Another Mod (regulated SatCure circuit)

[Steelwolf]

I recently completed another mod. It was based on this circuit and this circuit. They are both similar in that they are based on the SatCure circuit, but have added an extra transistor to regulate the output.

With any boost circuit like this, one of the most important components is the transformer. In this case, I used a minature toroid as a core. The item is Cat No. R5400 from **** Smith's Electronics and is described as:
"F43 Material - for RF suppression (30MHz-400MHz) in Amplifiers, etc. AL=520nH.
9.5mm(OD) x 5mm(ID) x 5mm(L)"

BTW, for those in the know, what exactly is F43 material? I used to use even tinier toroids which were made of F14 material and which were also described as meant for HF/VHF operation.

I wound the core using 0.25mm wire in the prescribed fashion to achieve 300uH in both coils. That took 22 turns. Other components were 2 2N2222A transistors, an output electrolytic cap of 220uF 16V, an input tantalum cap of 47uF 6.3V and 9 5mm 5600mcd 20deg LEDs wired in 3 banks of 3-in-series.

During testing utilizing a lab power supply, I ran the circuit between 6V and 0.7V. It appeared to the naked eye that regulation was achieved between 3.5V and 6V. This is higher than shown in the chart on the first webpage but should be expected since I am running 9 LEDs to his 6. I didn't really dare to go higher than 6Vin, nor did I feel the need to since I was testing the circuit for future use in a 4AA mod. Perhaps next time. /ubbthreads/images/graemlins/smile.gif

Power draw was propotional with reduced current draw for increased voltage. 6V-0.10A, 5V-0.11A, 4V-0.13A, 3V-0.14A, 3.5V-0.15A. Below 3.5V, it appeared to go out of regulation and power draw diminished, ie, reduced volts reduced amps. 3V-0.14A, 2V-0.10A, 1V-0.05A. It glowed all the way to 0.5Vin but only turns on again at 0.6Vin. Unfortunately, I soldered everything in place before thinking about checking the output power, so no numbers from there, suffice to say that each LED was glowing as bright as if they were individually powered at a nominal 3.6V 30mA.

I wonder if this behaviour is consistant with similar circuits that others have made before? Will reducing the number of LEDs also reduce the point at which the circuit falls out of regulation? Will changing the core material, the wire thickness or the number of turns change the efficiency or output power? Would it have been better to wire the LEDs completely in parallel or in series or in another configuration? (Actually tried this step during breadboarding and didn't notice any change in brightness between 6 in series or 6 in parallel.)

Obligatory beam shots and pics to follow.

 

[Steelwolf]

Photos of the mod can be found here. Will have to wait until it gets dark. Anyway, it is quite difficult to take beam shots as the spread is quite wide.

 

[dukeleto]

I played around with a 30-LED array a year ago for my bike; after many attemps at making a reasonable current regulator for all the 30 simultaneously (// or series, I tried both!)
off varying (albeit low) voltages I ended up powering the array of 4 separate convertors, apparently for the same reason as you: one single circuit went out of regulation too quickly... I don't really know why /ubbthreads/images/graemlins/thinking.gif
with 4 convertors it was great! Car drivers were occasionally very intrigued /ubbthreads/images/graemlins/grin.gif

 

[Steelwolf]

There was another guy who did a Mag mod using 3 separate LS and micropucks. I might have used a few separate boosters, but I didn't think it was necessary for so small a project. Afterall, I'm only providing about 1W to the LEDs. By all calculations, the manufactured components should be able to take a little more. It's just the inductor I'm worried about.

I tried running a SPICE simulation for the circuit I actually built (as oppposed to the circuits shown on the 2 websites) but according to the simulation, the LEDs shouldn't light up or should have blown up already. Huge spikes all over the place, and negative spikes too, and the mean current to the LED array never goes above a few microamps (not even milliamps).

