Current limiting stability problems

[David Hess]

There is still some noise on the output, especially when in current limiting. What can I expect here?
I measured my factory regulated power supply and that also gives quite a lot of noise when in current limiting state.

The bandwidth of your current control loop is lower which will have an effect on the output noise either way.

If the frequency compensation is marginal then the output noise will be higher.  I would do a transient response test to check how stable the output is.

The ACS712-5 you used as a current sensor may be inherently noisy.  I seem to recall that being the case with hall effect sensors.

Operation in current limiting is not usually designed with low noise in mind.

[akis]

In terms of reaction times and oscillations / noise, if your pass element was a simple transistor/FET eg an MJL3281A/MJL1302A or FETs like the IRF9540, all of which react in a given time, there are no sub-circuits lurking in there to generate delays, oscillations or whatever. Grounding the pin of the voltage controller will bring the output low, but what are the characteristics? I think it would be much simpler to use a simple transistor driven by an op-amp rather than a linear voltage regulator which does not have an adjustable current limit ability and you have to "hack" its ADJ pin to implement the current limiting. I know it is very common method and I have used it too, an LM317 and a BC327 grounding the ADJ pin if needed.

In terms of protection my bench PSUs have a diode (Schottky 20A) at the output, so if you plug in a large battery (or similar) nothing bad will happen. Actually the PSU resembles a battery charger which also has a diode at the output (mine do anyway). Sure there are a couple of Watts wasted there, but the PSU is almost indestructible. The voltage drop on the diode is of no consequence since the voltage is sampled in two separate places past the diode.

[David Hess]

I have always had excellent results using an integrated regulator as a pass element.  A 317 or 7805 provides fault protection in the form of thermal shutdown, safe operating area protection, and a peak current limit.  Used in combination with thermally mated pass transistors, an integrated regulator can somewhat provide these protections to them in a high output current design.

Old application notes show a very similar design with an external reference, error amplifier, and Kelvin connections to considerably improve low frequency noise and stability.  Last time I designed one, I used an OP-27 class operational amplifier for low noise and tweaked the frequency compensation for best transient response.  Line and load regulation was so good that I was not able to measure it on my best voltmeter.  I think a discrete output stage would have been lower noise and faster but not by a lot.

[akis]

You got it right, the external bypass transistor must be mounted on the same heatsink... I had for years a PSU with LM317/337 bypassed by MJL3281/1302 and the LMs blew many times, more times that I care to remember. Accidental shorts in particular would simply blow them up, I do not know why. I ended up bolting a fuse on the output as a way to prevent the LMs from dieing.

[David Hess]

Those are mighty fast transistors so it would not surprise me if parasitic inductance allowed the output to rise above the input, if leakage inductance in the transformer allowed the input voltage to rise above the maximum rated input of the regulator if the input capacitance was low, or if just a parasitic oscillation was the culprit.

I have had a couple of failures like this but they all came down to the output voltage being higher than the input voltage for whatever reason and that is easy enough to protect against with a shunt diode across the regulator and any pass elements.  A shunt diode to the adjustment terminal might be needed as well if it is bypassed.  Good application notes usually cover adding these two diodes but most designs do not need them.

[Remondepemon]

Hi guys,

Thanks for all the advice and tips! Designing a power supply is not so simple as I expected. I couldn't get the LT series stop oscillating and I decided to try the good old 2N3055 transistor instead. And what a blessing: simple, robust and stable! Have been playing around with the current limiting as well and was able to get it free of oscillations very simple.

So this is my schematic so far. Now I have to continue the digital part. The pots for the current and voltage will be replaced by an 12 bit DAC and controlled by a uP.

What do you all think?

Remon

[dannyf]

High speed devices are evil. Glad that you found it out.

you will need to provide a means to get the PS started. Depending on how you implement it, you can implement soft on or off, or one key on off here.

the voltage error amp needs an led in its output.

the current sense amp can be simplified.

[Remondepemon]

the voltage error amp needs an led in its output.

Why do I need an LED there? Only 0-100 uA runs there.

the current sense amp can be simplified.

I do not want to use a negative rail and my ACS give 2.5V at 0A. I decided to first zero the ACS to 1V@0A (0V is not possible without a negative rail) and then set the slope to 1A/V. What can be simplified to that?

Thanks,
Remon

[dannyf]

You can pad the pots output to null the 2.5v output, via a resistor or in software.

also, you need r2ri opamps here, or on that will swing to the negative rail, or one with pnp input stage. Lm358 or 393 would work here.

