()blivion
Active Member
Hello everyone. I've got a project I'm working on finishing for submission to the circuits page of this site, and was wondering if anyone wanted to see it first and provide input on it. A new pair of eyes might catch something I missed.
Synopsis.
The main circuit is a boost design I started working on yesterday and have almost finalized. It was designed based on parts from three different circuits. The first part is the infamous "Joule Thief" LED boost circuit. I started the design with this as the base circuit because of it's great efficiency and amazing ability to "steal" usable power from very low battery's. The second circuit inspiration came from Mr RBs great two transistor boost circuit (which you can read about *HERE*). I really liked his two transistor positive feedback system because it slams things on and off, making switching very efficient. The third pieces of circuitry are very simple but effective modifications I came up with myself when playing with Mr RBs circuit. These modification give these circuits more energy efficient regulation and a flatter more noise free voltage output, suitable for more sensitive electronics.
Not to toot on my own horn, but these three expert designs coming together create probably one of the best simple and efficient boost regulator circuits on the internet that doesn't involve an integrated circuit and can be purchased for less than 5$. For it's ability to suck the last bit of juice right out of a battery and put it to work for more than just lighting LED's, I am naming it "The-Scavenger". I still need to make 100% sure it works in hardware, and figure out a few caveats if I can. But as is the circuit should already be workable, and the future of modifications to the base design look very promising.
The circuit.
There are two versions to look at right now. I'm mostly going to focus on the first one, the second one will be explained in relation to it.
The first one I will present is 100% working in theory and needs little to no modifications. It pretty good in it's own right. It's only real problem is a minor efficiency quirk that I would like to do away with. Here is the schematic of this version...
View attachment 67221
Notes:
(1)R2 Should be 1K, not 10K. I'm too lazy to fix it right now. It actually is 10K in the second version of the circuit though. So don't confuse the two.
(2)R1 may need to be adjusted, lower values are for lower frequency's and lower Vin conditions and higher currents. Higher R1 does the inverse of this. R1 can also go on either the top or the bottom of the sense half of the transformer T1, it doesn't seem to matter much electronically. Putting it on the bottom allows one to use a center tapped transformer though. This is easy to make by hand wrapping a toroidal inductor. All of my builds will likely use this method for T1.
(3)R3 is optional, it is the main RC filter component. Having it installed creates a much flatter Vout, but causes a little bit of extra loss, especially at higher loads. You can replace it with a piece of wire if energy efficiency is more important to you than low noise.
(4) Under certain conditions, there may be a significant ripple current on C2. Such could also be the case if you remove R3 as per above. C2 needs to be able to handle this ripple current or it may heat to the point of destruction and/or you may loose efficiency. Using a polly cap or large Tantalum/silver mica is preferred. If you want efficiency, AND want ripple free Vout + C2, It should be possible to replace R3 with a suitable inductor. I have not tested this yet, but I do plan to.
And here are the main points....
Pros.
It has decent load and line regulation, At around 1500uV with the filter circuit.
It uses very common parts with a low low parts count, 11 parts with the filter, 10 Without.
It is very forgiving with parts tolerances. You can vary most of the parts by 50%.
It will run from a supply all the way down to 700mV!!! Allowing you to use the deadest of battery's.
It has a high energy efficiency when under normal load. And a decent efficiency when lightly loaded.
Using a transformer for current sense instead of a resistor allows for more efficiency.
Cons.
Has some ripple on the Vout if there is no filter circuit.
Has lower efficiency when lightly loaded, needing to shunt Q1's base current directly to ground.
Can't have very high a Vout filter capacitor, or it will start to effect the rest of the circuit.
The circuits frequency varies widely with operation, causing it to possibly emit annoying noises.
Requires an odd transformer, larger ones are more efficient, but will sacrifice other parameters.
It can only boost, and will only work well at no more than 1/2 to 3/4 of target Vout.
Going Farther.
The main problem, and the entire reason for the second version is because the above version is less energy efficient. When regulating for a light load, it shunts some mA of current directly to ground. I will be building this first version soon and I expect no major divergences from simulations as the main sub circuits have been proven elsewhere.
The second version I came up with to combat and correct for the problems of the first. But it too has problems. Here is the schematic of it...