 

[Doug Owen]

A couple of thought come to mind on this most interesting circuit. First off, I'm not sure 2N2222s are a good choice. At the kinds of Ic you're running they're running out of gain rapidly, keeping them saturating as they heat up (you're on the wrong side of the curve already) could be problematical. It could quite literally burn up on you.

Mr Al's suggested Zetex parts might be a better choice? This just doesn't seem like a 'general purpose' application.

The overall efficiency would be interesting to know. For sure if you really have a circuit that will drive nine 3.6 Vf devices at 30 mA each (.972 Watts) from a 3.5 Volt 150 mA source (.450 Watts) PM me and I'll make you rich!

Doug Owen

 

[e=mc²]

I have dabbled with transistorized circuitry as I cannot solder SMD components such as Zetex or Maxim brands offer. Does anybody know if either of these two mfrs. make regulator/boost chips that have similar .10" pin spacing as in DIP packages. That's about as small as I can resolve with these aging, diabetic eyes 'o mine.
BTW, with two transistors and plain old RF chokes, (much easier than hand winding ferrite beads but prob. less efficient) I have gotten runtimes on a single AA cell in excess of 72 hrs with 9600 mcd 5mm Nichia white LEDs. Unregulated, mind you, but I love the runtimes that I receive. My circuit is based on the Brinkmann Long Life LED light, changed the choke value from 81 to 100 micro henries, and the base resistor I have lowered to 1.2K ohms. I use 2N4401/2N4403 PNP/NPN pairs as the Brinkman circuit uses that approach.

Ed.

Ed.

 

[Steelwolf]

Hey Doug, You're right. I never even checked my figures. /ubbthreads/images/graemlins/blush.gif Those figures are impossible. But the LEDs are bright. Looks like I'm going to have to desolder at least 1 LED and try and find measure current and voltage and all that. I definitely made a mistake somewhere. My bad. /ubbthreads/images/graemlins/twakfl.gif /ubbthreads/images/graemlins/whoopin.gif

 

[Steelwolf]

[ QUOTE ]
Doug Owen said:
First off, I'm not sure 2N2222s are a good choice. At the kinds of Ic you're running they're running out of gain rapidly, keeping them saturating as they heat up (you're on the wrong side of the curve already) could be problematical.

Mr Al's suggested Zetex parts might be a better choice? This just doesn't seem like a 'general purpose' application.

Doug Owen

[/ QUOTE ]

Correct me if I'm wrong. I know the circuit suggests 2N2222 or 2N4401 mainly because he was looking at only 1 LED (maybe a maximum of 3), so the current draw isn't very high. That's where I began. But I'm using 2N2222A, which I found while searching for the 2N2222. I was given the impression that this was capable of handling much higher currents, that is why I'm using them. Another point that drew me to them was that they are meant for high speed switching.

Anyway, I just bought a couple of 2N3019 and 2N3904 transistors. The 2N3019 is supposed to be able to handle heaps of current as it is supposedly used in high frequency power amplification. The 2N3904 is labelled as a low level amp. I don't know which were the Zetex parts suggested by MrAl and I can't get them easily in Oz either. (No local stores stock them).

As for the mistake in my circuit, the signal for the base of the second transistor is tapped from only 1 LED line, but I have all 3 lines hooking up to a common ground from which the signal is then tapped. So perhaps I'm not driving the LEDs at the full current, since the base is getting the current from all 3 lines and so would be cutting in earlier than expected. I'm going to try and fix that and see what happens. The simulation doesn't show whether this was the problem, but it is likely.

And still I would like to know, how to choose core material for these flyback transformers. What is F16, F19, F29, F43 material? Between an Iron powder core toroid of "AL=52nH" and an F43 toroid of "AL=520nH", which is the better core for this purpose?

 

[Doug Owen]

[ QUOTE ]
e=mc² said:
I have dabbled with transistorized circuitry as I cannot solder SMD components such as Zetex or Maxim brands offer. Does anybody know if either of these two mfrs. make regulator/boost chips that have similar .10" pin spacing as in DIP packages. .

