Buck driver design advice needed

How did you calculated this 21.5 mOhm?

I do not understand it, can you explain it?

It’s a simple omhs law calculation

I know that 200mV/21.5mOhm= 9.3A
How did we get the 21.5 mOhm?

You do the opposite:

0,200V/9A = 22,2mOhms.

OK. There is a misunderstanding.

There are 2 current sensing resistors on my drawing. Both of them are 43 mOhm. But I use only one of them. They have different sizes, I only have the option to use 1206 or 2512.Only for flexibility… And it is also written on the drawing, 1206 or 2512.
Lexel calculated the equivalant resistance of parallel resistors and he got 21.5 mOhm and from this 9A.

But the current should be only 200mV/43mOhm= 4.6A
But it could go higher and was not constant as I changed the input voltage.

If your output gets higher with more input voltage there is definately something wrong with the switching frequency or current sensing
the buck has to stay stable in output

I also see no 1uF directly connected to the buck input Vcc that is described in the datasheet
with 3.3uH I calculated 200kHz
also the driver design tool points more to 4.7uH

also too low input caps can cause a drop, but not a increase with voltage over the current that the sense resistor should give

also an output cap is possible to reduce noise

I have read the datasheet of the evaluation kit you linked.
Evaluation kit 16820
I can not see any differences between this and my schematics.

My design
16820 EV Kit

According to the datasheet, C1 and C3 are optional.
“Typically, capacitor C1 is not required if the power supply is relatively close to the EV kit. If long wires are used
to connect the power supply to the EV kit, install up to
10µF of bulk capacitance at the surface-mount 2220
pads provided for C1.”

“Typically, the LED ripple current equals the inductor ripple current. To reduce the LED ripple current, install
optional output capacitor C3. The EV kit provides surfacemount 0603 pads for a capacitor nominal value of 0.1µF.”

The input capacitor you did not see, is there. This is C4 1uF capacitor on my schematics.

I recalculated the design with Maximintegrated’s design tool:
Design tool

Changing the switching frequeny in the excel sheet does not really influence the L value calculated by it .
And you can also change the input voltage in quite a big range without influencing the L value.

The only new info I got from the eval kit datesheet is:

_To enter hysteretic mode, the following input requirements must be met: set VIN above 5V or above VFLED +
4V (whichever is greater), and provide 1.1A of input current to the EV kit. If low-voltage or low-current input conditions fail to meet the input requirements for hysteretic
mode, the MAX16820 controller operates in linear
mode providing DC current to the LED load._

The excel sheet does not have any warnings about it. According to it the design is feasible.

So I think it means that for a 6V led I need at least 6+4=10V input voltage. Othervise the driver will operates in linear mode. Most probably that is why my output current changed as I changed the input voltage. But if I make a simple calculation: 6V*4,8A=28,8W this is the power consumption of my XHP70. If I increase the input voltage to 10V, the input current would be around 2,8A.
If I increased the output voltage of my power supply over 7,4V (simulating a 2 cell lipo) the current was also increasing until my power supply limited it at around 5A.

So I still do not undertand what the problem is.

the tool say you only if the switching frequency is working or not

If you look on the switching frequency diagrams the linear mode is not included, this is why I guessed its not supportet, sure any buck will fall out of regulation at a point,
but as I pointed out its not a specified low dropout like the LM3409 which will fall out of regulation when the Input voltage is Output + losses, you said it right its likely 4V
I simply had not the time to read all datasheet but the diagrams tell me wehats going on on normal operation

At this point I would simply go with a lower shunt like 100mOhm to see if its working properly right there and which current you get with it

The XHP70 should use at 4.8A 6.6V, but also shunt resistor, MOSFET, PCB, MCPCB, Digital Multimeter and wires ect. increase the real voltage so this will be more like 31W
Power measurements are only valid if you use a clamp meter and measure the voltage over the driver oputput terminals not the LED board

Hi,

I still have problem with my design.
I also contacted maxim’s support and got the following answer:

_I tested on the EVKIT and changed 2 serial LEDs about 5V. When VIN is in the range 5V~20V, the LED current is basically stable and won’t go too high.
For the inductance, the formula V=L(di/dt). The inductance must be higher than the calculated value.
And also for the di, 1A LED current has about 0.3A di, 4.8A LED current has about 1.4A~1.5A di. But the EVKIT assembled 56uH inductor, so for sliding scale, 4.8A LED current also need a inductor bigger than 10uH._

I made a new driver with a 10uH inductor that was suggested by the support, but it did not work properly. The current is still not stable:

I also made some measuremets at the point indicated on the last picture of the link below:
https://imgur.com/a/hYBgiya

1. measurement at D2 at 25%pwm dimming signal
2. measurement at D2 at 50%pwm dimming signal
3. 4. and 5. measurement at D2 at 100%pwm dimming signal

My measurement (if it is correct) shows 3 MHz switching frequency?
Of course the driver is still to hot.

