High Current Constant-Current Source

[Roff]

Thanks for the responses guys! The schematic in Roff's post is exactly what I was envisioning, but was second-guessing myself. Let me see if I have this straight (I reposted the schematic for easy reference).

**broken link removed**

Q2-Q9 each need ~1.4v B-E bias, which is provided when ~14mA flows through R1.

R2-R9 provide parallel transistor compensation. Assuming each transistor is equally sharing the 15A load, that means each transistor is passing 1.875A.

R2-R9, being 0.22Ω will have 0.4125v dropped across each @ 1.875A.

So really, each transistor needs 1.8125v bias (B-E of 1.4v and 0.4125v on R2-R9), which means 18mA will have to flow through R1 to bias each of those transistors (1.8125v/100Ω).

Since the hFe of each PNP is ~1000, that 18mA should be good for 18A total on all transistors.

Do you think I should maybe drop that 100Ω resistor to something like 47Ω to drive the bases a little harder? I just don't want to assume or count on each PNP having 1000 hFe (I know the datasheets says 1000 min gain, but I like to over-engineer).

Also, I see that R2-R9 are 0.22Ω. Can I use 0.1Ω so there is a little less v drop on them? Is there a rule of thumb to determine the "proper" size resistor in this application?

I started by somewhat arbitrarily deciding to use 8 TIP127's because it keeps the current below 2A (1.875A) which is well below the 3A spec for minimum beta=1000. In fact, the beta typically peaks at Ic≈2A (see Fig. 1 in the Fairchild datasheet).
With 15A total collector current, the maximum total base current will be 15mA (min beta=1000). With Vbe typically 1.7V@1.9A (Fig. 2), and another 0.4V across the emitter ballast resistor (0.22Ω), the drop across R1 will be ≈2.1V (simulation says 1.9V). Thus, the current through R1 will be 21mA. Adding this to the maximum base current of 15mA, the Lm317 max current is 36mA. In the sim, the actual total base current is about 4mA, and the LM317 current is about 23mA, due to beta being higher than 1000 and Vbe being less than 1.7V.
You don't "drive" the bases in this circuit. They draw what they need, due to the feedback.
If you make the emitter ballast resistors 0.1Ω, variations between transistor Vbe's will result in more unequal distribution of currents from on transistor to another.

Why only 6A? I would think it would be closer to 10A (15A * 0.707), or am I way off here?

I just looked at the graph of the pulsed current and eyeballed the duty cycle as being about 40%. I just now checked it on the simulator, and it measures 6.1A. 0.707 has nothing to do with the duty cycle here. It is mostly determined by the dropout voltage of the circuit. In fact, when your batteries are fully charged (3.6V per cell), the duty cycle (and average current) drops by a significant amount.

 

[BrianG]

Thanks Roff! That clears things up.

I just noticed something about that schematic that I missed before: Is there any particular reason Q2 is not connected the same way as Q3-Q9? Q2's collector is tied to the output of the LM317, while the collectors of the other PNPs are tied to the other side of the programming resistor.

Seeing as how the average current is "only" 6.1A, I might lower the programming resistor (and add a couple more PNPs) to bring the average current closer to ~10A. Or I may "lightly" filter the rectifier to bring the average up a little.

 

[Roff]

Thanks Roff! That clears things up.

I just noticed something about that schematic that I missed before: Is there any particular reason Q2 is not connected the same way as Q3-Q9? Q2's collector is tied to the output of the LM317, while the collectors of the other PNPs are tied to the other side of the programming resistor.

Seeing as how the average current is "only" 6.1A, I might lower the programming resistor (and add a couple more PNPs) to bring the average current closer to ~10A. Or I may "lightly" filter the rectifier to bring the average up a little.

You can connect all the collectors to the top end of the sense resistor. It will be more accurate. I did it the original way because I figured 0.68Ω is more available than 0.086Ω, which is what you will need. I can't imagine why you would need a battery charger to be a precision instrument.

 

[BrianG]

OK, that makes sense. No, it doesn't need to be that precise current-wise, I kinda plucked 15A charge rate out of my butt. The actual charge rate isn't terribly important as long as it doesn't exceed the safe rate for the cells I want to use, but if too low, it will take too long to charge. The important thing is constant current up to 3.6v/cell. Normally, at this point, a real lithium charger would switch to constant-voltage mode, but the cells I'm planning to use (A123 cells) don't take lots of charge during that phase so I'm just going to terminate the charge then instead. Makes things much simpler.

