If I was going to drive 9 x nubm44's I would regulate the current not the voltage so I would regulate @ 4.5a and make sure there was 42 volts available although it would likely draw closer to 41v @ 4.5a
Typically I run a nubm44 at 4.5a and it will draw about 4.5v however the factory data projector drives at a lower current that's sustainable with active cooling, it also adjusts for demands and compensates for wear over time, that's why we can push them so hard, there is a lot of headroom built in.
Anyway if you regulate current the diodes will draw the voltage they need, so you want your driver to be able to supply it, also laser diodes don't like spikes or noise so you wouldn't want to use just anything, actually a linear would make sense for 9 x nubm44's in series and would be pretty easy to make.
Yea they are wired in series so for sure 30V-36V is probably the drop and I would bet they don't go over 3A. They have to be able to compete with the life of LED sources to be competitive in the market so it means running them way below their potential or where previous generation dodes which usually were only rating a life of 500-5K hours. You can look at the power tests and see the difference in the percentage back of foldback that is on the datasheet for the rated currents on previous generation diodes compared to thess. Also maintaining specific wavelengths is very important as they tweak the ratios trying to get the best color balance with the limitations of the system which has been hit and miss for hybrid as you have to have the same blue for the blue and the conversion of blue to green which I think is one reason for after a bit having two types in the array(a-type m-type). Some go for all laser color conversion which probably gives a little better control of the color but more loss weather they pass the blue or convert the blue to slightly different blue as well as converting the green and red from the blue. Anyway current and temp is important to maintain that especially in an environment where you may run it continuously for hours or event days. I remember after just 20 minutes you could fry and egg on the sink in an A140. Burnt the crap out of my hand and not sure if that was a engineering feat for the small form factor units that has improved and is not as much of an issue in the larger units.
If you are still planning on using that 3S lipo pack or packs, you may have to contend with spikes as high as 30 volts p-p. They are highly inductive and when the current goes from max to 0, this is when these spikes appear. It has been shown several times from what I have been able to ascertain that buck drivers fail using these lipo packs because of these spikes.
x RedCowboy and Paul: I received another interesting reply from the creator of the BlackBucks drivers, x-Wossee.
About the voltage spikes:
"Most likely You meant inductive voltage spikes that come from mechanical switching ringing.
I see one simple solution for this: an electronic power switch or/and a TVS diode in the power circuit.
About the diode in series:
"This idea is great as long as we do not use the calculator Why?
Vf of NUBM44 @5A = 4.8V, 9x diodes = 43V;
Operating voltage of Li-ion cell is from 4.2 to 2.8V;
You need to have a minimum of 45V from the battery when they are almost discharged, i.e. 16 cells.
When 16 cells are charged they give 67V. Your linear driver will emit 110W of heat at this voltage at 4.5A
of operating current! Are you ready for this?
In addition, the LM338 is not tolerant to 67V.
Finally, the cost will be much higher than 9 of the usual buck drivers "
I have seen the TVS diode recommended in many places as a possible remedy for these spikes. But, it one engineering blog, it was tried with less than satisfactory results. It appears that in addition to the diode you need some high value caps located close to the driver. Don't know how well this works as it wasn't confirmed. My thoughts are to dispense with the lipo packs altogether.
So, I'm in a dead end.
Also putting 3 or 4 diodes in series with 3 or 4 LM338 will generate a lot of heat to dissipate and complicate the circuit a lot.
This is the latest message from x-Wossee about the LiPo voltage spike issue:
"Transient voltages suppressors are a separate class of devices - chips, TVS diodes, etc.
This driver does not have full protection against poor supply voltage.
Of course, some of the spikes will be absorbed by the input ceramic capacitor, a fast fuse
will also help protect the driver from "bad quality" of the supply voltage,
but parasitic inductive voltage or other anomalies can damage any device, not only this driver.
Your task as a developer is to ensure an acceptable quality of the supply voltage.
How to do it I described before.
Right now I was trying to kill this driver with my 3 cells 3000mAh 25C Lipo,
and I could not do it.
I made about 100 on/off cycles and to no avail.
But maybe you can create conditions under which the driver will be beheaded,
I really do not know.
I'm sorry that I can't give you unconditional guarantees.
So, the LiPo choice is a risky choice, I can be lucky or not.
I throw an idea here:
use a 2S LiPo pack instead 3S with a high discharge value (like 50 or even better 100c).
In that way I will reduce a lot the voltage spike issues while providing all the needed amps.
Also, I do not know if my original design - 4 cells in parallel in a battery holder, with three battery holders in series -
can works correctly.
Even a small resistance in a contact can unbalance the system and cause overload in one or more cells.