Shall I change to LM3409? Can it be better for my configuration (xhp 50 , xhp 70; 2 and 3 cell drivers)

on page 7 of the datasheet is written how the switching frequency can be calculated
if they sufggest delta I 1.5A, then with R=43mOhm you get 65mV Ripple voltage over the resistor

so calculating with Vin 3S 12V to Vout 7V
f sw=(12V-7V)*7V*0.043Ohm)/(12V*0.065V*0.0000047H)
I get with 4.7uH 410.5kHz switching frequency

Are you sure on your circuit no bad things happen like voltage spikes coming from your bench power supply
Had you run it on a battery pack?

I would modify only one thing in your calculation, they also suggested to use at least 10 uH inductor.

According to it:
fsw=(12V-7V)*7V*0.043Ohm)/(12V*0.065V*0.00001H)
It would result 192,9 kHz

I connected it to a 8V battery and measured the LED current with a clamp meter, at 100% pwm and it was around 4,5A The inductor, fet and current sense resistor got quite hot after a while.

Than I connected it to a 3s lipo (I have 3 modes, 25,50 and 100% pwm dimming) at 50% pwm something got smoking hot, so I disconnected it immediately, did not want to kill the driver.

I do not know what happens inside the my circuit but based on this test I do not think that my power supply cause it.

I'am thinking about to try LM3409 and made a calculator.

http://www.filedropper.com/lm3409calculatorv1

Not tested!

I have designed a new driver based on LM3409

https://imgur.com/a/3hnAzgB

R_dynamic (ohm) 0,12 ohm LED dynamic resistance (5,9375-5,375)/(4,8-0)
V_input 11,1 V Input voltage (min 6V)
V_input_max 12,6 V Max. input voltage (max. 42V for LM3409 and 75 for LM3409HV)
V_output 6 V (n*LED Vf)
R_led_dynamic 0,12 ohm LED dynamic resistance
f_sw * 500 kHz target switching frequency
I_led * 4,8 A target LED current
Di_L-PP * 1 A target inductor current ripple
Di_LED-PP * 1 A target LED current ripple
Dv_in-PP 0,5 V Input voltage ripple, 2%-10% of V_input
V_turn-on * 8,4 V target turn on threshold
V_HYS * 0,6 V expected hysteresis
n 0,95 x100% Expected efficiency
C_off 470 pF OFF time capacitor (470pF to 1 nF)
R_off * 7,6 kohm OFF time capacitor
R_off 7,59 kohm OFF time capacitor selected (closest value)
t_off 861 ns OFF time
f_sw 500,6 kHz switching frequency
L1 * 5,2 uH inductor
L1 4,7 uH selected inductor (closest value)
Di_L-PP 1,099 A Inductor current ripple
I_L-max 5,35 A max. inductor current
V_ADJ 1,24 V The IADJ pin is left open forcing V_ADJ = 1.24 V
R_sns * 0,046 ohm calculated sense resistor
R_sns 0,043 ohm selected sense resistor (closest value)
I_led 5,22 A LED current
t_on 1,14 us ON time
C_in-min 11,86 uF min. input capacitor
C_in 23,72 uF min. input capacitor
C_in1 = C_in2 .... 10,0 uF selected input capacitors C_in/number of parallel capacitors
I_in-rms 2584 mA input current RMS
V_T_max 12,6 V Q1 min.voltage
I_T 2,97 A PFET current
R_ds-on 18 mohm PFET Rdson
I_T-rms 3,94 A PFET current RMS
P_T 280 mW PFET power dissipation
V_D-max 12,6 V D1 Schottky diode min. voltage
I_d 2249 mA D1 Schottky diode current
V_d 470 mV D1 Schottky diode forward voltage
P_d 1057 mW D1 Schottky diode power dissipation
R_uv2 * 27,3 kohm calculated R_uv2
R_uv2 27 kohm selected R_uv2 (closest value)
V_HYS 0,6 V hysteresis
R_uv1 * 4,68 kohm calculated R_uv1
R_uv1 4,70 kohm selected R_uv1 (closest value)
V_turn-on 8,4 V turn on threshold

Selected components
C_F (*1) 1 uF
C_off 470 pF
R_off 7,59 kohm
L1 5 uH
R_sns 0,043 ohm
C_in_1 10,0 uF
C_in_2 10,0 uF
R_uv1 4,70 kohm
R_uv2 27 kohm

I would really appreciate any comments and suggestions.

