• State TI Thinks Resolved
  • Locked Locked
  • Replies 10 replies
  • Answers 3 answers
  • Subscribers 63 subscribers
  • Views 1328 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Original question:

LM5116: LM5145- Using Synchronous Buck DC-DC Controller design variable/Programmable wide rage Vout

LM5145: Solution for multiple high power +24V outputs

Chris van der Aar
Intellectual 570 points
Part Number: LM5145
Other Parts Discussed in Thread: CSD18563Q5A, LM5116, CSD18531Q5A, LM5146-Q1

For a new project for a customer we want to make a design where we need 3* +24V/15A and 1* +24V /6A out of incoming 230Vac per unit (each machine will contain 42 of these units).

Therefore we are thinking of a Vicor unregulated 1750W converter (BCM6123TD1E5135T00) to create +40Vdc out of +320V dc (=rectified 230Vac in) and using 4* LM5145 to create max +24V/15A max at each output. This is purely based on very worst conditions where we think in practice this will never occur but in theory we have to take account with it. It may also be another converter but I think the LM5145 looks the best solution for us at this moment (also wrt costs)

I already did a simulation with the LM5145 set at +24V output and this showes that the impact of a voltage change at the input has not much effect on the output voltage voltage which is important for us.


Still some questions when using this solution:

1) Do you think this solution will work? Of course the input voltage can vary because the converter is a non regulated version (1:8) so each input change (e.g. ripple) will be visible at the output of the Vicor but it looks the LM5145 could handle over a wide input voltage range and simulation shows not miuch impact at a voltage change from +30V to +50V at the +24V output.

2) The Excel calc sheet shows 350uF cap at the output with an extreme low ESR of a few mOhm. The very best cap I can find is already 13mOhm. Or is it OK when we put a ceramic cap parallel with the big cap at the output to reduce ESR? Or has the ESR much impact on the output ripple? The Excel sheet shows 35* 10uF cap in parallel (I think for the low ESR) but this looks not very practical for the final design...At your eval board I see a lot of 47uF ceramic caps in parallel with the output elco but I guess these will be not cheap for higher voltages (+35V or higher) 

3) Are there some special precautions required with a load of 15A at +24V? The Excel sheet shows a power loss of 4W for the upper MOSFET at 15A load. With a thermal resistance of 50C/W this would result in 200C temp rise. We have a 32 degrees coldplate for the converter so maybe we can find out if we can manage to connect the FET to the coldplate? Or are there MOSFETs with lower thermal resistance with similar specs so the temp rise will be less? The dissipation in the coil and lower MOSFET is lower according the datasheet so probably not a problem..

4) We need 4 of these converters at one PCB all connected to the same output of the high power 1750W converter. My idea is to sync one converter with the neighbour to use shifter clock and maybe running the third one at a different clock speed and sync this one to the fourth so we have 4 converters running at different clock edges and frequencies to minimize interference and EMI. We have to deal with EMI Standard class A (we prefer class B when possible)

5)  T.I. has an evaluation board with the LM5145 but with +5V output. Do you think we could use this for +24V (out of +40V in) and by changing some component values modify it for +24V output? The inductor (3.3uH) and MOSFETs look the same as I get as suggested values from the Excel calculation sheet at +24V output. So when we could change this by only changing some resistors it would be nice.

6) At your EVAL board I see a large Schottky diode in parallel with the bottom FET. However in webench designer and in the Excel sheet schematics I don't see this diode. Is it required for the design or is there a special reason that it is not shown at the Excel sheet?

I hope you can help us with these questions so we can continue with investigating the final solution

Best regards

Chris van der Aar

Sr Hardware Engineer

NTS Systems Development

Eindhoven, The Netherlands

  • 0
    TI__Guru 74295 points
    I think that the LM5145 should work
    Using a mix of ceramic and Al. output cap is a good way to get the low ESR.
    The FETs will definitely need some heat sinking. You will need to look at this carefully.
    Synchronizing the converters as you mention is a good idea.
    I would use the EVM from TI and modify it to allow you to experiment and check the design before production.
    The Schottky diode may not be needed, however it is good to leave a place for it on the PCB.
  • 0
    TI__Guru* 76022 points

    Hi Chris,

    It appears you can use 60V NexFETs for this application - for example, CSD18563Q5A - for both high-side and low-side MOSFETs. Please send the Excel quickstart file for review as the power dissipation in a 5 x 6mm MOSFET package generally should not exceed 1.5W (depending on max ambient temperature and airflow).
     
    The antiparallel Schottky diode only has incremental benefit as the LM5145's deadtime is very short (15ns). 30-70uF output cap is sufficient (e.g. 5pcs of 10uF/50V/1210/X7R), and the inductor should be selected to provide 40% ripple curertn at Vin-nom.

    Regards,
    Tim

  • 0 in reply to Timothy Hegarty
    Intellectual 570 points

    Hi Tim,

    Today I tried the simulation again with the LM5145 in Webench Designer. I set the input voltage between +35V and +50V (we have around +40Vdc nom.) and the output voltage at +24V/15A

    However with these values the simulator tells me that the FET junction temperatuyre is above the max rating and the simulator places an ideal FET. See message below. I can not remember that I saw this a few days ago? Do you know if there has been something changed in Webench or why I get this message? I thought it should be possible to get 15A @+24Vdc with this converter (we have the possibility to mount the FET at a cold plate). When I reduce the current to 5A or less then it selects a certain MOSFET.

