EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: LoveLaika on May 08, 2020, 09:19:57 pm
-
I'm developing an adjustable voltage supply to give out positive and negative voltages as well as current from each supply, up to 2.5 amps. The schematic is shown below. I'm using the LM2673, working off of a 15 V input going to both of the regulators, and I have a 1-ohm sense resistor with an op-amp for each supply in order to adjust the current. I ran this in SPICE, and it seems to work well in practice, but I wanted to get a review of my schematic. The values for the components match up with TI's guidelines, but I was wondering if there were any improvements that I should look into or consider. One thing that concerns me as well is the output. The positive output is straightforward, but for the negative output, I labeled the connector accordingly based on how I had it hooked up in SPICE, so if I take the voltage difference between NEG_RETURN and NEG_OUT, NEG_OUT should be at a lower voltage. Overall, does this circuit look okay?
-
The CC loop might be unstable. What does pin 7 of the op-amp look like?
-
Well, the first op-amp acts as a differential amplifier (G=1) while the second one acts as a comparator, with pin 7 being its output. The diff. amp. converts the current measurement into a 1-to-1 voltage, going to the comparator. Pin 7 goes to around 3.5 V when the 'current' exceeds the reference voltage set by the potentiometer going into pin 6. Otherwise, it stays below 0.5 volts. If it goes above 1 volt, it turns on the diodes at pin 7 and controls the regulator feedback voltage that way, decreasing the output voltage in order to maintain the current limit. I put those LEDs there to act as an indicator when you're reaching the current limit, but I can change that around to swap it around with a regular diode or something.
I got the idea from this link below. The LM2673 feedback voltage pin can go as high as 14 volts, so this seems to be within the acceptable limits. Though, I do see your point. The comparator output behaves differently for both the negative and positive voltage regulators, but they both behave the same way. I've attached a waveform of how they behave along with my test circuits. The negative output voltage circuit is set up the way it is because of issues with the model, so that's why the output voltage for that one is positive.
https://www.youtube.com/watch?v=EVL7TzCde8I (https://www.youtube.com/watch?v=EVL7TzCde8I)
By the way, you think the 7805 regulator's okay? It's only powering the op-amp and the current reference, but those are only just a few milliamps.
-
Although there is a lot of activity at pin 7, it might not matter. It could be clean up with some capacities feedback on the compatitor op-amp but this will slow the response.
7805 regulators don't make the best references but might be ok for your application. Check to see how much the output varies with change in input voltage with a real part.
-
Thanks for your reply. Considering that my timescale is in milliseconds, if a small capacitor can clean that signal up fast enough, then I'm all for it. Is it as simple as adding a capacitor at the output in order to reduce that noise? Would it be a good idea to add a small resistor like 1k to limit the op-amp current? Might be best to clean up that signal so as to not affect the regulator feedback.
The 7805 has a dropout of roughly 2 V, so since it's being powered at 15 V, as long as it's above 7V, it should be fine, right? What other ICs would you recommend here, or should I just splurge on a pre-made buck module fixed for 5V?
-
Thanks for your reply. Considering that my timescale is in milliseconds, if a small capacitor can clean that signal up fast enough, then I'm all for it. Is it as simple as adding a capacitor at the output in order to reduce that noise? Would it be a good idea to add a small resistor like 1k to limit the op-amp current? Might be best to clean up that signal so as to not affect the regulator feedback.
The 7805 has a dropout of roughly 2 V, so since it's being powered at 15 V, as long as it's above 7V, it should be fine, right? What other ICs would you recommend here, or should I just splurge on a pre-made buck module fixed for 5V?
Experiment with a 10K in series with the inverting input and a capacitor from output to inverting input.
I usually use a 3 pin regulator followed by a TL431 shunt regulator.
Just lately I noticed the output of a 3.3V 3 pin regulator varying voltage with change of input while well above dropout. I suspect it was a thermal effect.
-
Hi,
Assuming that you are making a power supply with a dual output. You will connect the POS_RETURN and the NEG_RETURN together at the load.
