Constant current source review

[superfrog]

Sorry to be pushy, I'd love some feedback on this. This is prety much a repost from the end of my previous thread.


After going over pretty much every post in the forum relating to constant current sources, I found https://www.electro-tech-online.com/threads/high-power-led-controller.95832/ which has some interesting design ideas.

I slightly extended the idea to work over the range I am interested in, ie rectified AC, in the following schematic. Is the design correct?


**broken link removed**

The zener voltage could be reduced to accommodate maximum gate voltage,
R8 would be used as sense resistor to create drop Vref at 20mA in my case but I guess R = Vref / I in the general case.
At 20mA with Vref being 2.5V, Heat losses in the sense resistor are .05W, which is acceptable.

I guess R6 and R7 need dimensioned, and I probably need advice on that, as I believe they are actually the critical bit of the design, in term of responsiveness and stability.

With R6 at 300k, I should get a bit less than 1mA at 300V , which feels like enough to feed the gate. (and I happen to have a reel of 1206 100k that could be put to good use there)

R7 should limit the rise speed, I am not sure of which value would do, but i suspect it is very much a wet finger estimate type of thing. Maybe an extra 100k?

Would that work?

Thank you in advance for your answers.

 

[MikeMl]

To create a constant-current source where the TL/LM431 controls the gate voltage as you show, the FET has to be an NFet.

What is the voltage at "PWR-FLAG"?

How many LEDs will be in the string?

 

[superfrog]

Rectified 240AC, so 340 ish, the bunch of led are just to represent the load.
In this case it could be anything from 1 to 120 (leds being up to 2.5V,20mA).

In all likelyhood, it will be for testing small led modules of up to 40-60 units, so most of the power will go to the mosfet heatsink.

In the longer term, this will be replaced by a buck-boost smps, so efficiency is not a concern for this design. My mosfet can easyly whistand the full possible wattage(a tad less than 7W) at 20mA.

Indeed the mosfet is the wrong one here, I got confused by the arrow direction that looked right to me on schematic, I am not used ot this notation. My bad.

EDIT: UPDATED SCHEMATIC

**broken link removed**

 

[MikeMl]

Rectified 240AC, so 340 ish, the bunch of led are just to represent the load.
In this case it could be anything from 1 to 120 (leds being up to 2.5V,20mA).
...

I simmed it, and found a few interesting things. The Zener and its current limiting resistor are not needed. The LM431 shunt regulator does just fine in limiting the NFET's gate voltage. Adding the capacitor provides transient pulse protection.

The power dissipations are of most interest: The DarkGreen trace is the dissipation in the LM431. The Purple trace is the dissipation in the NFet (average = ~4W). The dissipation in R8 is the Gray trace (average = ~ 550mW, so it needs to be a 1W resistor).

 

[Mr RB]

I think the design is fundamentally flawed in that there is no need to power any of the control circuit from the high side (with the big V drop losses).

 

[MikeMl]

I think the design is fundamentally flawed in that there is no need to power any of the control circuit from the high side (with the big V drop losses).

With only 4 LEDs in the sim it wouldn't matter much if R8 was connected to the drain of the FET instead of the rectified line. If there were ~50 LEDs, the power dissipation in the FET and R8 would be significantly reduced.

Or do you have someplace else in mind from where to power the Gate circuit?

 

[superfrog]

thank you for the analysis.
I shall try to get myself to use spice, or rather gnu-cap in my case. I did that a few years back but can't remember any of it (apart for the fact that the syntax was only second to TeX, in the ranking of the most horrible declarative languages). On top of that, in most cases the spice circuit definitions were generated from other forms of design for me so I could probably count of my fingers the circuits I have actually described in spice.

I should have said before, I pulled the input stage from a pc power supply (which is pretty much what I have in the next schematic). The input should be pretty close to constant.

**broken link removed**

The transformer is not one but rather coupled inductances, and the first inductor is really a choke, but I could not find symbols for it in kicad. I know I need a bleeder resistor for the capacitor as the current circuit is extremely dangerous with the 680uF electrolytics charged at 320V (and staying there for over a week). X caps between live and neutral, Y caps between either of the previous and ground, as well as for the mysterious capacitor from the rectified negative to ground. (probably to keep negative close to ground, I'd love to know what it is for). 1 amp fuse, slow. No idea of the characteristic of the inductors, but I do not see how too big can be a problem here.