    With the design conditions, either the IC or the selected FET junction temperature is exceeded above the maximum rating. Hence, this design is created using an ideal FET. Please note that the resulting FET parameters are ideal, so the efficiency/loss opvals have been disabled. Also, the schematic/PCB export and Thermal simulations will not work with the ideal FET.

  • 0 in reply to Chris van der Aar
    Intellectual 570 points

    Hii Timothy!


    (see my previous question concerning Webench above)

    Today I tried the MOSFET Power Loss Calculator from your website. However the LM5145 was not in the list but I tried with the LM5116. When I fill in my requirements (40Vdc in, +24Vdc/15A max out) the high side MOSFET shows a dissipation around 7.5W when switching at 500kHz. With Rja=50C/W this would mean that the junction temperature rises far above the limit. So maybe that is the reason why Webench does not show candidates for MOSFETs with these settings for the LM5145?

    When I reduce the switching frequency to 100KHz (larger inductor value) the power loss will be around 2.2W so we are wondering if we have to choose a higher inductance value here (and lower switch frequency) or do you think it is possible to use lower inductor values and higher switch frequency here? Or can we use MOSFETs with a lower thermal resistance here?

    Regards Chris

  • 0 in reply to Timothy Hegarty
    Intellectual 570 points

    Here I send you the data I used for our application with the LM5145 and CSD18563Q5A MOSFETs. The calculator shows a lot of influence on the power loss by Rdson but also by the gate charge value. There are some MOSFETs with extreme low Rdson but most of them have a higher gate charge or a very high input capacitance. The calculator show a power loss of 1.9W at 200kHz switching frequency and 24V/15A load so probably the junction temperature could be at the limit in that case (i calculated with a Rdson of 8mOhm because I guess the case temp will be higher so Rdson will be higher) but we have the possibility of using our cold plate that we already use for cooling our DC/DC converter for converting the rectified net supply to 40Vdc for the LM5145 so probably the case temperature will be not too high for the high side MOSFET. The lower MOSFET shows around 1W loss at max load so I think not a big problem. Besides this is all based on a very worst case condition in our case which probably  will never occur (but we have to take account for it).

    I canreduce the power loss a little bit by lowering the switching frequency to 100kHz but in that case I need a coil with a higher inductance value. With 200kHz I can use 10uH (for a ripple of 4.8A) and looks acceptable. I could go up to 500kHz and use 3.3uH but in that case the power loss in the high side MOSFET will be very high so I think this would be an acceptable option . Especially when we are able to cool down the MOSFET by our cold plate (active water cooling).

    NOTE: I was not able to insert the original Excel file here so I generated a pdf.

    LM5145 Wide Vin Synchronous Buck Controller Quickstart Calculator.pdf

  • 0 in reply to Chris van der Aar
    TI__Guru* 76022 points
    Hi Chris,

    Thanks for sending the file.

    Try the CSD18531Q5A 60V NexFET for the high-side as it has higher transconductance and will slew the current faster during the switching commutation. Also, use the Rds(on) value at Vgs = 7.5V (i.e. the VCC level of the LM5145).

    Regards,
    Tim

    http://www.ti.com/widevin
  • 0 in reply to Timothy Hegarty
    Intellectual 570 points

    Hi Timothy,

    Still a question concerning PCB layout for the LM5145. In your datasheet and your evaluation board the suggestion is to keep the power loop as short as possible. This means that you need a PCB with components placed at both sides so the converter chip will be close to the inductor and capacitors at the other side of the PCB.

    However this will increase the costs for our PCB and at this moment there is no real need for 2 side component placement except for  the DC/DC converters.

    Do you think it is possible to keep the converter and inductor and caps and other small SMD components all at one side of the PCB and try to keep the power loop as short as possible? Or is this not something you would advise? I don't know if you have experiences with these types of converters and all components at one side of the PCB? And if you have an example for this is it possible that i can check the layout for this? Otherwise I will follow the layout rules in your datasheet and evaluation board but that means 2 side component placement.

    Another question concerning output caps. One of our component suppliers said that the availability of ceramic caps with large values is still a problem worldwide and therefore they suggested to use polymer caps instead of ceramic. Besides polymer caps have the advantage that the capacity will not reduce at higher voltages like ceramic caps. Do you have experiences with polymer caps in combination with the LM5145 or simular converters and do you see some improvements regarding ripple etc?

    Regards


    Chris van der Aar

    NTS Systems Development

    Eindhoven the Netherlands

  • 0 in reply to Chris van der Aar
    TI__Guru 74295 points
    I will close this due to inactivity
  • 0 in reply to Frank De Stasi
    Intellectual 570 points
    Please could you answer my latest question concerning PCB layout for the LM5145? I was wondering if it would be possible to keep all components for the converter (including inductor and MOSFETs) at one side of the PCB? Otherwise I have to make a PCB with 2 sided SMD components only for this converter part. I understand that the power loop has to be short as possible so I can imagine that you want to place the converter as close as possible to the MOSFETs and inductor and in theory this looks only possible by placing the converter at the other side of the PCB. But I was wondering if their are examples of a PCB layout with all components at 1 side with this type of converter.
  • 0 in reply to Chris van der Aar
    TI__Guru* 76022 points
    Hi Chris,

    Take a look at the LM5146-Q1 EVM layout as an example of a single-sided layout: www.ti.com/.../lm5146-q1-evm12v

    In fact, keeping all power stage components on the top side is preferable as it has benefits in terms of EMI.

    Also, using polymer caps instead of ceramic shouldn't be a problem. Just keep in mind the ESR value and thus the ESR zero frequency - this is cancelled by a compensator pole (typically the RC network across the upper feedback resistor). See the datasheet for related detail on compensation.

    Regards,
    Tim