Here is the third of three short tech nuggets, by Rich Nowakowski from TI, before our “Ask the Experts” session on Wednesday, July 23, at 1:00 p.m. ET (10:00 a.m. PT). Hopefully these brief blogs will help stir your curiosity and questions, which you can post on our chat site here either before or during the session.
Deep sub-micron process technology is impacting the power management solution of performance DSPs, FPGAs, and ASICs. Advanced processes require tighter core voltage regulation accuracy, even during load transients. Faster switching frequencies and newer control modes have been shown to help reduce voltage over- and under-shoots and during load transient events. Here is an article that discusses improving DC output voltage accuracy with proper feedback resistor selection, but perhaps at the expense of system efficiency.
Output-voltage accuracy
Using large feedback resistances to increase efficiency is common. However, choosing resistances that are too large affects the converter’s output-voltage accuracy because of leakage current going into the converter’s feedback pin. Figure 1 shows the current paths at the resistive feedback divider (R1 and R2). For a fixed feedback leakage current (IFB ), current through R1 (IR1) decreases as the values of R1 and R2 increase. Therefore, an increase in divider resistance means that a larger percentage of IR1 leaks into the feedback pin, and the current through R2 (IR2) decreases, causing a lower feedback-pin voltage (VFB ) than expected. Since VFB is compared to an internal reference voltage to set the output voltage, any inaccuracies in the feedback voltage create inaccuracies in the output voltage.
Equation 1 can be derived from Kirchhoff’s Current Law, showing VFB as a function of R1 and R2. Note that IFB is not fixed in a real system and can vary from device to device and over the operating conditions. To generate a worst-case estimate of the output-voltage change that is due to the leakage current, the specified maximum value of IFB is used in the calculations.
How important is output voltage accuracy in your application, and what other techniques or features can help improve accuracy?
“However, choosing resistances that are too large affects the converter's output-voltage accuracy because of leakage current going into the converter's feedback pin”
Steve, you are right about the leakage current and wastage in terms of heat. But is it possible to reduce such wastages; difficult.
@Steve: With large resistor networks, time constants will increase degrading the response time of the feedback network.
Steve,
From the figure I see that at 3.3V and total R1+R2=1K te VOUT starts dropping. This suggests that the current in thefeedback resistor divider is nearly 3.3mA. I think that is huge power loss? Am I missing something here?
“what other techniques or features can help improve accuracy”
@Steve: I think whereever the VFB goes to inside a chip can be a CMOS device, so that the input current (leakage current can be reduced), which can further help increase the R1+R2 resistance.
@amrutah—Right about large resistors slowing down the response time. It all depends upon the stray capacitcance of the PC board and the IC itself to make that RC combination
oops! Looks like I read it wrong, it is 1000K and not 1K.
@amrulah: Ohh a big difference indeed, isn't it ? Anyway still the price tag is quite high
@Steve: Does it have any direct links with the PC processor speed as well ?
“Does it have any direct links with the PC processor speed as well ? “
Chrish, PC processor speed depends up on many factors. Motherboard, RAM, Clock speed capability etc.
Computer processor might be changed by CPU architecture: Reduced Instruction Set Computer (RISC) and Complex Instruction Set Computer (CISC). Each of architecture has its own advantage. But when look at this blog of response time, RC time could be most likely depending on resistor and capacitor component.
“Computer processor might be changed by CPU architecture: Reduced Instruction Set Computer (RISC) and Complex Instruction Set Computer (CISC). Each of architecture has its own advantage. But when look at this blog of response time, RC time could be most likely depending on resistor and capacitor component.”
DaeJ, you may be right normally as I mentioned response time in PC depends up on many factors like clock speed, RAM, Accelerators etc
I am not happy about the use of the term “Feedback Current”. Current flowing in an input node is normally referred to as “Bias Current”. I should be very surprised if there were much leakage current to external circuitry.
@James: Yes you do have a good point here but what do you suggest instead of it ?
Use “Bias Current”, as analog engineers have for about half a century.