View attachment 67222
The difference and thing to notice with this second version, is that instead of shunting Q1 base current directly to ground to stop Q1 from turning on, I instead block it altogether with a high side PNP switch. This work just as well to keep Q1 off (in simulation), but I have removed the quiescent current associated with grounding R1. The problem this creates is that the point at which Q2 turns on is now determined by Zd1 + Vin. This means to make a target Vout of 5 volts, you need Zd1 to breakdown at 4 Volts AND you need Vin to be exactly 1 volt. This is very not optimal.
The whole point of this circuit is to use as much of the systems battery energy as possible. This means we want to BOTH convert the energy efficiently, AND drain as much as we can from the battery's. So as you can see, each version has a road block that needs to be solved before it is ready for service.
Ideas?
So I'm looking for ideas, I left the first circuit up as a fall back design because it almost certainly works. And because it may be found out that it's actually possible to take it farther than the second one. But I really want to focus on the second version.
With the second version, all current STOPS in the simulator when the regulation kicks in! The only current that is still flowing is the load and a very tiny Zd1+R2 current. When the regulation kicks on in the first version, most every thing stops, but there is still a decent amount of current drain from R1 being shorted to ground. Under light loads, the global power drain is actually dominated by this current. It's only a few mA with the right choice of R1. But if this current can be removed while the circuit is still functioning, it will be nothing but profit. So I'm at least giving it some deserved thought.
I have considered somehow stabilizing the voltage that is fed to the sense half of T1 so that it will not vary with supply. But I think this will add as much or more power loss as the first circuit and it will raise the floor for how low Vin can be. Both these things are very counterproductive. I have also considered and tested the idea of using different parts and configurations for blocking Q1 base current. The PNP seems to be the best choice so far, though the sim isn't very good with MOSFET's. I was also looking into possibly using the negative kickback produced on T1 sense as the voltage feedback instead of Vout. This voltage kickback voltage seems to be directly proportional to output voltage, so it seams just as viable to create a breakdown based control circuit here as it is for how I already have i. The only difference is this area creates a negative voltage that may allow for tricks that wouldn't be possible otherwise.
Conclusion.
Anyway, that's where I'm at now. I would love to hear some input from the elites and hopefully some good advice as how to proceed. If not, I will prolly just run with version number one as it's only major problem right now is efficiency and the fact that it has not been physically built and tested yet.
Thanks for reading.
-()blivion
Synopsis.
The main circuit is a boost design I started working on yesterday and have almost finalized. It was designed based on parts from three different circuits. The first part is the infamous "Joule Thief" LED boost circuit. I started the design with this as the base circuit because of it's great efficiency and amazing ability to "steal" usable power from very low battery's. The second circuit inspiration came from Mr RBs great two transistor boost circuit (which you can read about *HERE*). I really liked his two transistor positive feedback system because it slams things on and off, making switching very efficient. The third pieces of circuitry are very simple but effective modifications I came up with myself when playing with Mr RBs circuit. These modification give these circuits more energy efficient regulation and a flatter more noise free voltage output, suitable for more sensitive electronics.
Not to toot on my own horn, but these three expert designs coming together create probably one of the best simple and efficient boost regulator circuits on the internet that doesn't involve an integrated circuit and can be purchased for less than 5$. For it's ability to suck the last bit of juice right out of a battery and put it to work for more than just lighting LED's, I am naming it "The-Scavenger". I still need to make 100% sure it works in hardware, and figure out a few caveats if I can. But as is the circuit should already be workable, and the future of modifications to the base design look very promising.
The circuit.
There are two versions to look at right now. I'm mostly going to focus on the first one, the second one will be explained in relation to it.
The first one I will present is 100% working in theory and needs little to no modifications. It pretty good in it's own right. It's only real problem is a minor efficiency quirk that I would like to do away with. Here is the schematic of this version...
View attachment 67221
Notes:
(1)R2 Should be 1K, not 10K. I'm too lazy to fix it right now. It actually is 10K in the second version of the circuit though. So don't confuse the two.
(2)R1 may need to be adjusted, lower values are for lower frequency's and lower Vin conditions and higher currents. Higher R1 does the inverse of this. R1 can also go on either the top or the bottom of the sense half of the transformer T1, it doesn't seem to matter much electronically. Putting it on the bottom allows one to use a center tapped transformer though. This is easy to make by hand wrapping a toroidal inductor. All of my builds will likely use this method for T1.
