load (capture the signals as well). 

By the way, you can send the data to my TI email account so I can respond immediately.

Best regards,

Excel