MOSFET-based LM334 regulator

[evan9162]

I've worked a lot with the basic LM334 current regulator circuit. I wanted to get this circuit working with a wide input voltage range, and also be able to handle high current and power dissipation while maintaing 0.1V or less of dropout. After playing around with a few different low saturation BJTs from Zetex, and trying to cascade the base current of the main transistor through another PNP transistor, I found it would be difficult to keep a very low dropout at low voltages, while keeping base current reasonable at higher voltages. I also found that my older design was causing up to 40mA of current to flow through the '334, which is bad, since the part is only rated for 10mA.

So I looked into using a MOSFET instead. I needed a P-channel MOSFET that could handle high current, could dissipate a lot of power, and had the lowest possible Rds with a Vgs of 2.5V or less


I found two MOSFETs that fit the bill nicely - one in SOT223 - the Fairchild FDT434P, and another in D2PAK (TO-263), the Fairchild FDB6021P.

Here's the circuit:

[edit] - this is the new circuit per the discussion below.

I built a low power and high power version.

Here are the two regulators:

The one on the left has the SOT223 part, and is set for ~550mA of current, while the right has the D2PAK part and is set for around 1100mA of current.

That's a quarter for size reference. The larger is 0.4" x 0.75" x 0.4". The smaller, 0.4" x 0.6" x 0.35". The smaller regulator can be made thinner by using a single board instead of two stacked board, and could end up being around 0.25" thick. Both are potted in epoxy for additional strength.

Both have an aluminium plate epoxied to the transistor for additional heatsinking. They both can dissipate between 1-2W alone. The smaller could probably dissipate 3W with additional heatsinking to the plate; the larger, could easily dissipate over 10W with a sufficient heat sink.

Both should work with up to 20V of input.

The zener and 4.7K resistor were added since both of these MOSFETs have a maximum Vgs of 8V, so the zener clamps that voltage, allowing a wider input voltage range.


I've done some measurements of both regulators:

Dropout of the high power regulator is 0.075-0.085V at 500mA and 0.085-0.095V at 1A.
Dropout of the low power regulator is 0.09-0.1V at 500mA and 0.1-0.11V at 1A.



I've been thinking about doing a run of these - though I'm not sure if there's enough interest. Price would likely be in the $5-$6 range. Or I could just let everyone build their own /ubbthreads/images/graemlins/tongue.gif

 

[moraino]

It looks interesting. Could you provide circuit efficiency if you have them for 4 and 8 cells etc. Thanks for sharing.

Henry

 

[evan9162]

[ QUOTE ]
moraino said:
It looks interesting. Could you provide circuit efficiency if you have them for 4 and 8 cells etc. Thanks for sharing.

Henry

[/ QUOTE ]

Being a linear regulator, any voltage above the Vf of the LED is simply dropped and dissipated as heat. So, calculating efficiency is easy - it's Vf/Vin.

If you're driving a Lux III at 700mA with a Vf of 3.5V, with 4 alkaline cells efficiency will start out at 3.5/6, or 58%, and get better as the batteries drain (and voltage drops). The regulator will initally be dissipating 1.75W, and would need additional heatsinking.

With 4 NiMH cells, and assuming 1.25V as an average cell voltage over the vast majority of battery life, then efficiency will be 3.5/5 or 70%. The regulator (on average) will dissipate just over 1W. But, with fresh cells, which can be 1.4V, efficiency would be about 62%, and the regulator would dissipate about 1.5W.


Driving a Luxeon V at 700mA from 8 cells, the efficiency numbers should be identical.


These regulators are best suited for when input voltage is close to Vf of the LED. For instance, driving a Luxeon I or III from 3 alkaline/NiMH cells, or a single Li-Ion cell.

For 3 alkaline cells, efficiency would start out at 77%.
For 3 NiMH cells, efficiency starts at 83%, and would average ~90%. These numbers are about the same for a single Li-Ion cell.

The above numbers would hold true for a Luxeon V and 6 alkaline/NiMH cells, 3 123A cells, or 2 Li-Ion cells.

 

[Doug S]

Darin, I've used a somewhat similar circuit with great success. I'm looking at some pages in one of my lab notebooks where I had tested arrangements where the capacitor is from gate-drain like yours and also gate-source. In my notebook I indicate that gate-source was the preferred solution. With your arrangement gate-drain the Mosfet will be fully on when the circuit powers up causing the LED to be hit with a narrow high current pulse until the regulation takes effect. At anyrate, judging from your choice of C value, you discovered as I did that if C is too low the current will oscillate in a sawtooth pattern [you would not know it without a scope] and the current peak-peak decreased as you raise C towards its minimum stable value. Versions of this circuit that I actually put into the field used a 10:1 ratio of upper to lower resistor [since I did not need as high a Vin range] and also I saw no need for the zener in my app since the Vin max was below my mosfet Vgs max and with the cap placed gate to source the cap provides some degree of ESD protection. With the lower resistor at 10K I found the lower limit for C for stability to be 1.5uf. That's using a temperature stable dielectric like X5R.
On further thought, even if the Vin max exceeds Vgs max in your circuit you don't need the zener. If the circuit is functioning it will prevent the Vgs from rising further than necessary to turn on the Mosfet.