When you do this you will short R6 and lose the current sensing in the negative half of the supply.
[attachimg=1]
In addition the 1 \$\Omega\$ sense resistor are high values for a 2.5A supply. Dissipation is 6.25W at 2.5A
Regards,
Jay_Diddy_B
-
I think the feedback capacitor will need a resistor in series with it also.
-
The CC loop might be unstable. What does pin 7 of the op-amp look like?
The response though the output LC filter is so slow that it might be stable.
-
The CC loop might be unstable. What does pin 7 of the op-amp look like?
The response though the output LC filter is so slow that it might be stable.
It could be just ripple getting through, not loop oscillation.
-
Thanks for your reply. I found a 1-ohm sense resistor on digikey rated for 6 watts, but in the worst case, one trick I saw from the video is to take precision, high-watt resistors in parallel to get down to 1-ohm and not worry about power dissipation.
I see what you mean regarding the load connections. Depending on the load, that would be a problem. Working with the negative circuit by itself, it seemed okay when I was running it in SPICE, but now I see the problems that occur when the load output share a common ground. It's always the issue with the negative output current sense circuit that's causing all sorts of problems, and here I was hoping that I had it figured out. If I move GND to NEG_RETURN, R6 can't get an accurate measurement of the current going to the load from the negative side, and the feedback loop won't be able to accurately compensate for R6's voltage drop.
-
What do you mean by might be stable? You think there's too much capacitance at the output that's affecting the comparator op-amp? I went with high capacitance in order to reduce the output ripple.
-
So, with a capacitor by itself, it'd be an integrator. A capacitor/resistor parallel together would form an LPF. That's an interesting thought. I'll give that a try. Thanks for the comments. I'm learning more issues with my circuit as I read responses along with some ideas that I'd like to try out.
-
So, with a capacitor by itself, it'd be an integrator. A capacitor/resistor parallel together would form an LPF. That's an interesting thought. I'll give that a try. Thanks for the comments. I'm learning more issues with my circuit as I read responses along with some ideas that I'd like to try out.
A capacitor by itself forming a Miller Integrator works with linear designs, then it occurred to me that extra lag caused by the inductor would likely cause it to make things worse.
Putting a resistor in series with the capacitor allows some proportion response, reducing lag.
If the CC loop is really unstable and oscillating, the resulting audible squealing could be annoying.
-
What do you mean by might be stable? You think there's too much capacitance at the output that's affecting the comparator op-amp? I went with high capacitance in order to reduce the output ripple.
I mean the LC output filtering on the switching regulator may reduce the high frequency gain enough before a phase lag of 180 degrees is reached that the feedback loop may be stable.
-
So, with a capacitor by itself, it'd be an integrator. A capacitor/resistor parallel together would form an LPF. That's an interesting thought. I'll give that a try. Thanks for the comments. I'm learning more issues with my circuit as I read responses along with some ideas that I'd like to try out.
A capacitor by itself forming a Miller Integrator works with linear designs, then it occurred to me that extra lag caused by the inductor would likely cause it to make things worse.
Putting a resistor in series with the capacitor allows some proportion response, reducing lag.
If the CC loop is really unstable and oscillating, the resulting audible squealing could be annoying.
Then, perhaps a different method might be worth looking into for current limiting. I found a video online showcasing the LMP8646 current sensing amplifier, and using that in place of the TLC272 seems to work pretty well. There's a lot less ringing at the feedback node, and it's more suited for stuff like this.
-
One possible way of looking at it is, assuming that the output is shorted to ground, the CC loop is effectively regulating voltage across the CS resistor.
The voltage drop across the CS resistor needs to be transferred to the FB pin without much gain applied.
The DC to low frequency gain can be high without causing a problem.
-
One possible way of looking at it is, assuming that the output is shorted to ground, the CC loop is effectively regulating voltage across the CS resistor.
The voltage drop across the CS resistor needs to be transferred to the FB pin without much gain applied.
The DC to low frequency gain can be high without causing a problem.