[/ QUOTE ]

Actually the excellent Zetex transistors can be had in 'TO92 like' .10 inch pitch 'transistor packages' for under a buck. I bought mine from Digikey.

If you want to screw around with SMD ICs, I'd suggest you investigate 'surf boards' a very small 'flag' PCB that converts the surface mount part to .10 inch pitch pins.

Doug Owen

 

[Doug Owen]

[ QUOTE ]
Steelwolf said:
Looks like I'm going to have to desolder at least 1 LED and try and find measure current and voltage and all that.

[/ QUOTE ]

It should be fairly easy to measure the voltage 'on each side' of one LED string from V minus. The voltage across the resistor (lower LED voltage) will give you the current by ohm's law, the difference the voltage delivered to the load (LEDs). Divide power out (Vled times Ires) by power in (Vbat times Ibat) and you'll know. No desoldering required.

FWIW, one of the guys you cite claims 65% efficiency *with a special transistor* and lower output levels. IMO you'll be very lucky to see 50% with this circuit and 'garden variety' bipolars driving that many LEDs.

Still, a very interesting circuit.

Doug Owen

 

[Doug Owen]

[ QUOTE ]
Steelwolf said:

Correct me if I'm wrong. I know the circuit suggests 2N2222 or 2N4401 mainly because he was looking at only 1 LED (maybe a maximum of 3), so the current draw isn't very high. That's where I began. But I'm using 2N2222A, which I found while searching for the 2N2222. I was given the impression that this was capable of handling much higher currents, that is why I'm using them. Another point that drew me to them was that they are meant for high speed switching.

Anyway, I just bought a couple of 2N3019 and 2N3904 transistors. The 2N3019 is supposed to be able to handle heaps of current as it is supposedly used in high frequency power amplification. The 2N3904 is labelled as a low level amp. I don't know which were the Zetex parts suggested by MrAl and I can't get them easily in Oz either. (No local stores stock them).

As for the mistake in my circuit, the signal for the base of the second transistor is tapped from only 1 LED line, but I have all 3 lines hooking up to a common ground from which the signal is then tapped. So perhaps I'm not driving the LEDs at the full current, since the base is getting the current from all 3 lines and so would be cutting in earlier than expected. I'm going to try and fix that and see what happens. The simulation doesn't show whether this was the problem, but it is likely.

And still I would like to know, how to choose core material for these flyback transformers. What is F16, F19, F29, F43 material? Between an Iron powder core toroid of "AL=52nH" and an F43 toroid of "AL=520nH", which is the better core for this purpose?

[/ QUOTE ]

If you look at the curves for the 2222 you'll notice that gains at the fairly high current levels you're using is low and goes *down* with temperature (this means it gets worse as it heats up). Remember the real current is much higher than the average since it's off a lot of the time. 2222s are good GP switches, but *at lower currents*. You really do want high gain, low Vsat parts that are intended for the current levels you're running. The fellow's claim of 65% efficiency on a more modest circuit *with a more appropriate transistor* is probably quite real.

Bummer about the Zetex parts, they're Brit you know, you'd think you could get them easier than I?

You didn't make any mistakes I know of, the current through each string should be the same as the 'base string'. Set it to say 25 mA (28 ohm sense resistor) and you'll get 25 mA *per string*. At least that's how I think you've done it.

In my experience 'softer' cores (lower F numbers) tend to work best in this sort of blocking oscillator. YMMV as they say.

Doug Owen

 

[Steelwolf]

[ QUOTE ]
You didn't make any mistakes I know of, the current through each string should be the same as the 'base string'. Set it to say 25 mA (28 ohm sense resistor) and you'll get 25 mA *per string*. At least that's how I think you've done it.

[/ QUOTE ]

What I was trying to describe was that I joined all the strings before the tap. So the current that the transistor was sensing was the total through all 3 strings, instead of the current through 1 string.

I've made the fix. You know, I never even thought about measuring the output the way you mentioned. Shows you what a well trained technician I am. But I still must not be doing something right.

