A circuit to make regulators go to 0v.

[VEGETA]

Peace be upon you guys...

I have a project in mind of a portable battery powered power supply with isolated USB charging. I wanted to use cheap and suitable DC-DC converters so my main power design is a single cell Li-ion battery (18650) followed by a boost converter to make it around 30v (1.5A maximum). Then the true regulator is a buck converter to output 0-25v or something like that.

The suitable regulator that I found is called AOZ1280 and it is capable of 1.2A max with 0.8-26v output voltage which is just good enough for my design because I don't need huge power, it is just a simple low profile regulator.

However, as you notice, the minimum output voltage it gives is 0.8 which is a troll really, why not true 0v?! I know it is due to the internal voltage reference but I really hate it so much and I want true 0v. So I thought of using the idea everyone know which is feeding it of a negative voltage.

For example: feeding it with -0.8v gives us 0v because they cancel each other, and so forth. But I must offset the setting voltage all together by a factor of -0.8 to be suitable for all uses. Thus comes this circuit:

http://www.partsim.com/simulator/#57645  << This simulator is really great and totally free, so you can simulate the circuit directly!

As you see, I use a +30v positive rail (Boost converter), and most importantly, a negative -5v rail! this is the key part because the summer OP-AMP needs a negative supply (differential amplifier) to be able to negatively offset the voltage. Another requirement is to set a fixed -0.8v on the positive input to be "added" to the DAC\PWM output (0-2.5v = 0-25v) as you see.

Now, what I need is your recommendation of negative output regulator that is cheap enough (SMD of course, because I plan to manufacture and sell this thing). I thought of taking your thoughts before I search myself because I need to know if it will sink some current or should it be high current output or not. The part that I found good enough is MAX1720 but it's specs like output current might not be suitable. Please tell me your suggestions.



AOZ1280 datasheet: http://aosmd.com/res/data_sheets/aoz1280ci.pdf
MAX1720 datasheet: http://www.onsemi.com/pub_link/Collateral/MAX1720-D.PDF

thanks!

VEGETA

[Zero999]

Add, the LT3081 (a low drop-out regulator which can go down to zero) on the output.
http://cds.linear.com/docs/en/datasheet/3081fc.pdf

[VEGETA]

Add, the LT3081 (a low drop-out regulator which can go down to zero) on the output.
http://cds.linear.com/docs/en/datasheet/3081fc.pdf

Thanks for your feedback.

I don't want to use any linear regulation in this design, the purpose is designing a switching one. so that buck converter is all i need on the output.

[uncle_bob]

Hi

To have a switcher cover a range like 0 to 30V, you will need a set of relays to change the output capacitors, the inductor, and likely some of the other parts as well. Take a look at the stability criteria and the duty cycle range limits.

Bob

[VEGETA]

Hi

To have a switcher cover a range like 0 to 30V, you will need a set of relays to change the output capacitors, the inductor, and likely some of the other parts as well. Take a look at the stability criteria and the duty cycle range limits.

Bob

Hi

well, can you explain where in the datasheet you read this from? because I couldn't find it. The AOZ1280 buck converter says it can output from 0.8v up to 26v, so this range is not a problem. What I tried to achieve from this circuit is to operate it less than 0.8v until 0v. So my range became 0-25v assuming a constant 30v input voltage (from the boost converter).

I looked into its datasheet and I found the reference voltage is, as expected, 0.8v and it is on a comparator op-amp exactly like linear regulators. Now, Dave achieved 0v from LT3080 by cancelling out the Vref (it is a constant current reference, but same idea) using the same technique or similar. He explained mine but said it costs money.

thanks!

[uncle_bob]

Hi

To have a switcher cover a range like 0 to 30V, you will need a set of relays to change the output capacitors, the inductor, and likely some of the other parts as well. Take a look at the stability criteria and the duty cycle range limits.

Bob

Hi

well, can you explain where in the datasheet you read this from? because I couldn't find it. The AOZ1280 buck converter says it can output from 0.8v up to 26v, so this range is not a problem. What I tried to achieve from this circuit is to operate it less than 0.8v until 0v. So my range became 0-25v assuming a constant 30v input voltage (from the boost converter).