[void_error]

Either use rail-to-rail opamps or use a negative supply. It can be easily done using a voltage inverting charge pump.
The problem is that the voltage at the base of the TIP31 isn't going low enough. You could use a NPN transistor with your current loop opamp.

[Remondepemon]

Hi guys,

Did some basic ADC tests in my digital part, but just couldn't stop thinking about the analog part.

@dannyF: I want to have maximum resolution in my ADC conversion, so zeroing everything in software will decrease my resolution to much.

@void-_error: I changed one of my opamps to a CA3140 because of the strobe input and being it rail-to-rail as well.

On the current limiting part I got rid of the pots and replaced it with fixed resistors. Now, 0A gives me about 1.1V output but I discovered that the first amp is not to linear. Hopefully somebody can give me some advise on this. The goal is to get 1V/A as linear as possible from the ACS712. The main problem is to get rid of the 2.5 base voltage at 0A.
As I only have to get rid of 15W of heat in my linear regulator, I got rid of the TIP31 and the second 3055 transistor.
Then I discovered something else while performing high current tests. It seems that the higher the current, the lower the maximum output voltage. To be sure that my input voltage was high enough, I raised it to 30V. To be sure that my current limiting was not kicking in, I disconnected pin 8 (strobe) of the CA3140. Then I connected three loads to the power supply and these are my findings:

1. fixed resistor of 30 ohms
maximum voltage: 26.1V
current at 26.1V: 830 mA
P = 26.1 x 0.830 = 21.7W
R = 26.1 / 0.830 = 31.4 Ohm

2. car lamp of 12V/21Watts
maximum voltage: 16.1V
current at 16.1V: 1.85A
P = 16.1 x 1.85 = 29.8W
R = 16.1 / 1.85 = 8.7 Ohm

3. the lamp and resistor parallel.
maximum voltage: 12.7V
current at 12.7V: 1.95A
P = 12.7 x 1.95 = 24.8W
R = 12.7 / 1.95 = 6.5 Ohm

Hopefully somebody can answer the question why my maximum autoput voltage goes down.

Here is my latest schematic diagram.

[David Hess]

On the current limiting part I got rid of the pots and replaced it with fixed resistors. Now, 0A gives me about 1.1V output but I discovered that the first amp is not to linear. Hopefully somebody can give me some advise on this. The goal is to get 1V/A as linear as possible from the ACS712. The main problem is to get rid of the 2.5 base voltage at 0A.

Linear output hall effect sensors are not known for their linearity or even accuracy.  When accuracy is required, I have only seen them used in balancing configurations where their gain and linearity are irrelevant and only offset matters.

Maybe I am reading the datasheet wrong but it looks to me like the ACS712 is only good to about 6 bits.

How were you testing the linearity at low output currents?  Won't using the the CA3140 strobe that way cause oscillation?

[dannyf]

I would try to make sure that this thing works in DC; then to make sure that it is AC stable; and go from there.

[Remondepemon]

Maybe I am reading the datasheet wrong but it looks to me like the ACS712 is only good to about 6 bits.

@David: Can you explain me what you mean by 6 bits? In the ACS712-05 dataset the error is 1.5% over the full range. 1.5% over 5V is 0,075V. with an ADC of 10 bits, 1023 steps, the 0,075V is 16 steps is 4 bits. 

How were you testing the linearity at low output currents?  Won't using the the CA3140 strobe that way cause oscillation?

@David: I wrote down the voltage output 0-2A in 100 mA steps. From the ACS712-05 sensor it was linear. On the output of the opamp it was not linear below 1A. What can this have been? Do you suspect a measurement error there?

Thanks,
Remon
 
 

[David Hess]

Maybe I am reading the datasheet wrong but it looks to me like the ACS712 is only good to about 6 bits.

@David: Can you explain me what you mean by 6 bits? In the ACS712-05 dataset the error is 1.5% over the full range. 1.5% over 5V is 0,075V. with an ADC of 10 bits, 1023 steps, the 0,075V is 16 steps is 4 bits.

Exactly.  10 bits - 4 bits = 6 bits.  They way I came up with that number is 1.5% is 1 part in 66 which is close to 1 part in 64 which is 6 bits and that is only the *typical* error listed in the specifications.  It could be larger or smaller.

How were you testing the linearity at low output currents?  Won't using the the CA3140 strobe that way cause oscillation?

@David: I wrote down the voltage output 0-2A in 100 mA steps. From the ACS712-05 sensor it was linear. On the output of the opamp it was not linear below 1A. What can this have been? Do you suspect a measurement error there?