 

[evan9162]

Doug,

I'll have to re-test, but I think I was very close to exceeding Vgs(max) when driving 3 Luxeon IIIs in series, at the point where Vin was just below allowing regulation.

[Edit]

Yep - it looks like it is needed. Driving 3 Luxeon IIIs in series, I measured Vgs in slight excess of 8V (the max for these parts), and the circuit wasn't in regulation yet, so it would have gone even higher. Vgs rises steadily with input voltage up to the point of regulation, then drops rapidly.


Do you think there's any danger of leaving the cap where it is now?

 

[Doug S]

[ QUOTE ]
evan9162 said:
Doug,

I'll have to re-test, but I think I was very close to exceeding Vgs(max) when driving 3 Luxeon IIIs in series, at the point where Vin was just below allowing regulation.

[Edit]

Yep - it looks like it is needed. Driving 3 Luxeon IIIs in series, I measured Vgs in slight excess of 8V (the max for these parts), and the circuit wasn't in regulation yet, so it would have gone even higher. Vgs rises steadily with input voltage up to the point of regulation, then drops rapidly.


Do you think there's any danger of leaving the cap where it is now?


[/ QUOTE ]
Ahh, I had not considered that you were driving a series string. In that case consider using a standard +/-20 Vgs rated part. Most +/-20 Vgs rated parts turn on by 4V or so. If you have two or more lux in series it should do fine. Alternately there are plenty of low Vgs turn-on parts rated at +/-12Vgs.

I don't know about your second question beyond my earlier comment about the brief full DD inrush.

 

[Doug Owen]

Evan,

Well done, glad you decided to 'come out' with this. If you decide to make it a go, count me in for a few.

As I said in our PM exchanges, I fiddled with this idea as well, but settled on keeping the PNP and using it to drive a N channel MOSFET in 'source follower' (LED from V+ to drain). This allows smaller C values, but needs an additional transisor and another resistor.

Since I'm dealing with lower voltages no zener is needed. FWIW, if driving more than one LED (from a higher voltage source, of course) I'd probably just put the LEDs in series with the supply.....

Doug Owen

 

[evan9162]

[ QUOTE ]
Doug S said:
[ QUOTE ]
evan9162 said:
Doug,

I'll have to re-test, but I think I was very close to exceeding Vgs(max) when driving 3 Luxeon IIIs in series, at the point where Vin was just below allowing regulation.

[Edit]

Yep - it looks like it is needed. Driving 3 Luxeon IIIs in series, I measured Vgs in slight excess of 8V (the max for these parts), and the circuit wasn't in regulation yet, so it would have gone even higher. Vgs rises steadily with input voltage up to the point of regulation, then drops rapidly.


Do you think there's any danger of leaving the cap where it is now?


[/ QUOTE ]
Ahh, I had not considered that you were driving a series string. In that case consider using a standard +/-20 Vgs rated part. Most +/-20 Vgs rated parts turn on by 4V or so. If you have two or more lux in series it should do fine. Alternately there are plenty of low Vgs turn-on parts rated at +/-12Vgs.

I don't know about your second question beyond my earlier comment about the brief full DD inrush.

[/ QUOTE ]


yeah, that's a good solution, though I'd have to build separate models for high and low voltage operation with different parts. One of my goals was to try to make a "one size fits all" device. So the first thing I looked for was a turn-on voltage of around 2V or less. The trade-off for that was a lower Vgs(max). For series luxeons, and Luxeon Vs, the higher voltage part would eliminate the need for the zener. Though the zener is not too big of a deal, it only cost 4 cents

If I had an o-scope I could investigate the inrush current thing more. Might have to bump up the priority on getting one.

 

[HarryN]

I have some interest in this for 2 x 123 for 1 x Lux III, and potentially 12 - 15 V driving 3 x Lux III in series.

 

[evan9162]

Doug (S),

I moved the cap between the gate and source, and things tested out good /ubbthreads/images/graemlins/thumbsup.gif

I'll have to see how that change will affect the board layout.

 

[Doug Owen]

[ QUOTE ]
evan9162 said:
Doug (S),

I moved the cap between the gate and source, and things tested out good /ubbthreads/images/graemlins/thumbsup.gif


[/ QUOTE ]

Gents,

Ya got me all mixed up here. Source is connected to the plus rail, a low impedance point, right? Now we're not talking snubber any more but low pass filter driving the gate???

Doug Owen

 

[evan9162]

Doug,

This is how my test rig is wired now:

I tested a single Lux III at 1A up to 6.7V in, and two Lux III in series up to 12V in...no stability problems.

With my limited equipment, I was able to see some high inrush current when applying power with the original design, but wasn't able to reproduce that condition with the cap moved to the new spot.