I looked into its datasheet and I found the reference voltage is, as expected, 0.8v and it is on a comparator op-amp exactly like linear regulators. Now, Dave achieved 0v from LT3080 by cancelling out the Vref (it is a constant current reference, but same idea) using the same technique or similar. He explained mine but said it costs money.

thanks!

Hi

They sort of left that all out didn't they? More or less, they are assuming you can spend a lot of time working it all out on your own. If you take a look at the data sheets from some of the other outfits, they are a bit more careful. They go a bit further explaining how to keep the part from blowing up on you.

One simple way to dig into it:

Take a look at the inductor and capacitors their math comes up with for 30V out and for 0.8V out. Hmmm .... they don't go into that either. There's a *lot* missing from that data sheet.

I'd bet that somewhere there is a set of app notes on the part. They go into the details of how to use it. Where they are ... no idea.

Good luck with that part !!

Bob

[VEGETA]

Hi

To have a switcher cover a range like 0 to 30V, you will need a set of relays to change the output capacitors, the inductor, and likely some of the other parts as well. Take a look at the stability criteria and the duty cycle range limits.

Bob

Hi

well, can you explain where in the datasheet you read this from? because I couldn't find it. The AOZ1280 buck converter says it can output from 0.8v up to 26v, so this range is not a problem. What I tried to achieve from this circuit is to operate it less than 0.8v until 0v. So my range became 0-25v assuming a constant 30v input voltage (from the boost converter).

I looked into its datasheet and I found the reference voltage is, as expected, 0.8v and it is on a comparator op-amp exactly like linear regulators. Now, Dave achieved 0v from LT3080 by cancelling out the Vref (it is a constant current reference, but same idea) using the same technique or similar. He explained mine but said it costs money.

thanks!

Hi

They sort of left that all out didn't they? More or less, they are assuming you can spend a lot of time working it all out on your own. If you take a look at the data sheets from some of the other outfits, they are a bit more careful. They go a bit further explaining how to keep the part from blowing up on you.

One simple way to dig into it:

Take a look at the inductor and capacitors their math comes up with for 30V out and for 0.8V out. Hmmm .... they don't go into that either. There's a *lot* missing from that data sheet.

I'd bet that somewhere there is a set of app notes on the part. They go into the details of how to use it. Where they are ... no idea.

Good luck with that part !!

Bob

I've seen similar regulators online and they all seem to be the same settings and even circuits. They all don't mention these stuff. thx
brb

[uncle_bob]

Hi

To have a switcher cover a range like 0 to 30V, you will need a set of relays to change the output capacitors, the inductor, and likely some of the other parts as well. Take a look at the stability criteria and the duty cycle range limits.

Bob

Hi

well, can you explain where in the datasheet you read this from? because I couldn't find it. The AOZ1280 buck converter says it can output from 0.8v up to 26v, so this range is not a problem. What I tried to achieve from this circuit is to operate it less than 0.8v until 0v. So my range became 0-25v assuming a constant 30v input voltage (from the boost converter).

I looked into its datasheet and I found the reference voltage is, as expected, 0.8v and it is on a comparator op-amp exactly like linear regulators. Now, Dave achieved 0v from LT3080 by cancelling out the Vref (it is a constant current reference, but same idea) using the same technique or similar. He explained mine but said it costs money.

thanks!

Hi

They sort of left that all out didn't they? More or less, they are assuming you can spend a lot of time working it all out on your own. If you take a look at the data sheets from some of the other outfits, they are a bit more careful. They go a bit further explaining how to keep the part from blowing up on you.

One simple way to dig into it:

Take a look at the inductor and capacitors their math comes up with for 30V out and for 0.8V out. Hmmm .... they don't go into that either. There's a *lot* missing from that data sheet.

I'd bet that somewhere there is a set of app notes on the part. They go into the details of how to use it. Where they are ... no idea.

Good luck with that part !!

Bob

I've seen similar regulators online and they all seem to be the same settings and even circuits. They all don't mention these stuff. thx
brb

Hi

There are indeed a lot of poorly spec'd parts out there. I've even seen cases where they send you to the competitor's app note library to find out how to use their part !! The gotcha is that circuit theory still applies. Strange (and possibly alarming) things will happen if you don't get the values right.

====

What are you going to do about current limit?