Thanks

Unless there are vias under pads that I can’t see (shouldn’t be done anyways), there’s a few things that need attention in your layout -

- Check your MOSFET connections again, I think you got them backwards (D & S are backwards in a P-FET)

- Does the high current path from BAT+ to FET “drain” (see above) go through a single via and a thin trace? You got plenty of space there and can beef that up A LOT; since it can’t sink heat into the GND plane you have to max out that copper area

- the LM3409’s power pad should be connected to GND, same story (heat)

- I’d clean up all those little copper islands on GND

- the TPS709 datasheet says: EN can be left floating for operation; if pulled high, it should stay below 6.5V (7V absolute max) :open_mouth: :confounded:
(you could use a smaller 3-pin LDO anyway)

  • You could also add a voltage divider to an ADC pin for batt voltage readout, maybe shared with the LM3409

edit - I missed the biggest one - you can make use of the 3409 dimming function, look up the GT driver. (Dimming via the EN line is still needed for <25% or so)
Can’t find a schematic right now, I think this here is basically the same, just using different ATtiny pins

Thank you vary much for the comments. I have modified the circuit and the pcb design according to them.

This is the new version.

https://imgur.com/a/QvVo2ZT

I do not understand this:

edit - I missed the biggest one - you can make use of the 3409 dimming function, look up the GT driver. (Dimming via the EN line is still needed for <25% or so) Can't find a schematic right now, I think "this here":https://budgetlightforum.com/t/-/41130 is basically the same, just using different ATtiny pins [/quote]

I also checked the schematics you linked but there are also a few things I do not understand. Both of the EN and Iadj pins are used. Why?

And there is something that is not clear from the LM3409 datasheet:

I wanted to use the EN pin for dimming. Attiny13's output is a 5V, 500Hz pwm signal.

LM3409 needs at least 1.74V on the EN pin to be enabled. Does it means that if I have a 25%pwm signal LM3409 does not enabled at all?

1,74V/5V*100-->34%

So do my pwm duty has to be between 34% and 100%?

And what happens if I connnect a voltage divider between the mcu output and gnd to get a max 1.24V output and use the analog output function of the mcu to dimm the LM3409?

The link to the calculation is not working
I would compare it to my own calclation table

I found one starnge thing with the LM3409 calculations all driver seem to run with about 50% more switching frequency than calculated

I build all my drivery with 3-4 times of C min from the calculations
XHP70.2 has about 6.3V output voltage plus a bit on wires on your curent
You need to calculate with a reasonable voltage drop from the batteries, so use 4V or 3.9V per cell on your calculation

the P-FET power loss is more than just the ohmic loss I could give you a calculation from TI, not sure if your choice of FET is good

you should also use a separate logic ground thats starting point id on the LM3409 leg containing full LDO and MCU path as well as all LM3409 related parts, I had problems not doing so with th LDO freaking out because of a ground loop that acted as receiver of the switching node
also do a low pass on Vin to the LDO with 10Ohm and 4.7uF

also put in an output cap of 1uF per A

using just the EN pin on the LM3409 will have you in all brightness levels PWM which is bad, especially at that low 500Hz frequency

use here I adjust over a low pass filter to set an analog signal for it, and only PWM for very low modes

PWM is PWM it has always 5V when enabled dont worry for low PWM value “average voltage” the signal is always 5V strong

Now the FET connections (D & S) are correct in the schematic but wrong in the layout. (Sorry, I think I was a bit unclear in my previous post.) Now you have also in effect shorted its source to GND (it is not an isolated pad!)
In short:
#Source (pin 5,6,7,8 + pad) goes to BAT+ and this copper area has to be maxed out.
#Drain (pin 1,2,3) goes to the node between L1 and cathode of D2.
#Q2 has no physical connection to GND.

As a reminder (CSD25402, other P-FETs in the same package have different pinouts):

I think you pad clearances to GND could be a lot bigger, you’ll have to check with the PCB manufacturer what’s the permitted minimum

Regarding the filtered PWM dimming via the I_adj pin, here’s a TI app note that describes it neatly. In the pdf there’s a link to an Excel spreadsheet that’s very handy for calculations.

at least the PWM + analog part, the actual dimming in their example is in reverse

The ATtiny (like many other 8-bit microcontrollers) has no analog output function a.k.a DAC. Hence the filtered PWM signal which is in effect an analog voltage.

I can totally back this up I was sloppy having a FET changed to more effective one and suddenly it acted just as a diode
In the end I could scrap a whole sert of produced PCBs and reverse drain and source on new design

Always look in the datasheet where the pins are connected internally

also placing ground viases around not ground connected copper areas that have heat sources on it nas having also viases on the outer edge of that area helps

Thanks for the suggestions, I will try to interpret them, but most probably I will ask again.
Until that I have aploaded my calculator here again:

LM3409calculator