Are there any problems with the cap acting as a LPF rather than a snubber?

 

[NewBie]

Evan,

During a negative transient, such as happens in an automobile (if you were using it for this environment) during a load dump, the diode in the P-Channel MOSFET would be forward biased, exposing the LED to the negative voltage, which would forward bias many of the ESD diodes found in white, blue, and green LEDs. The current would only be limited by the 0.12 ohm resistor, and I've seen -40V load dumps before.

So if you desire to utilize it in an automotive environment, you might think about it...

 

[MrAl]

Hi folks,

With regard to the capacitor, if i rem right the circuit
involving the LM334 is basically an oscillator until
you add the cap, which acts as an integrator to stabilize
the circuit (which then provides a smooth output dc).

Take care,
Al

 

[Doug Owen]

[ QUOTE ]
evan9162 said:

Are there any problems with the cap acting as a LPF rather than a snubber?

[/ QUOTE ]

Not if it works. It's hard to argue with success, and I always go with 'the guy that's been there'.

It's just that this points to an entirely different problem than I think most folks are thinking about. It doesn't have anything to do with the stability of the transistor as such, but feedback involving the 334.

What you've done is slow down the response of the amplifier to it can't 'chase it's own tail fast enough'. The cap should work equally well from gate to *ground*. The snubber on the PNP assumes the transistor itself is unstable and couples the output (from the collector) to the input (base) 180 degrees out of phase to counter it. In the process it too lowers high frequency response.

As it is, the cap integrates the rapid changes in input to the gate. If this does the job, it does the job I guess. But I think I need to rethink how I'm driving the N channel MOSFET.....

Doug Owen

 

[evan9162]

[ QUOTE ]
NewBie said:
Evan,

During a negative transient, such as happens in an automobile (if you were using it for this environment) during a load dump, the diode in the P-Channel MOSFET would be forward biased, exposing the LED to the negative voltage, which would forward bias many of the ESD diodes found in white, blue, and green LEDs. The current would only be limited by the 0.12 ohm resistor, and I've seen -40V load dumps before.

So if you desire to utilize it in an automotive environment, you might think about it...

[/ QUOTE ]

Definitely something to keep in mind. I don't think I'll address that in this design, I'm more targeting battery driven/DC power supply driven applications. Adding a diode to the input side should take care of that though.

 

[OddOne]

I'm looking into current regulation with an LM338 myself, and as it turns out your circuit is largely similar to what I've been coming up with.

My target current is a bit higher though - 2.5A to 5A with a 24VDC source - so my design is a bit bigger. (TO-220s on a heatsink plate handle much more heat than TO-263s on a PC board, additional aluminum notwithstanding.)

oO

 

[evan9162]

[ QUOTE ]
OddOne said:
I'm looking into current regulation with an LM338 myself, and as it turns out your circuit is largely similar to what I've been coming up with.

My target current is a bit higher though - 2.5A to 5A with a 24VDC source - so my design is a bit bigger. (TO-220s on a heatsink plate handle much more heat than TO-263s on a PC board, additional aluminum notwithstanding.)

oO

[/ QUOTE ]

With the larger package, and higher working voltage, you'll have no problem finding parts with extremely low Rds(on), and high power handling.

The TO-263 isn't actaully on a PC board - the tab is thermal epoxied to the aluminium bar, so it can handle quite a bit of heat when attached to a larger heat sink.

Are you trying to drive a Lamina cluster?

 

[OddOne]

Yep, a Lamina BL-3000 is the target, and by adding a second sense resistor and MOSFET in parallel to switch it in and out of circuit with the existing, I can provide an "overdrive" option by cutting the sense resistor's effective value in half. (I've used that trick before - makes for very easy variable current control.) /ubbthreads/images/graemlins/grin.gif

Now that I think about it, nothing says you couldn't use a TO-220 packaged MOSFET sticking out from the board and use the board as a glorified connect point for it. Be super-flexible that way AND have great power handling and heat dissipation. /ubbthreads/images/graemlins/grin.gif

oO

 

[evan9162]

[ QUOTE ]
OddOne said:
Yep, a Lamina BL-3000 is the target, and by adding a second sense resistor and MOSFET in parallel to switch it in and out of circuit with the existing, I can provide an "overdrive" option by cutting the sense resistor's effective value in half. (I've used that trick before - makes for very easy variable current control.) /ubbthreads/images/graemlins/grin.gif

Now that I think about it, nothing says you couldn't use a TO-220 packaged MOSFET sticking out from the board and use the board as a glorified connect point for it. Be super-flexible that way AND have great power handling and heat dissipation. /ubbthreads/images/graemlins/grin.gif

oO

[/ QUOTE ]


Yep, the only thing to watch out for is that usually the drain of the MOSFET is connected to the tab, so you'd have to make sure you weren't shorting the output to ground. When I epoxied the TO263 to the aluminium plate, I made sure it was electrically isolated, so there's no problem attaching these regulators to a grounded heat sink.