 

[Roff]

OK, that makes sense. No, it doesn't need to be that precise current-wise, I kinda plucked 15A charge rate out of my butt. The actual charge rate isn't terribly important as long as it doesn't exceed the safe rate for the cells I want to use, but if too low, it will take too long to charge. The important thing is constant current up to 3.6v/cell. Normally, at this point, a real lithium charger would switch to constant-voltage mode, but the cells I'm planning to use (A123 cells) don't take lots of charge during that phase so I'm just going to terminate the charge then instead. Makes things much simpler.

It's not gonna be constant current to 3.6V per cell (43.2V) if you don't have a filtered supply, because the duty cycle drops as the voltage rises. To get constant current at 43.2V, you will need, as you said, about 50V DC. I would use a switcher. Otherwise, the rectifiers and capacitors are gonna be big.

 

[BrianG]

Yeah, I'm seeing that. I'll probably have to filter it then and deal with the number of caps needed and wasted heat. I would prefer a switcher, but would be considerably more complex. The linear circuit is easy and simple enough to use perfboard and uses parts I already have on hand for the most part. If I get too complex, I might as well get a true CC/CV supply.

 

[Ubergeek63]

I need a constant current source able to deliver 15A. I know how to configure an LM317T as a CC source with the programming resistor between the output and adjust pins (and the power required, which in this case is 1.25v * 15A), and use a pass transistor to handle the extra current.

However, I'd like to use multiple PNP pass transistors to help spread the current (and thermal) load. I know that paralleling transistors directly is not advised without compensating for transistor variances by using a small value resistor. Most paralleled transistor circuits I've seen use this resistor in the emitter, but not sure exactly how it would work in this application.

BTW: I seem to have a ton of TIP127 PNP Darlington transistors for some reason, so I'd like to use those if I can.

Edit: Oh yeah, I did try the search, but the only results I seem to get are circuits for LED drivers - I know how to regulate lower currents...

and why not use a chip designed to be a current source as a current source? an HV9910 LED driver would work fine.

 

[BrianG]

Three reasons:
1) That specific chip is only rated up to 1A. I need around 15A. I'm not sure if there is an easy way to boost that current with added FETs/transistors.
2) Linear is simpler, albeit much less efficient.
3) I already have the 317, TIP127s, transformer, heatsink, fans on hand. The only thing I might need to get are some caps, but I think I have enough 1000uF/100v caps laying around to get by.

It would be nice if there was a cheap switcher chip I could use to get what I want.

 

[Ubergeek63]

Three reasons:
1) That specific chip is only rated up to 1A. I need around 15A. I'm not sure if there is an easy way to boost that current with added FETs/transistors.
2) Linear is simpler, albeit much less efficient.
3) I already have the 317, TIP127s, transformer, heatsink, fans on hand. The only thing I might need to get are some caps, but I think I have enough 1000uF/100v caps laying around to get by.

It would be nice if there was a cheap switcher chip I could use to get what I want.

where did you get the idea that there was any current limit on the HV9910? it is a gate driver and the current and voltage limitations are only limited by your NFET (and the 15nC drive limitation at higher frequencies that is easily overcome with and external gate driver)

now if you were looking to limit your inductor size and wanted to complain about the 280nS blanking period and the 150nS delay getting in the way of the higher frequencies, you MIGHT have a point.

there is no actual current rating... they have this 1.4A demo board: https://www.electro-tech-online.com/custompdfs/2010/05/HV9910BDB1.pdf and I personally use it at 1.8A with a 4uS off time.

 

[BrianG]

Well, I did a Google search and found a PDF, and when I read the description, it said "up to more than 1.0A" (last line of first description paragraph). I confess I did not read the whole thing.

Ok, so say I try this route. Since I'm using a switcher to gain efficiency, I might as well use rectified 120V AC line voltage. According to the formulas in the datasheet (assuming 50kHz), I would need a 150uH inductor and around 1000uF 250v capacitor. I imagine I'm going to also need a pretty beefy Schottky diode as well as inductor to handle the current. I don't know, it's a better design to be sure, but I am still leaning to the linear design for simplicity despite the size/weight/inefficiency. If I do try the switcher route, I should probably get a few of everything because I'm sure flames will destroy the first one.

 

[Ubergeek63]

Well, I did a Google search and found a PDF, and when I read the description, it said "up to more than 1.0A" (last line of first description paragraph). I confess I did not read the whole thing.

