LM5122-Q1: Minimum Recommended Switching Frequency

Thank you, Niklas,

I am using the equation:  4 on page 17 of the LM4122 (it is not a Q1 as my thread orinally suggested; and in the current design I find a 301kOhm part, which would yield a fsw of 30kHz, its a -32 to -43 to -28 buck application, the - 48 is tied to Agnd and Pgnd and the 0 V to vin for a duty ratio of 0.6586.6kOhm resistor for a slope of nearly .9.  The slope compensation makes sense due to the duty cycle, but fsw seems way too small.  Given the slope depends on the fsw, and that is in question, can you provide typical bounds for the slope for the duty cycle calculated? 

The slope resistor is 86.6k, with a duty cycle of 0.65 

Hi Guillermo,

Can you share the schematic of the design you have found?
Is it a PMP reference design from TI, or was it created by a third party?

Thanks and best regards,
Niklas

I Nikla, I am unable to do so, while the application is "standard" I am bounded by the Arms Export Control Act, clearly stated in the schematic - we do sensitive work for the U.S. Navy - apologies.  

On another note, what is the quickest way to tap into the fsw out side of putting a second tap point on the sense resistor? - or measure the inductor current - 

We have two rails, 34 to 43 to 28, and its negative mirror. The positive down voltage is done by a LM46002-Q1, whose "hang out" components makes sense to the specs - this one is throwing me for a loop - all pun intented. 

Hi Guillermo,

I understand that you cannot share the schematic.

I would recommend to have a look at our quickstart calculation tool for LM5122
https://www.ti.com/tool/download/LM5122-BOOST-CALC
All equations and calculations are the same for Q1 and non-Q1 version, so this does not make any difference.
The calculator does not allow negative voltages, but if you consider the -48V level which is tied to AGND as "0V", the calculations for duty cycle, slope compensation, etc become valid again.

Regarding the switching frequency:
This device is optimized for switching frequencies of 50kHz to 1MHz. If the oscillator is set to 30kHz or 2MHz, the device will try to operate as such, but it is outside our recommended range, as there are no qualifications or tests run at these frequencies.

I cannot explain why this reference design runs at 30kHz, but I would recommend to increase it at least to 50kHz.
If higher switching above 1MHz are desired, I would recommend to consider our newer generation parts, like LM5125, which support switching frequencies up to 2.2MHz.

Best regards,
Niklas

To measure the actual switching frequency, the switch node voltage or driver signals are the best indicators.

Thank you, se have a TP on SW - Question:  Driver signals - you mean HO and LO? those are gate signals to the MOSFET - I am always hesitant to  probe gates as they are fast switching entities, and in my previous life (FA) dropping a pad down on the gate, maybe problematic as it may load it and drain charge from it - can you be specific on the which driver signals you are suggesting the Source / Drain of the MOSFET?  - depending on configuration? 

Thank you, se have a TP on SW - Question:  Driver signals - you mean HO and LO? those are gate signals to the MOSFET - I am always hesitant to  probe gates

Hi Niklas, I did not follow your format for the Duty cycle, I simply did a voltage energy conservation on the inductor (Vin-Vour)D+(1-D)(Vdd-Vout)=0 Vin for me is zero, Vo =-28, Vdd=-43 and Vout =-28, that led me to D43=15, or the same duty cycle if you assume 28 at the input and 48 at the output -the only values your calculator accepts.  If you do it that way - which mimicks your schematic - then get using energy conservation:  28/43=1-D, with a D=0.3488I chose 300kHz because somewhere in your schematic, I read this thing is optimized for 300 to 400kHz for voltages in the mid forties....Question - can I test this on  a bread board before I go PCB cost and if so, which one would you recommend.  I am aware of your evaluation board as well - so I do want to order that but with more tap points so that I can get a better understanding...

Hello - Will the set up above provide the fright components for our set up of 0 to Vin and -43 to agnd and pgnd with a Vout of -28?

Hi Guillermo,

Guillermo Aldana said:

Driver signals - you mean HO and LO? those are gate signals to the MOSFET

Correct, I was referring to the HO and LO signals.
For switching regulators, it is no problem to place probes there. The device has 3A gate driver strength, so the controller doesn't mind if a little charge is drained.

 If you place a probe on the switch node (between inductor, high side FET and low side FET), you essentially measure the drain-source voltage of the low side MOSFET, so it works the same way as measuring the LO signal output.
Both ways are valid to measure the switching signal and the according switching frequency. It is more about convenience where you have a better place on the PCB to place the probe.