I am not sure what would cause that result.  The output of the LM358 should be close to 2.5 volts when the current is zero so it is operating well within its linear range.  Maybe it is oscillating.

[mij59]

I am not sure what would cause that result.  The output of the LM358 should be close to 2.5 volts when the current is zero so it is operating well within its linear range.  Maybe it is oscillating.

As mentioned, the asc712 is very noisy, with a filter cap of 1nF on pin 6 it outputs an equivalent of about 0.5A of noise.
In the schematic there's no filter cap.

[Remondepemon]

Exactly.  10 bits - 4 bits = 6 bits. 

I am not sure what would cause that result.  The output of the LM358 should be close to 2.5 volts when the current is @David: zero so it is operating well within its linear range.

At 0A, the ACS712's output voltage is 2.5V. For every Ampere, the voltage changes 185 mV. This allows you to measure negative and positive current. I just want to get rid of the 2.5V because I don't want to measure negative currents, but amplify my output voltage because of my ADC.

In the schematic there's no filter cap.

@mij59: You are right, the 1nF cap is there, but I forgot to include it in my schematic.

Leaves my most important question open; why does my output current affect my maximum output voltage?
Can anybody give his opinion about that?

 

[void_error]

You might want to try replacing the ACS712 with a differential amp and a power resistor (0.5 ro 1 ohm) and see if you get the same results. That way you can see if the problem is the ACS or not.

[Remondepemon]

After drinking a beer and relaxing downstairs, I continued my search why my output voltage had a relation with my output current and why my output current was not linear below 1A. Guess what... FIXED!

The voltage problem
My 3055 was pulling 12 mA from my opamp when I experienced a voltage limit. The CA3140 can deliver only 10 mA, so that was the reason.
I have added my TIP 31 again, and all is fine now. (See schematic)

The linearity problem
I did not connect the opamp power supply to the input voltage of 26V, but after my pre-regulator.
So the input voltage of the opamps was also a factor that caused the non-linearity in my current measurement.

New issue
After introducing the TIP31C, my voltage does not go down to 0V.  It doesn't go lower than 0,8V...
 

[dannyf]

why does my output current affect my maximum output voltage?

The output voltage is Vbe lower than the potential at its gate. Vbe goes up as Ic goes up.

Plus, if your current sensor has voltage drop (generally true but don't think it is true here), you lose some voltage there.

Not to mention the output resistance from the power source (transformer, diode's Vfwd, wiring, ...).

[dannyf]

My 3055 was pulling 12 mA from my opamp

You will need to deliver considerably more current to drive that 3055 at high current levels.

The use of a darlington (discrete or otherwise) will drop more voltage on the regulator -> more heat.

Think about some high current integrated voltage regulators in place of that 3055 - Linear has quite a few. They are much easier to drive.

[void_error]

New issue
After introducing the TIP31C, my voltage does not go down to 0V.  It doesn't go lower than 0,8V...
 

Your triple darlington doesn't turn off completely. Check the base-emitter voltages, opamp output voltages, base current (voltage on the 36onm resistor).

You have three options here:

1. Put a load resistor between the emitter of the 2N3055 and ground (220ohm/5W)
2. Put a 100ohm (or thereabouts, can't be bothered to do the math right now) resistor across the BE of the 2N3055
3. Use the TIP31 only, it'll be able to handle the dissipated power with the pre-regulator in place.

If none of the above work use a low voltage (-3V ish) negative rail.

[Remondepemon]

@Void: Thanks for all the advise I tried all options, but decided to move to an TIP142 darlington transistor. Might be a bit more expensive, but easy to work with and it saves me a lot of extra components.

For now, I am off to La France for a nice holiday. So in 2 weeks more opdates on the digital part that will be built around an Ardiuno and an external 12 bit DAC. Not sure yet if I want the microprocessor to set and measure or only to measure Voltage and Current.

Also, I have to focus on my 5 volts for the digital part. I don't want an extra transformer or dissipate too much heat from the 26V supply. I was thinking about using a transformer with a center tap, but am sure how to work out the diodes. (see: https://www.eevblog.com/forum/beginners/diode-bridge/)

Regards,
Remon

[void_error]

No idea why I assumed the TIP31 was a darlington... my bad.

For the digital supply you could use a LM2594 (or equivalent, 500mA output current should be enough).

[dannyf]

dissipate too much heat from the 26V supply.

In the case of a high current and large voltage range power supply, the same concept (of using a center-tap'd transformer) also applies to minimize power dissipation: you can switch between a full wave and half wave power supply to minimize the voltage drop on the regulator when  a low output voltage is needed.

Relays can switch between the two configurations.