Ok, so say I try this route. Since I'm using a switcher to gain efficiency, I might as well use rectified 120V AC line voltage. According to the formulas in the datasheet (assuming 50kHz), I would need a 150uH inductor and around 1000uF 250v capacitor. I imagine I'm going to also need a pretty beefy Schottky diode as well as inductor to handle the current. I don't know, it's a better design to be sure, but I am still leaning to the linear design for simplicity despite the size/weight/inefficiency. If I do try the switcher route, I should probably get a few of everything because I'm sure flames will destroy the first one.

40uH ... three of these in series works: IHLP6767GZEB150M01 Vishay/Dale Power Inductors

constant off time: https://www.electro-tech-online.com/custompdfs/2010/05/AN-H50.pdf allows the frequency to vary eliminating sub harmonic oscillation at high duty cycles

just watch the gate charge on the FET as you might need a gate driver in addition to the HV9910

 

[BrianG]

:sigh: It looks like I really need to study/experiment with FETs at long last. I don't usually make circuits capable of currents over 3-5A, so transistors generally work well, but for switching and larger currents, it looks like FETs are the way to go. Looks like gone are the days where I can make RadioShack parts work for those times where ordering online takes too long to experiment on one-off circuits. I probably should start off with a couple simple lower-current circuits to get my feet wet. Are there any good and accurate FET tutorials? I know the net is chock full of reference, but it's hard to know what's right and what isn't.

Thanks for the input Uber!

 

[Ubergeek63]

np, not all that familiar with the ones online, i learned in the pre-net era.

basically you need to get the gate from one point to the other and it hesitates in between as the drain voltage changes. The faster you get it to switch the less power you lose but the more EMI you generate. You get it to switch faster by making more current available to it since the gate is basically a capacitor. the hesitation is the capacitance from the drain to the gait charging as the drain voltage changes. Finally the resistance from the drain to the source dissipates power as the FET conducts.

Dan

 

[BrianG]

The majorty of what I know (or think I know) of FETs comes from dealing with speed controllers for R/C use. I was under the impression that, since the gate was basically a capacitor, it was a "voltage-driven" device (very high input impedance) as opposed to a bipolar, which is current-driven. I know of the rdson value, which is the small resistance present between the D-S when it is fully "on". I also know that switching the gate as fast as possible is desired for less rise-time losses. The FET's D-S rise-time is also a factor in this. I also know they parallel nicely. That's about all I "know". How much to drive the gate, polarity, what precautions to take, etc are the things that are fuzzy.

 

[Ubergeek63]

The majorty of what I know (or think I know) of FETs comes from dealing with speed controllers for R/C use. I was under the impression that, since the gate was basically a capacitor, it was a "voltage-driven" device (very high input impedance) as opposed to a bipolar, which is current-driven.

true ... you charge the G-S capacitance, and as the device switches you have the D-G capacitance as well. it is easier to look at the total charge spec if you are running near the tested values.

I know of the rdson value, which is the small resistance present between the D-S when it is fully "on". I also know that switching the gate as fast as possible is desired for less rise-time losses. The FET's D-S rise-time is also a factor in this. I also know they parallel nicely. That's about all I "know". How much to drive the gate, polarity, what precautions to take, etc are the things that are fuzzy.

not really a heck of a lot else accept that you need a little gate resistance to prevent oscillation and you need to use a fast parallel diode or synchronous rectification to prevent the inductance from blowing out your FETs.

 

[Boncuk]

Hi BrianG,

here's a proper heatsink for your project.

Depending on the number of "tunnels" and fans Rth(K/W) is 0.15 down to 0.05.

Large aluminum fin type heatsinks are Rth(K/W) 0.7 at best.

Here is a datasheet.

Boncuk

 

[Ubergeek63]

Hi BrianG,

here's a proper heatsink for your project.

Depending on the number of "tunnels" and fans Rth(K/W) is 0.15 down to 0.05.

Large aluminum fin type heatsinks are Rth(K/W) 0.7 at best.

Here is a datasheet.

Boncuk

as a curious side note on the linear side, while you realize you are throwing away power, what you might not realize is that the complexity and cost of mucking around with a heat sink add up to more... Rth 0.7 ain't cheap and is in fact useless if the parts are not thermally connected. If you need to electrically isolate them it becomes problematic if your calculation did not include the electrical barrier.

dan