Electronics > Projects, Designs, and Technical Stuff

LTC3895 step down converter burns off

<< < (5/6) > >>

Srijal97:

--- Quote from: Srijal97 on January 14, 2019, 04:43:36 pm ---I've attached the new design with this post.

--- End quote ---

So I noticed that the board pdf does not include the name of nets so I exported high-resolution images in Eagle and have them attached in this post. Sorry!

mvs:

--- Quote from: Srijal97 on January 14, 2019, 04:43:36 pm ---It would be great if I can get some improvements and comments before I place an order for manufacture, thank you very much for your time!

--- End quote ---
Sense+ pin current is +/-1µA, resistance of your current sense divider 277K (280K||10Meg), this gives you current sensing error of +/-277mV.
Current sense threshold is around 50mV (43-57mV), so guess why some of your ICs are blown and some do not start.
Reduce resistance of your divider by factor 100 and increase capacitor C9 to 1µF.

Next point are caps. You are trying to build 20A supply, there will be some ripple current, usually around 30% or 6A, perhaps even more. I have never seen a 22µF aluminum electrolytic cap that has ripple current rating of more then 0.5A. Look for low ESR caps, their typical capacitance range is 100-470µF and use more then one to split the current load. Put caps both on input and output.

Ceramic output caps look also quite small. Are they in 0805 package? Why? Either they have very low capacitance, or they will loose most of their capacity with bias voltage above 15V. Go for the larger footprint (at least 1210), look for better material (X7R, X7S or at least X5R). Some caps may have "capacitance change vs DC bias voltage" diagram in datasheet, like Samsung CL32B475KBJVPN, so you can compare their capacitance at desired 32V output voltage.

And last but not least - please do not order PCBs with black soldermask for prototypes. It is very difficult to inspect and rework them.

capt bullshot:
Some of your component values look wrong:
C10 / C11 - typically one finds here a few pF or nothing
R8/R7 - The datasheet states this resistor should be in the 2k ballpark. What's that R9? Looks pointless.

You need an electrolytic input capacitor additionally to your ceramic ones. Placing an 100R in series with an 1u ceramic is pretty pointless, a typical value would be 0.1R to 0.5R, but this is not required if you have an electrolytic. Some 200uF are required here.

Did you check INTVCC with your prototype?

Srijal97:

--- Quote from: mvs on January 14, 2019, 08:04:00 pm ---Sense+ pin current is +/-1µA, resistance of your current sense divider 277K (280K||10Meg), this gives you current sensing error of +/-277mV.
Current sense threshold is around 50mV (43-57mV), so guess why some of your ICs are blown and some do not start.
Reduce resistance of your divider by factor 100 and increase capacitor C9 to 1µF.

--- End quote ---


--- Quote from: capt bullshot on January 15, 2019, 06:53:43 am ---R8/R7 - The datasheet states this resistor should be in the 2k ballpark. What's that R9? Looks pointless.

--- End quote ---

Okay yes, we messed up. We did not quite understand the current sense threshold and +/-1uA parameter and were happy to see the large values of resistors having low power dissipation allowing us to use 0805s. Now that it was pointed out(Thank you!), we spent some time figuring it out and it makes sense. Considering C9 as 1uF, R1||R2 comes out around 2.9k and so R1 is 3.9k with 0.8W max dissipation across it. I'll be adding a through hole with 1-2W rating and a vertical fold to save space. Also, R2 comes out to be 12k (which is R9 in my schematic, with earlier value 10M), so we need to put that in right? Or can you directly connect R1||R2 in place of R1, without R2? I have attached the calculations done from the datasheet, just in case.


--- Quote from: capt bullshot on January 15, 2019, 06:53:43 am ---C10 / C11 - typically one finds here a few pF or nothing

--- End quote ---
Those were the values suggested by the simulation...should I change those?

Regarding the input caps, we hesitated with electrolytic ones because of the drastic loss in capacitance at frequencies of 800-900kHz...but I guess adding large values >100uF would get the effective value to the required range.

And for the output, the required value came out to be very low and we did see DC bias graphs for KEMET MLCCs, so based on that we decided to go with 0805s with some of them rated up to 50VDC. But we'll change it to a larger footprint to get more options with the required voltage ratings and dielectric choices as suggested.


--- Quote from: capt bullshot on January 15, 2019, 06:53:43 am ---Did you check INTVCC with your prototype?

--- End quote ---

Not yet, and it did cross out mind after the tests. We'll do it again tomorrow.

Other than that, do the GND planes, switching node and overall layout look decent? Thank you very much for your support!

capt bullshot:
The numbers for C9 / R8 / R9 look plausible to me now.
Layout looks OK now.

C10/C11 - your arrangement looks unusual to me (but I didn't verify anything here), the datasheet suggests to not have C11 and treat C10 as optional. So one might first try without these.

The inductor value of 4.7uH is rather low, expect large ripple and higher losses. In general your circuit looks quite tiny for your power goal (7A / 32V, up to 140V input). This is 224W output power, and optimistically assuming some 94% efficiency (may be better at low input voltage, worse at high input), your circuit will dissipate about 13W.

So I'd recommend you to slowly increase input voltage / load and watch the power components temperatures.

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod