Make VCC for TPS92075

Thanks for fast reply Tim!!

Tomorrow i try to calculate some main inductance value and auxiliary winding, i've tight space and board can receive SMD components only in bottom layer for cost reduction.

With axiliary winding in main inductor this controller is still compatible with triac dimmer? I've to maintain the retrofit compatibility with this kind of dimmers but also i've to design a specific dimmer that act in to ISNS signal, but this is my other task :)

How do you select Δil also for this point? Do you prefer CCM instead DCM for emi filter or other reason?

I don't know if the required inductance in CCM for my LED stack voltage and current that can be realized in a small custom part...

Can i put in a stand-by operation this controller by pull down the COFF pin by a NPN? I've to use less components count for space and BOM problem.

Thanks in advance, Luca.

Hi Luca, I will respond to your points above:
i) I have attached an example of a coupled inductor we have used to generate an AUX voltage. If you are not familiar with transformer design it is not something to easily describe in an E2E post. You could go to WURTH design services or other service with standard configurable parts. Or start with the part attached and simply adjust the secondary turns to suit your needs.

ii) There are two ways to configure for use with an AUX winding: (a) a linear regulator for start up (as in the normal EVM design: resistor/zener/N-FET) and set at 'X' volts for example, and then you configure an AUX winding to diode connect to the linear output at 'X'+2 to 3 volts, for example, to effectively turn the linear supply off when the AUX voltage is present.     (b) you do not have a linear but use a resistor only. At turn on, VCC charges through the resistor until VCC is above VCC UVLO. When the part turns on VCC UVLO must have a large enough hysteresis that the converter can run long enough to bring up the AUX winding and supply the VCC voltage.(in our case this would require a VCC cap of ~22uF)  This method eliminates the linear.  In each case as the output is dimmed, the LED voltage drops and then the AUX voltage drops.
Using method (a) is the most compatible with TRIAC dimming because as the output voltage drops the linear will again begin to supply the controller current and the part can continue to run.   in (b) there are two issues, becuase the AUX voltage will drop you must either live with a smaller dimming range (the output will shut off earlier) and because at turn on, if you are dimming, the input voltage is lower you may not turn on until the dimmer is moved higher in range to allow VCC to charge enough through the series resistance. 
So the best choice for dimming is to still rely on the linear or your range suffers and the dimmer range will have a lot of hysteresis. Meaning you will dim the light and then shut off, but then you will have to increase the dimmer a lot to start up again. For a dimming solution, operation with the supplemental linear is recommended. You increase the efficiency, but do not save the linear parts. In your case this may not be the best option if you're trying to save parts and space.  Some efficiency examples are shown below, for comparison. The 6W and 10W are the end solution output power. 'FET AUX' is the solution described above in (ii) (a)  and 'Resistor, AUX' is the solution described in (ii) (b).  'FET No AUX' is a solution similar to the EVM where to AUX winding is used.

iii) Δil with some DCM is possible with the TPS92075, you can see the inductor valley shown in the calculator tool. You can adjust your parameters so it is mainly DCM, but note that the tool will not accurately calculate the sense resistance value as it assumes the current is mainly CCM. CCM requires diode recovery, but has lower peak currents. Designing a TPS92075 soltuion that is constant DCM is really not possible. To maintain DCM in the center of the line cycle will require very deep DCM at the edges. The best trade off for this controller is ~1/3 of the cycle being DCM.

iv) You will have to compute your inductance and peak current and then start shopping for the smallest part.

v) Standby with COFF low: no, this will not work as the part has a maximum off-time. (the max-off timer ensures the part will start)  If you use the linear as in the EVM design (resistor-zener diode-NChannel FET that is providing the VCC bias) you can use a PNP to pull down the zener voltage and shut the linear off, disabling the part.  This also saves the losses in the linear as well and is a better approach.

Hope this helps...!

7065.750311736_0_coupled indcutor for bias.pdf

Thanks a lot Tim! Very useful and appreciate.

My final goal is a 160V LED stack with 120mA average current for standard 23W 150cm neon replacement with 230Vac 50Hz nominal input voltage.
I've just made the first prototype board that works very well at this point but now i need to re-design it for space and components saving.
Actually my board make VCC with simple linear regulator with FET, resistor, zener, diode and capacitor but i need to reduce this components... i think i've need to discuss about retrofit triac dimmer compatibility with commercial agents before continue...

My first idea it was to stop the device by shut down internal oscillator, like for example with UCC28070 or similar devices.
Can i stop the controller if i add a voltage signal, that come by external circuitry, summed with a diode in to ISNS pin?
I mean to add this signal  in to ISNS pin and the device decrease the output current, basically i've a dimmer effect, if I increase this signal until the controller enter continuously in OCP i can shut down the led? Then if i reduce this signal the led can turn on well again? If this solution is applicable i can save a P device in series with VCC and other little components so i can do a dedicated external dimmer that work with this method.

If i use a 220k for start up resistor with the AUX winding, can i use 10uF capacitor instead 22uF? Have you done some test before?

Can i attach the excel sheet, and schematic, with the values for my design in order to discuss togheter?
Sorry for my not perfect english.
Thanks in advance, also for your patience and time.

Regards, Luca

Hi Luca,  The designs are very different if you need to make it dimmer compatible or not. Non-triac dimmable is much simpler.  For non-triac dimmable you could use the resistive start-up with reasonable (~200ms to 300ms)  turn on delays and use the coupled inductor bias.
One way to increase efficiency and still use the linear is to power it from the bottom of the LED stack (LED -) Then you are not dropping the whole line voltage across the linear. With a high stack voltage like you have it will help. The trade off is that you need to increase the VCC cap from 0.1uF to 1uF up to 10uF to 22uF as the VCC cap will need to provide the controller current when the line is below the LED stack voltage.

There is not a way to shut down the oscillator in the TPS92075. The part was designed mainly with bulb applications in mind, so there was not really a need to disable switching as the input power on-off control would accomplish this. The main way to disable the part is to remove VCC. You could also do this with a series FET as another simple method.  

Yes, you can control the output current by adding current to the ISNS feedback, for example, if you wanted to do analog dimming or adjust the output current in some other way - but you cannot fully disable the output this way.  You could 'trick' the peak current sense the way you mention by forcing a lower signal on ISNS, like 0 to 300mV, for example, to keep the peak current sense from ever being reached.  This would stop switching, but the stopped state would be: switch ON! Which would not work. Forcing ISNS voltage higher just ends the ON-time and starts an off-time, creating a narrow pulse every ~280us. The current would be very low, but not zero. 

Yes, you could use a 220k and 22uF. The trade off is just a longer start-up time, but less drop on VCC when the part starts running. This would possibly allow fore more variation in your LED stack voltage. I believe we did use 22uF for some of our testing.

Yes, you can send a schematic and I can look at it. Thanks, Tim