Power in is 3V 0.15A = 0.45W

The LEDs show 9.52V across a bank of 3 LEDs. I used a 22ohm current limited resistor, which shows 0.2V across it. This means I have only 9mA flowing through the LEDs? But they are not that dim!

However the power out looks about right. 9mA through 1 bank = 27mA through 3 banks. Power out is about 0.26W. About 57% efficiency. I suppose this is barely acceptable as a boost circuit? I think running the original bulb would have been as efficient.

 

[Steelwolf]

Doug: Since many parts easily available in USA are not available here (namely the Zetex stuff), except from online merchants and bulk suppliers, I've been trying to find suitable substitutes. But I'm not quite sure I understood you correctly. Should I be looking for higher Vce(sat) or lower? Gain is labelled hfe right? Should that be high or low? I'm guessing low Vce(sat), high gain especially at the currents I'm working at. And also I'm guessing a high Ic should be good too?

I was looking at the specs for the BC337. It seems to be comparable to both the Zetex FMMT617 and ZTX650, which were the Zetex transistors mentioned in some similar circuits. Would it work?

http://www-us.semiconductors.philips.com/acrobat/datasheets/BC337_3.pdf

 

[Doug Owen]

[ QUOTE ]
Steelwolf said:
Doug: Since many parts easily available in USA are not available here (namely the Zetex stuff), except from online merchants and bulk suppliers, I've been trying to find suitable substitutes. But I'm not quite sure I understood you correctly. Should I be looking for higher Vce(sat) or lower? Gain is labelled hfe right? Should that be high or low? I'm guessing low Vce(sat), high gain especially at the currents I'm working at. And also I'm guessing a high Ic should be good too?

I was looking at the specs for the BC337. It seems to be comparable to both the Zetex FMMT617 and ZTX650, which were the Zetex transistors mentioned in some similar circuits. Would it work?

http://www-us.semiconductors.philips.com/acrobat/datasheets/BC337_3.pdf

[/ QUOTE ]

Steelwolf,

Yes, your sorta on the right track. High gain (Beta fr hfe typically, but the ratio of Ic to Ib) *at the currents used* and low Vbe (sat) are key specs. What you need is a 'saturated transistor switch' not a 'general purpose switching transistor' or worse still a GP part intended for linear use like these, 3904s, 2222s and so on, since typically high speed switching transistors are *never* allowed to saturate (slows them way down). It's a complex deal, having to do with the number of available carriers and the physical sizes of the parts of the transistor structure.

Vsat is very important. Consider that at the very low supply voltages you're working with, the spec sheet you suggest shows this value to be 700 mV at half an Amp (realistic peak currents for the average 150 mA you're running at, sure to get worse as you try to get more light out) in a *seriously overdriven* state (forced gain of ten, note Ic/Ib is only ten?).

Note from the graphs that things are 'falling rapidly off' as currents go up (and it sounds like you need to raise yours), you need more 'margin' (you need to be further away from the drop off).

Also note not all 337s are the same, if possible you at least want the 'hotter dash numbers'. Such parts should be considerable improvements over 'the plain vanilla' part for your use.

Contrast this with the data from the 617:

Zetex transistors

Note that Vsat is typically 70 mV, max 100 at one Amp and a forced gain of 100. This is ten times better performance, at higher current with one tenth the drive. I trust you can see how these are not "comparable"? This is the sort of part I suggest you might want to look for.

If you think about it, I'm sure you'll understand how important this is. If you only have 1.5 Volts to start with 'leaving half of it across the transistor' (rather than the transformer) automatically condemns you to 50% or less efficiency. Even though this is a simple circuit in terms of parts count, the demands on the transistor are not at all simple. General purpose devices are almost sure to be poor choices.

Notice that one of the authors of the sites you point to discusses how important this is and suggests a source for the special sort of transistor you want, camera flash circuits. The flash needs just this sort of transistor to run the converter to raise the few volts in the battery to the hundreds needed by the flash. I suggest you look into salvaging this part from a disposable camera flash?