Bob

[VEGETA]

Hi

There are indeed a lot of poorly spec'd parts out there. I've even seen cases where they send you to the competitor's app note library to find out how to use their part !! The gotcha is that circuit theory still applies. Strange (and possibly alarming) things will happen if you don't get the values right.

====

What are you going to do about current limit?

Bob

Hmm I saw ALL switching regulators the exact same. Even Dave's ones are similar (pre-regulator stuff), I guess the bad things will happen when you wanna design something high power. But 1-3 amps is not I guess.

What about current limit? you mean a constant current feature of the power supply? of course I will make one! but the circuit here is for voltage control only which is the current problem of getting true 0 volts.

I will use the same method used by Dave and the others of making an op-amp comparator with a mosfet. I don't *fully* understand how it works but generally it is not easy. The only thing I don't know is the mechanism of limiting current by pulling the adjust pin low when the comparator is on. Meaning, even if i totally didn't know about it, the circuit still works perfectly... I just hate doing something I don't know.


I plan to design a PCB using CM then order the parts from digikey when I have enough money. There are lots of stuff needs to be done aside from main power supply circuit.

But if you are so pessimistic, what will you do if you were in my place?

always thanks for your support!

[StillTrying]

I don't want to use any linear regulation in this design, the purpose is designing a switching one. so that buck converter is all i need on the output.

How are you going to provide the 0.8V - 26V feedback to the buck, in theory you could feed it back 0.8V+ when its output was less than that. 

[uncle_bob]

Hi

There are indeed a lot of poorly spec'd parts out there. I've even seen cases where they send you to the competitor's app note library to find out how to use their part !! The gotcha is that circuit theory still applies. Strange (and possibly alarming) things will happen if you don't get the values right.

====

What are you going to do about current limit?

Bob

Hmm I saw ALL switching regulators the exact same. Even Dave's ones are similar (pre-regulator stuff), I guess the bad things will happen when you wanna design something high power. But 1-3 amps is not I guess.

What about current limit? you mean a constant current feature of the power supply? of course I will make one! but the circuit here is for voltage control only which is the current problem of getting true 0 volts.

I will use the same method used by Dave and the others of making an op-amp comparator with a mosfet. I don't *fully* understand how it works but generally it is not easy. The only thing I don't know is the mechanism of limiting current by pulling the adjust pin low when the comparator is on. Meaning, even if i totally didn't know about it, the circuit still works perfectly... I just hate doing something I don't know.


I plan to design a PCB using CM then order the parts from digikey when I have enough money. There are lots of stuff needs to be done aside from main power supply circuit.

But if you are so pessimistic, what will you do if you were in my place?

always thanks for your support!

Hi

The only practical way to come up with a clean supply is to have a linear post-regulator. You run the switcher down to (maybe) 3V when the output is at zero volts. The linear takes care of the current limit and the switching noise. You add another $1 and get a 10X better supply.

Bob

[Jay_Diddy_B]

Hi,

This can be done quite easily, provided that the regulator has a tracking capability. Here I am using the LT8640-1 to illustrate how to do this:



The circuit was designed to have a normal output voltage of 25V by choosing the divider resistors.
The TRACK/SS pin is essentially another reference. The chip will regulate to the lower of the references, either the internal 0.97V reference or the TRACK/SS pin voltage.

I have added a divider to the TRACK/SS pin so that 0-5V, V-CTRL, gives me 0-25V Output.

Here are the Modelling Results:



Because the LT8640-1 is a synchronous buck, operated in the forced continuous mode, it is able to discharge the output capacitors and return the energy to the input. This is happening when the output is being programmed to a lower voltage.

I have included a resistor from INTVcc to FB to introduce a small offset. This make sure that the output voltage can be commanded to 0V.


I have attached the LTspice for those playing along at home.

The LT8640-1 is a 'Silent Switcher'. These have significantly lower noise and EMC than most other switching supplies. I have include a small post filter using a ferrite bead to clean up the output.

Regards,

Jay_Diddy_B

[ebclr]

You can try put a Series diode and your 0.8V will drop to 0.2V

[VEGETA]

I don't want to use any linear regulation in this design, the purpose is designing a switching one. so that buck converter is all i need on the output.

How are you going to provide the 0.8V - 26V feedback to the buck, in theory you could feed it back 0.8V+ when its output was less than that. 

What does that relate to linear supplies?

I will feed it a voltage drawn from the 30v boost converter before it via op-amp x10 gain circuits, controlled by a microcontroller output (pwm\dac0).

Hi

The only practical way to come up with a clean supply is to have a linear post-regulator. You run the switcher down to (maybe) 3V when the output is at zero volts. The linear takes care of the current limit and the switching noise. You add another $1 and get a 10X better supply.

Bob

I wanted a pure switching supply with tolerable amount of output noise. I know Dave put an LT3080 and controlled it via the same method. However, from the datasheets, I assume I can do CV and CC modes with the buck converter. Maybe later on when I finish it I will consider trying a post regulator.

Jay_Diddy_B

I know that device! it is really great but it is like 9$ from digikey and I don't need all that power, my maximum is 1-1.2 amps. I don't quite understand how you achieved true 0v from this circuit. What I know is that you must eliminate the voltage reference effect to achieve 0v. I used a negative voltage offset method but you seem to have a cheaper method because it doesn't use external parts.

If you explain how to use it with my buck regulator or in general, I will be happy xD. thanks for your feedback!

[VEGETA]

Jay_Diddy_B

I studied what you wrote and I think I understood the concept, I will explain it to you and you tell me if it is wrong or good please. Look at this internal diagram of your chip:



The comparator equation is like this:

0.97 (internal ref) + TR\SS voltage - FB voltage = ouput

your resistor divider on the TR\SS pin outputs exactly 0.97v @ 5v max so this makes the output equation like this: 

V_out = 0.97 + 0.97 - FB

while normally TR\SS is tied to ground and by the effect of the internal current source it makes it nearly 3.4v according to the datasheet because if it is > v_ref the thing will not be tracked and activate the ramping feature. However, they explicitly say (and encourage) that you drive this pin directly from external voltage ranging from 0-0.97v to ***override*** the internal 0.97v reference voltage... this is the quote that I got this from:

For output tracking applications, TR/SS can
be externally driven by another voltage source. From 0V to
0.97V, the TR/SS voltage will override the internal 0.97V
reference input to the error amplifier, thus regulating the
FB pin voltage to that of TR/SS pin.

This is where I ask, how to get 0-25v by making a 0-0.97v TR\SS voltage? I mean, the error amplifier voltage is 0.97+0.97-FB = 1.97 - FB. The datasheet says the TR\SS voltage "Overrides" the internal reference voltage... where is that in the error amplifier? the error amplifier has 2 + and one - inputs... there is no where in the schematic that shows how to eliminate the ref voltage.

If it is eliminated and we have a "variable reference" voltage, then it is solved! I can set it to 0v by a 0% duty cycle pwm and the output goes 0v. However, you put a 2.2Mohm resistor in the feedback of the output... I failed to understand its job and why it makes an offset. Please explain.


Finally, I have 2 questions:

1- Is it possible to have current limit of this regulator? what about constant current (fixed current with variable voltage)?

2- Is there other cheaper (2A max) regulators? maybe buck-boost or buck or sepic... There is no way in digikey that I can search from "tracking" switching regulators.

THANKS!

[tszaboo]

This can be done quite easily, provided that the regulator has a tracking capability. Here I am using the LT8640-1 to illustrate how to do this:

Did you actually try to do this? I'm guessing there will be some issues, when light load, because of the minimum on time, and switching frequency.

[VEGETA]

This can be done quite easily, provided that the regulator has a tracking capability. Here I am using the LT8640-1 to illustrate how to do this:

Did you actually try to do this? I'm guessing there will be some issues, when light load, because of the minimum on time, and switching frequency.

He put an LTSpice model for it, I don't know if he tried it on real design. Isn't switching frequency affect noise and inductor size? what does it have to do with this 0v operation?

What about minimum on time? If you mean something like soft-start, it is there using a capacitor on the tracking pin. And why light load?

Do you have a better solution?

thanks!

[uncle_bob]

This can be done quite easily, provided that the regulator has a tracking capability. Here I am using the LT8640-1 to illustrate how to do this:

Did you actually try to do this? I'm guessing there will be some issues, when light load, because of the minimum on time, and switching frequency.

He put an LTSpice model for it, I don't know if he tried it on real design. Isn't switching frequency affect noise and inductor size? what does it have to do with this 0v operation?

What about minimum on time? If you mean something like soft-start, it is there using a capacitor on the tracking pin. And why light load?

Do you have a better solution?

thanks!

Hi

Minimum on time = One of the specs that is missing on the data sheet of the regulator you are looking at. On a simple PWM regulator, the pass element is only just so fast. There is some minimum time that you can turn it on for (it will not go to zero nanoseconds). In a switching setting, that gives you a minimum output voltage before you have to go to something other than PWM. Exactly how your regulator handles this will impact it's ability to deal with low voltage outputs.

Bob

[Jay_Diddy_B]

See Below ...

[Jay_Diddy_B]

Edit - Corrected the gain

Hi,

I have not built this circuit, but I have run the LTspice simulation, which is normally very accurate.

I am going to explain some of the subtle features of the circuit that I presented.

2.2 M Ohm resistor

If you look at the partial block diagram:



You can see that there is a 1.9uA current source on the TRK/SS pin. This is required in normal operation to charge the SS capacitor. However, in the circuit that I presented this current  is undesirable. It creates an offset on the TRK/SS PIN.

Consider these models:



And the modelling result:



You can see that the voltage is higher on the circuit with the 1.9uA current source.

The offset can be calculated as the Thevenin equivalent resistance of the divider multiplied by the SS current. With the values chosen the offset is 15mV.

The whole circuit has a gain of 5 25 from the TRK/SS pin to the output, so the 15mV offset becomes 75 375mV on the output.

(The gain is from the TRK/SS PIN to the output. The gain is 5 from the from V_CTRL to the output)

Compensation

Ideally you would want to create a current source of the appropriate value and inject this current into the output voltage divider to create a matching offset.

The Thevenin resistance of the output divider is 10K//249K = 9.6 k Ohm

So to get 15mV of offset we need a current of 15mV/9.6k = 1.56 uA

You could go to a lot of trouble and design a 1.56uA current source, or you can make a first order approximation using a voltage and a resistor.
The INTVcc, at a nominal 3.1V is a suitable voltage.

Model



Modelling Results




So in the first order the offset caused by the SS current source has been cancelled. There is little point in having a better solution because the SS current source is specified as 1.2 to 2.6 uA over the full temperature range of the part.

Regards,

Jay_Diddy_B

[VEGETA]

Well, I investigated another similar IC called LT8611 which has a good unique feature of setting the output current limit by a voltage from microcontroller or just traditional power resistor limit. Other tricks and quirks can be done to do it different ways as long as the voltage is all we need.

After I read your explanation, I guess I understood your idea. You made a resistor divider that outputs 25v at all times if compared to the 0.97v internal regulator by this equation:

V_out = 0.979 * (R2/R1 + 1) ---> 0.970 * (249/10 + 1) = 25.123v which is the maximum that we have. Now your idea is to use the TR\SS pin to act as a variable gain from 0-1 which corresponds to 0 being 0 output voltage and 1 being 25v output voltage. You got the 0-1 from dividing 0-5. I don't know if this can be accurate but it works as you posted.

I was misguided by the datasheet saying driving TR\SS pin with voltage above 0.97 will make it return to the internal 0.97 reference so you must make it 0-0.97v with your resistor divider. What confused me more is where this gain of 5 came from, because of the output feedback resistor divider.... But now I understood your idea.

My concept was different and totally wrong! I wanted to feed a 0-25v to the feedback from op-amp circuits (just like linear regulators) and deal with the 0-0.97v TR\SS pin the same method.... However, I read the "Absolute maximum" specs that said FB and TR\SS MUST NOT exceed 4v!!! How the hell can I feed it 0-25v at this time?! My mistake is assuming the feedback mechanism is similar to famous linear regulators.

I wanna build this circuit up but no money now xD! plus I want to make a full PCB of the whole design because there is not much thing other than this issue here.

My choice will be using LT8611 instead of this device because it has great current monitoring and current limiting features which will enable me to measure output current and make constant current mode by one regulator. I saw it is similar to this one you have here in most cases. I still need to figure out how to control the current via the 0-50mv input control.... my guess is 0-50v = 0-2.5A (its max amp is 2.5). the only drawback from this is it is not silent switcher so I will use more ferrite beads.

I will keep you updated for sure! your help made great deal to me really!

[Jay_Diddy_B]

Hi,

I did make a mistake in the gain and I have corrected the post above.

Here is an explanation of the low level performance. As some members have pointed out, it is hard to make a power supply do very small voltages.

For a buck regulator:

Vout = duty-cycle x Vin

So as Vout -> 0 ,duty-cycle -> 0 and the ON time -> 0.

Chips have a minimum ON time, for the LT8640-1 it is a very short 30ns.

Model No load current, low output voltages




Modelling Results




Details



There are a couple of thinks happening. The Forced Continuous Mode, FCM, allows the inductor current to be negative. This means if you can have ripple current with an average of zero.

The other thing that has happened, is the integrated circuit is pulse skipping. The minimum on-time prevents the chip from running at the target frequency 1MHz, it operates at 100kHz to meet the duty-cycle requirements with a 30ns minimum on-time.

At the higher output voltage of 0.5V



The chip is still pulse skipping but the skipping is not as regular. This is not really a big issue.

Regards,


Jay_Diddy_B

[Jay_Diddy_B]

Hi,

If I try the same techniques on the LT8611, I get the following model:



These are the results:



And zooming in:



You can see that the results are not nearly as good as the LT8640-1. This is because the part lacks the Forced Continuous Mode. The inductor current is only allowed to flow from input to output. The chip is pulse skipping, but at a much lower frequency.

In addition the LT8611 will not discharge the output, returning the energy to the input:



I have attached the LTspice model.

Regards,

Jay_Diddy_B

[VEGETA]

LT8611 has minimum on-time of typical 50nS (30 min, 70 max) and it gets a little bit better if synchronized to an external clock. I don't know the benefit of it till now, is it essential for this application?

About pulse skipping, I saw the datasheets linking it to lowering frequency to match the minimum on-time and somehow it is linked to "Burst mode" which, I assume, doesn't exist in your regulator. Quoting datasheet of LT8611:

SYNC (Pin 1): External Clock Synchronization Input.
Ground this pin for low ripple Burst Mode operation at low
output loads. Tie to a clock source for synchronization to
an external frequency. Apply a DC voltage of 3V or higher
or tie to INTVCC for pulse-skipping mode. When in pulseskipping mode, the IQ will increase to several hundred
µA. When SYNC is DC high or synchronized, frequency
foldback will be disabled. Do not float this pin.

and most importantly:

The LT8611 will not enter Burst Mode operation at low
output loads while synchronized to an external clock, but
instead will pulse skip to maintain regulation.

Since the project will have a microcontroller, I guess I can use it to switch to "Burst mode" in these light loads (starting from what voltage or current?). But it all depends on synchronization or not.


further explanations:

For some applications it is desirable for the LT8611 to
operate in pulse-skipping mode, offering two major differences from Burst Mode operation. First is the clock stays
awake at all times and all switching cycles are aligned to
the clock. Second is that full switching frequency is reached
at lower output load than in Burst Mode operation. These
two differences come at the expense of increased quiescent
current. To enable pulse-skipping mode, the SYNC pin is
tied high either to a logic output or to the INTVCC pin.

Now this is interesting:

The LT8611 does not operate in forced continuous mode
regardless of SYNC signal. Never leave the SYNC pin
floating.

while LT8640-1:

At light loads, FCM
operation is less efficient than Burst Mode operation or
pulse-skipping mode, but may be desirable in applications
where it is necessary to keep switching harmonics out
of the signal band. FCM must be used if the output is
required to sink current.

Hmm so the only difference here is the ability to sink current or not. You demonstrated its usefulness very well but can the circuit work without it? I mean they claim LT8611 and many others do work like this. It is said also that burst mode and pulse skipping is more efficient, so maybe the current sinking is the only difference.

You didn't tell me how do you achieve constant current?

[VEGETA]

The best thing about LT8611 is current monitoring and control feature. This doesn't exist in LT8640-1, so how will you achieve constant current or fixed current mode from this regulator?

Will the circuit work without problems in low voltages (regardless of its efficiency)?