(3)R3 is optional, it is the main RC filter component. Having it installed creates a much flatter Vout, but causes a little bit of extra loss, especially at higher loads. You can replace it with a piece of wire if energy efficiency is more important to you than low noise.
(4) Under certain conditions, there may be a significant ripple current on C2. Such could also be the case if you remove R3 as per above. C2 needs to be able to handle this ripple current or it may heat to the point of destruction and/or you may loose efficiency. Using a polly cap or large Tantalum/silver mica is preferred. If you want efficiency, AND want ripple free Vout + C2, It should be possible to replace R3 with a suitable inductor. I have not tested this yet, but I do plan to.
And here are the main points....
Pros.
It has decent load and line regulation, At around 1500uV with the filter circuit.
It uses very common parts with a low low parts count, 11 parts with the filter, 10 Without.
It is very forgiving with parts tolerances. You can vary most of the parts by 50%.
It will run from a supply all the way down to 700mV!!! Allowing you to use the deadest of battery's.
It has a high energy efficiency when under normal load. And a decent efficiency when lightly loaded.
Using a transformer for current sense instead of a resistor allows for more efficiency.
Cons.
Has some ripple on the Vout if there is no filter circuit.
Has lower efficiency when lightly loaded, needing to shunt Q1's base current directly to ground.
Can't have very high a Vout filter capacitor, or it will start to effect the rest of the circuit.
The circuits frequency varies widely with operation, causing it to possibly emit annoying noises.
Requires an odd transformer, larger ones are more efficient, but will sacrifice other parameters.
It can only boost, and will only work well at no more than 1/2 to 3/4 of target Vout.
Going Farther.
The main problem, and the entire reason for the second version is because the above version is less energy efficient. When regulating for a light load, it shunts some mA of current directly to ground. I will be building this first version soon and I expect no major divergences from simulations as the main sub circuits have been proven elsewhere.
The second version I came up with to combat and correct for the problems of the first. But it too has problems. Here is the schematic of it...
View attachment 67222
The difference and thing to notice with this second version, is that instead of shunting Q1 base current directly to ground to stop Q1 from turning on, I instead block it altogether with a high side PNP switch. This work just as well to keep Q1 off (in simulation), but I have removed the quiescent current associated with grounding R1. The problem this creates is that the point at which Q2 turns on is now determined by Zd1 + Vin. This means to make a target Vout of 5 volts, you need Zd1 to breakdown at 4 Volts AND you need Vin to be exactly 1 volt. This is very not optimal.
The whole point of this circuit is to use as much of the systems battery energy as possible. This means we want to BOTH convert the energy efficiently, AND drain as much as we can from the battery's. So as you can see, each version has a road block that needs to be solved before it is ready for service.
Ideas?
So I'm looking for ideas, I left the first circuit up as a fall back design because it almost certainly works. And because it may be found out that it's actually possible to take it farther than the second one. But I really want to focus on the second version.
With the second version, all current STOPS in the simulator when the regulation kicks in! The only current that is still flowing is the load and a very tiny Zd1+R2 current. When the regulation kicks on in the first version, most every thing stops, but there is still a decent amount of current drain from R1 being shorted to ground. Under light loads, the global power drain is actually dominated by this current. It's only a few mA with the right choice of R1. But if this current can be removed while the circuit is still functioning, it will be nothing but profit. So I'm at least giving it some deserved thought.
I have considered somehow stabilizing the voltage that is fed to the sense half of T1 so that it will not vary with supply. But I think this will add as much or more power loss as the first circuit and it will raise the floor for how low Vin can be. Both these things are very counterproductive. I have also considered and tested the idea of using different parts and configurations for blocking Q1 base current. The PNP seems to be the best choice so far, though the sim isn't very good with MOSFET's. I was also looking into possibly using the negative kickback produced on T1 sense as the voltage feedback instead of Vout. This voltage kickback voltage seems to be directly proportional to output voltage, so it seams just as viable to create a breakdown based control circuit here as it is for how I already have i. The only difference is this area creates a negative voltage that may allow for tricks that wouldn't be possible otherwise.
Conclusion.
Anyway, that's where I'm at now. I would love to hear some input from the elites and hopefully some good advice as how to proceed. If not, I will prolly just run with version number one as it's only major problem right now is efficiency and the fact that it has not been physically built and tested yet.
Thanks for reading.
-()blivion
Last edited: