Having proportional unequal currents is definitely possible with the load share controller simply by designing each load share controller circuit to have equal CSO voltages at the desired current level. The CSO voltage sets the LS bus voltage; this LS voltage level is common to all of the load share controllers and is an input to all of the controllers' internal error amplifiers. As long as you provide the appropriate gain around teach of the controllers' current sens amplifiers the circuit should adjust the current in the desired proportion. Just make sure the TPS5430 is appropriate for load sharing. The converter needs to be able to source current but not sink current from the output.
Lisa Dinwoodie
Dear Lisa,
Are there any application notes for this user scenario?
Thanks for your attention!
I am not aware of any application notes that specifically address this. I believe there was reference to it in one of the Unitrode Power Supply Design Seminars (in either the 1997 or 1991 seminar books).
Thanks for your help!
Lisa,
The load share controller UCC29002 only work for Vout > 4V, since the internal non-inverting op-amp will clamp at 3V (EAO pin), ADJ has to be at least 4V to support the NPN circuit.
Is there a solution or alternative controller that can work for Vout = 1.2V?
Thanks,
Tim
The attachment shows how the UCC39002 was implemented on a 0.95V output rail by using a seperate -5V supply for the load controller GND reference and a seperate 8V supply for bias.
This is great! The DC/DC converter module that I'm planning to parrallel them up is the PKM4118L (Isolated 48Vin, 1.2Vout @ 100A), what is the best way to generate this -5V supply from these voltages (48Vin, 1.2Vout)?
Since 2x PKM4118C in parrallel, output current can be up to ~ 200Amps, is there any limit or issue of using the UCC39002?. I wouldn't think so but would like your comment.
PKM4118L Series datsheets: http://archive.ericsson.net/service/internet/picov/get?DocNo=28701-EN/LZT146397&Lang=EN&HighestFree=Y Thanks, Tim
PKM4118L Series datsheets: http://archive.ericsson.net/service/internet/picov/get?DocNo=28701-EN/LZT146397&Lang=EN&HighestFree=Y
One of the most common limitations is whether there is a resistor from the +Sense to the +Vout in the module that is less than the required ADJ resistor. This internal resistor is essentially going to be in parallel with the ADJ resistor and if calculations say you need 85 Ohms for load sharing adjustment range, and there's a 2 Ohm resistor therethen you will be hardpressed to put a resistor in p[aralllel with 2 Ohms to get a resultant 85 Ohms. But you are using an open frame module so that's nothing a soldering iron can't fix :-)
The other issue would be the adjustment range of the module. If the module can sink, as well as source current, then you would have to use a diode to block the sink current. The voltage drop across the diode will limit the available adjustment range of the load share and current share accuracy will be poor.
I would suggest a simple Flyback to generate the required bias voltages.
The internal resistor on +sense with respect to +Vout is ~ 100 Ohms, so the ADJ resistor value seems to be in a reasonable value.
What is the minimum voltage requirement for VDD (Pin 3)? Is it > 4.375V or in the range 13.5V ~ 15V?
What is the minimum regulation or tolerance requirement for the -5V bias as reference GND (Pin4)? I think this will affect how accurate load-share between the 2 paralleled modules is...
Thanks, Tim
Only after running the calculations can you determine whether the existing resistor between +Vout and +Sense is "reasonable". Each load share circuit is specific to the modules that are being shared as they depend upon the modules' loop response, adjustment range, internal resistance, load current, etc. It probably is okay, at least it isn't an exceedingly low value, so the odds are good.
VDD abs max is 15V, but it requires at least 4.575V for start up (always use the max value listed as any given device in any given lot may need up to this value for start up and still be a good device...process variations and component tolerances need to be considered). The higher the VDD voltage, the higher the CSO voltage can be (must be approximately 2V less than VDD) and the master's CSO voltage defines the LS bus voltage. Load sharing will be more accurate with higher CSO and LS voltages. So, 4.575V would be the minimum value but 10V, 12V, 13.5V, would be better. more operating range, more headroom...
The -5V reference should be as DC-like as possible as the internal thresholds are riding on it (25mV or so...). This is going to be the nice quiet "ground" for the load share controllers so that's the important point. One thing some customers have tried is to use the high current return path of thesystem load fo the UCC29002's GND and this just doesn't wor; the -SENSE Kelvin connection must be used becuause quiet is important.
Please note, when testing this, use the UCC29002/1 device...not the "non-/1" part. The response to transients and start up will be slower but less likely to oscillate. Just thought I'd throw that in there...
Lisa