Doug Owen

 

[Steelwolf]

Doug, thank you so much for your help. I'm starting to see what I should have been looking for in a transistor. Essentially I want to move away from high speed switchers (which I thought was needed because the circuit is supposed to oscillate at MHz rates). Low Vsat (Vce sat, Vbe sat or both? Probably both is better?) at high amps, and high Ic to Ib ratios.

So the search term I want is "power transistors"? Another term I've heard associated with "power transistors" is Darlington (I think). What are those? I can't use MOSFETS or HEXFETS right? (These came up in the search.)

So far, with all the transistors in my parts box, the 2N2222A transistors are still giving me the best run. /ubbthreads/images/graemlins/tongue.gif But then again, all I have are high frequency amplifier transistors. Crap.

 

[vicbin]

[ QUOTE ]
Steelwolf said:
Doug: Since many parts easily available in USA are not available here (namely the Zetex stuff), except from online merchants and bulk suppliers, I've been trying to find suitable substitutes. But I'm not quite sure I understood you correctly. Should I be looking for higher Vce(sat) or lower? Gain is labelled hfe right? Should that be high or low? I'm guessing low Vce(sat), high gain especially at the currents I'm working at. And also I'm guessing a high Ic should be good too?

I was looking at the specs for the BC337. It seems to be comparable to both the Zetex FMMT617 and ZTX650, which were the Zetex transistors mentioned in some similar circuits. Would it work?

http://www-us.semiconductors.philips.com/acrobat/datasheets/BC337_3.pdf

[/ QUOTE ]
Steelwolf,

Did you try 2SC2500 ? I think it is easier to find than Zetex part, especially in this region (Asia-Pacific).

Although it can't beat FMMT617 but still it has low VCEsat, @2A typical 200mv (max. 500mv) and also the ft=150Mhz

2SC2500 Datasheet (PDF)

Vic

 

[Doug Owen]

[ QUOTE ]
Steelwolf said:
Low Vsat (Vce sat, Vbe sat or both? Probably both is better?) at high amps, and high Ic to Ib ratios.

So the search term I want is "power transistors"? Another term I've heard associated with "power transistors" is Darlington (I think). What are those? I can't use MOSFETS or HEXFETS right? (These came up in the search.)


[/ QUOTE ]

That's the ticket. I wouldn't worry about Vbe, it's Vce we need to get as low as possible.

This is medium power, really. Power transistors tend to have lower gains, bigger total device dissipation than we want. Darlingtons are out. That's a scheme of cascading two transistors so their gains multiply (which is good) but Vsat is typically never under a volt or so (which is bad). FETs can also work, but not in this circuit.

You're looking for a 'saturated switching transistor', of they type typically used in switching power supplies (which after all is what you're making, right?

Doug Owen

 

[Doug Owen]

[ QUOTE ]
vicbin said:
[Steelwolf,

Did you try 2SC2500 ? I think it is easier to find than Zetex part, especially in this region (Asia-Pacific).

Although it can't beat FMMT617 but still it has low VCEsat, @2A typical 200mv (max. 500mv) and also the ft=150Mhz

2SC2500 Datasheet (PDF)

Vic

[/ QUOTE ]

Looks like a likely candidate to me. Can you get them?

Otherwise, like I said, I'd consider hacking into a disposable camera......

Doug Owen

 

[Steelwolf]

The 2SC2500 is not available from my usual source. I'm just going to have to figure out some way to get the Zetex parts, I guess.

Yeah, so far, I've been investigating every single transistor available from my usualy source and none seem to fit the bill. High gains have high Vsat. Low Vsat mean moderate gains, but the Ic is also correspondingly low, too low for this purpose.

Is it possible to harness 2 transistors together? Either in a parallel configuration, similar to how one would increase the current-carrying capacity of resistors, or to use a smaller transistor to switch on a larger transistor?

 

[Steelwolf]

Oh wait... that second point about using a smaller transistor to switch on a larger transistor... that's a Darlington isn't it? Won't work then. *sigh*