I have to change my design around due to issues with the output connection. The positive and negative outputs need a shared return to GND on the board, so the way I have current limiting now, with the sense resistor and IC, may not work.
I was thinking of reworking it by adding a 30k-pot (if I can find one) to the Iadj pins and limit the current through the inductor. Radj at that pin = 37125/(Peak Switch Current), and the peak switch current needs to be approx. 1.5x the max load current, so I could limit it that way as well.
-
If you don't mind me asking, you have any suggestions as to some alternatives that I can consider? Considering the type of output connection I want, with a single GND point, I'm looking into new ways of running my negative circuit.
Take this with the LT1074. Pretty simple, lots of documentation, but if I want to put in a sense resistor for measuring current, say to use as a warning indicator, how can I fold it into my circuit without affecting the output voltage?
-
Considering what one user said, the needs for my connection to have one single GND pin makes my circuit pretty much unusable. I'm looking into new methods, such as with the LT1074.
It seems to work well (and the model works in LTspice), but if I want to add a sense resistor, how can I incorporate it into my circuit without affecting the voltage?
-
Hi,
Take a look at this:
[attachimg=2]
Not finished, but an idea for you to work on.
Regards,
Jay_Diddy_B
[attachimg=1]
-
Thanks for your reply as well as your example circuit.
Sorry for asking like this, but I've been going back and forth with some others about my project, and I've been rethinking my original circuit. I was hoping that I could get some feedback on my thoughts.
With the LM2673, I can use the current limiting feature (I_adj) with a pot to limit the switch current, so as my load current increases, the output voltage will drop if it exceeds that limit. I verified that in SPICE, so in that sense, I can more or less current limit my regulators, plus or minus 25%. This is done automatically within the regulator, so that takes care of one problem.
I'd like to add an LED indicator of sorts so I can 'indicate' when a limit has been reached, so I still need a current sense feature, but now, it doesn't have to be folded back into the feedback loop. The key thing was my sense resistor at 1 ohm. I have to shrink it down to something like 1 milli-Ohm in order to have a negligible voltage drop, and this way, I can place it alongside NEG_OUT and POS_OUT without worrying about feedback and voltage drop. Does this sound more or less alright? Following this, it seems that the only thing I have to decide now is picking a current sense op-amp.
-
LoveLaika,
In the circuit I gave you. The differential amplifier U3 measures the output voltage directly.
The voltage from U3 output to the GND pin on U1 is the same as the voltage across the load.
The 1 \$\Omega\$ current sense resistor is to illustrate that this method works even if there are large current sense resistors. You can, and should, use a smaller value resistor.
1 milliohm sense resistor have other problems when you are trying to measure small currents.
The I-adj pin on the LM2673, will give you current limiting. Most bench power supplies are constant current. If you implement the current limiting with the I-adj pin the output current will vary with output voltage.
You can still use a current sense amplifier.
Jay_Diddy_B
-
Thanks for your reply. With your warning though, about current limiting with the I-adj pin, isn't that the same idea? If the load draws too much current, then the output voltage will decrease (or increase in the case of the negative output voltage). This occurs without the current feedback amplifier loop depending on how I_LIM/I_ADJ is set, so wouldn't that make the diff-amp redundant (unless you're using it for something else)?
-
Hi,
Even if you use the I-lim pin on the chip for limiting, you will probably need a sense resistor for metering.
Jay_Diddy_B
-
Thanks again for your help. I really appreciate all the help that you've given. If you don't mind, what do you think of my circuit below? I revised it following your example schematic, which helped a lot, and I still get the output that I want. Parts are subject to change depending on tolerances and values, but it seems to work in LTspice, at least up until I try and include the TLC272. The program just bugs out due to timestep stuff.
The INA293 with the sense resistor manages a 1-to-1 ratio for measuring current without a large voltage drop, which is used by the TLC272 to compare it to a reference voltage, and then it does whatever it wants, in this case, light up an LED. The respective op-amps are tied to the correct supply reference so there should be no issues with the current draw. The potentiometer at I_ADJ allows for easy adjustment for current.