Guillermo Aldana said:

I chose 300kHz because somewhere in your schematic, I read this thing is optimized for 300 to 400kHz for voltages in the mid forties

The device works equally good over the full frequency setting range. The final frequency depends on customer preference.
Lower frequencies mean lower switching losses, but larger inductance requirements to keep the VOUT ripple on the same level.
Some other customers need to avoid the AM/FM band, so they need to stay below 600kHz or go directly to 2MHz, and so.
I see no problem with a 300kHz selection.

Guillermo Aldana said:

I am aware of your evaluation board as well

I would definitely recommend to make some initial test with an EVM. Due to the switching behavior, the power stage can be a strong source of noise, so good layout is crucial for a stable system. (If you use a blank breadboard to build a boost converter, there would be so much noise and voltage overshoots that the design is basically unusable, as the jumping up and down of the switch node will work like an antenna.)

We have two EVM boards for LM5122. One with one phase and one with two phases interleaved.
Which one is better suited depends on the output power requirements of your design. 300W or less can normally handled by one phase, but for higher power levels we generally recommend a multi phase design to get thermals better under control.
This is the one phase EVM:
https://www.ti.com/tool/LM5122EVM-1PH

You need to see if this boards can be modified to better reflect your specs. This is something I cannot assess with the limited information. We do have a lot of additional tools for power design calculations and app notes though:
A general power stage calculation tool:
https://www.ti.com/tool/POWERSTAGE-DESIGNER

App notes on negative input, negative output DC/DC design:
https://www.ti.com/lit/an/slvaf68/slvaf68.pdf

Maybe there is something useful for you.
Best regards,
Niklas

Thank you for your input Niklas, I will inquire to understand if the EVB you mentioned can be bought unpopulated, so that we can configure it to what we need.  I am looking to understand what you would recommend as a "current load" demand circuit - that is a way to test my booster / buck controller for a given demand - I know a variable resistor of course is the obvious, but I am looking to include transients which mimick our current demand, which are low BTW, 1 1.5 amps - which is another point of concern, as the LM5122 is a CCM mode - but this seems like a low power application, meaning it will go into DCM which is my understanding be less efficient (but it does it automatically - which is good).  Is there another booster controller you can recommend which is designed for such an application? You mentioned the LM5125, with high switching freuquencies, this would allow less ritpple in the current but because I don't know my transient current demand - that is how quickly my load demands and shut down current, then its difficult me to gauge. Any additoinal advantages to the LM5125?

Hi Guillermo,

For 1.5A load, a one phase design should be fully sufficient.
For running load transient tests on the board, you can use active (electronic) loads, or passive (resistive) loads, both works.
The voltage over- and undershoots of the output depend on the loop speed and stability. With higher switching frequency, you can achieve faster regulation loops, but a fast and stable regulation can also be achieved at lower switching frequencies.
The quickstart calculation tool I mentioned in a previous reply comes with a bode plot calculation, which shows the expected loop speed and stability.

When it comes to CCM and DCM behavior, I would not say that DCM is automatically lower efficiency. It is rather the output voltage ripple which is generally higher in DCM operation. LM5122 allows FPWM (forced pulse width modulation), meaning it will always operate in CCM mode to reduce output voltage ripple. However, this will actually reduce efficiency a lot, as the current is just shifted back and forth between input and output side. So DCM is actually better for light load efficiency.

LM5125 is our newest generation for synchronous boost controllers. Features are mostly similar to LM5122, only the production technology is our latest and greatest, meaning lower Iq, improved adjustable dead-time, and spread-spectrum Fsw dithering for better EMI behavior. LM5125 is a dual phase controller, but the according single phase device (LM5126) will also be out within the next weeks.

Best regards,
Niklas

Niklas, Thank you for the feedback.  I have inquired about multiple options for your EVB - I will look into the LM5125 - dual phase, meaning internally? I am interested in the dynamic loads - AN-1733 Load Transient Testing Simplified This note eludes to a circuit that mimics my interest.  I would like to test the legacy and potential future design with a load that ramps up fast from zero to I, maintatins and then drops like  brick, followed by a train pulse of currents...I have found some hardware such as the Rigol DL3000 series, but if you have other or better recommendations, I am all ears - I am new to all of this..