That said, your schematic is so much better than mine that I might just use it, sure the LEDs are not on all the time, but I do not really care. I misinterpreted the maximum voltage for the TL431, it makes sense that 37V is the maximum regulated voltage, rather than the maximum voltage beetween anode and cathode.
The zener could still be of some use to avoid Vgs to go too high in the case of an open load, as it is in practice limiting Vg to the zener voltage.
For R8, I still think that 200-300k as multiple 1206 should save me, and keep me in the thermal envelope of the packages (1/4W, but I need to check).


Thank you so much again for that. It is probably mundane circuitry to you, but it teaches me hell of a lot.
Any chance you could provide the spice file as asci? I would be most grateful for that. And I could tinker with the params and simulate on my side.

Ben

 

[MikeMl]

Watch that you don't wind up with ~300V across a 1206. You need to check the max voltage allowed across a chip resistor.

 

[superfrog]

Can I reiterate my demand for the spice netlist from your sim.

You are very right, max operating voltage is 200V , 400V peak.
That is why I was thinking along the lines of 300K as 3 in series. I should receive a pack of 400V resistors before the end of the week anyway. I might just wait.
I got a partial reel of 1000 100K resistors for 2£ off ebay, so in the worst case, I will not be too sad if they go to resistor heaven.

I think I just got what Mr Black said earlier, and take the opportunity to also thank him.

The drain net could be use as the power input of the gate circuit. It would indeed reduce the consumption trough the resistor, specially when the voltage actross the nmos is smaller (while beeing the same in the general case).
This would solve the problem of this circuit, that will set the nmos to fully conduct if the load is open, and then only adjust the voltage to the gate to get the desired conductance for the gate.

I believe I would add yet another zener in series between load and nmos to make sure sufficient voltage to adjust the gate is present when a load needing close or more to the maximum regulation voltage is present as I believe that would be a saner behaviour. I would have to check but I believe for the mosfet gate I selected, a 5 volts zener should be enough.

Pardon me for stuffing zeners everywhere, I discovered their existence very recently and find them extremely useful.

I ll think about it and botch another diagram tomorrow.

Cheers,
Ben

 

[Mr RB]

With only 4 LEDs in the sim it wouldn't matter much if R8 was connected to the drain of the FET instead of the rectified line. If there were ~50 LEDs, the power dissipation in the FET and R8 would be significantly reduced.

Or do you have someplace else in mind from where to power the Gate circuit?

Yeah I got all that before I said it. But if he's only powering a couple of LEDs from mains with a resistive CC it's fundamentally flawed again, just in a different way.

Anyway it looks like he's going to a transformer driven design now so that fixes that.

 

[MikeMl]

I just thought of another serious problem that the previous sim didn't model. With only four LEDs in the path, and with the 330V peak input, the voltage at the drain of the NFet goes much too high. You would have to buy a super high voltage FET!

 

[kchriste]

What's wrong with simply using the proper resistor (Ohmage and Wattage) to limit the LED current? Or use the Xc of a cap to conserve power dissipation. These schematics, from QSLnet illustrates the idea (You'll need to mod for 240Vac):
**broken link removed**

 

[indulis]

A note of caution when using 431's... place a cap from the cathode to the adjust pin even in a DC application just to insure it behaves. If you look at a simplified schematic of the device, you'll find an opamp in there.

 

[superfrog]

Taking in account the comments from all replies

**broken link removed**

The mosfet is rated 600V, and shoud do the trick.
I ll leave the 15V zener to clamp the gate voltage, in case of overload.

I added a Capacitor, and added a resistor to avoid keeping the gate voltage high if the load is disconnected.

r9 shall be 200kΩ and r10 shall be 100kΩ, r12 1.2MΩ, c10 100nF(-20/+80), which give a half second 95% capacitor discharge and keeps leaks reasonable.

D24 makes sure that 5V head is available to keep the gate in control when in overload.

What is more amusing is that I did a BOM for this circuit, and it is way under a pound, about a pound with an extra bridge rectifier, about 2 if you add a 33uF electrolytic cap to filter the bridge (which should be enough to keep variation less than 10V up to 60mA, and it is actually as 2 200V caps in series which is cheaper)

Add a scrap heatsink, and you have a linear current limiter.
I am quite impressed. I did not count the disc capacitor to stabilise the gate, but this cannot be too expensive.

Now I believe is the time to start trying. I ll get a prototype working on 16V unregulated psu (without the 15V zener, and with R9 0Ω), to test the concept, and I ll see how much stuff I can have self combust tonight.

If it works, I ll just be courageous, hope that royal mail deliver my resistors, and implement the 240VAC version shortly.

Thank you guys.