The only way I see that being possible is with inductive kick. It would be reproducible with a bench power supply, since the internal resistance there is effectively zero. All lengths of wire are inductors, although small.
There are lots of ways to deal with switching noise. Common mode chokes, TVS, lead twisting, LC filters, extra capacitance, RC filters, or a simple zener to name a few.
The LiPO is not to blame. It is the lead length, and the input filtering of the driver that is potentially at fault.
The problem is that something like a LM338 (or LM317 etc) is designed as a voltage regulator, not a constant current source.
They can be used as such though, and that works fine in most applications.
One issue with them is how they regulate a constant voltage difference between output and reference pins: This circuit is basically just an opamp driving a power transistor, and that opamp as a limited bandwidth and slew rate.
So in a steady state this is all well, but in quickly changing state you can run into problems: A sudden change in load or input voltage can result in a brief transient of overcurrent. In most applications this is no problem at all, as these overshoots are usually in the microsecond to millisecond range, and most loads will tolerate that.
Laser diodes, however, do not: in a millisecond you can have catastrophic optical failure.
The only way to verify that a driver doesn't produce such spikes is looking at the output voltage on a scope whilst connected to a dummy load. A digital sampling scope would be most practical here, though you can see it using an analog scope as well.
I have used the LM317 to drive laser diodes with no problem as current regulators, the reason I don't use a linear for nubm44 is the efficiency, however with 9 diodes in series if your batteries are up to it the efficiency is good, also the cost is good, but everyone is responsible for protecting their own diodes, I take no responsibility.
Oh yes on my cheap mastech supply I was killing drivers by connecting them with the supply set above 10V which I only started having issues with when the newer higher power diodes came along as I use 12V with the 3.5A and over as I am lazy and don't want to change the jumper all the time on my test load or have to remember which setting I have it on when I put a new driver on it as most the time I am setting drivers I try to do things efficiently so I will take a block of orders and see the number of drivers of different types and what currents to do them all like an assembly line having them ready for the builds or if I have a large order where I am setting many many drivers I want to just book through them only pausing when my massive test load needs some cooling.
the setting I leave it on will be over 2V with 100mA so good with any of the boost divers through their full range of input voltage which is right now between 2V-6V(I think the new boosts BBe has will be able to have higher input and higher output capabilities if i remember correctly) and the didoes i use them with only up to 1200mA for use with the high power reds I offer.
For the 1.5A-2.5A it drops 4V-5.5V so good testing bucks with 9V on the supply between those ranges and where I had not had issue in the past with loading form the supply.
For 4A and over that setting will start dropping around 8V and so I have to use a higher input voltage if I do not want to have to change that pin selection going from the first two groups to the higher group and keep a mind on where it is as if you are setting for 4A or 3.5A and have it on this settin using 9V you might actually get a reading while turning the pot but fall out of regulation right near your target output current but keep turning and keep in mind going out of regulation is then basically going back to voltage regulation with the power source minus the dropout in the driver being the fixed voltage and if you are near 3.5V or 4V you might get some increase in current with heat rise when out of regulation that you think is from your turning of the pot so you are thinking you have one current and then when a lower load is placed on there and it goes back into constant current it set higher maybet to the detriment of the diode.
anyway long explanation why I started using 12V to test higher than 3.5A setting of a buck so I can hot swap quickly off the pinned breadboard that the drivers fit in on my test load that Lazeerer built. Pretty efficient setup actually but started killing SXD's and blackbuck drivers when hot loading with 12V which is way faster than turning voltage to zero removing the driver, placing the new driver and tuning backup to 12V all the while trying to remember to rinse an repeat without having those fubar moment when flying though setting drivers and kill one by not following the sequence. That is what led me to landing on and purchasing the BK unit I have now as it was supposed to have a pretty small overshoot when loading which every power source will have.
And the real long story short I can save you time and drivers to confirm a cheap power supply will blow driver's hot swapping or connecting when set to some voltages with the current maxed not being regulated other than by its max potential.
I can see the utility in having a jig to set drivers on, but I would never run them through on a hot supply as that is the best way I know to kill one. I know it is a PITA, but setting your supply every time when I am setting up a driver has saved me countless times from really stupid errors because you can't be aware of everything going on all the time. It also works quite well when driving LDs to check the wavelengths when you have twenty or more to do.
For 4.5a the resistor would be 0.2667 the cap is not needed AFAIK
The input capacitor is a good thing. Mostly it will work without, especially when powered from a battery with short leads and such. But it doesn't do anything bad here either, and can prevent problems with oscillations - just leave it in as long as you have physical space for it.
Also make sure you get the proper resistors. As drawn, the single 0.2 one drops 0.9 volts at 4.5 amp, dissipating 1.8 watts. You'd need a 2 watt resistor, but make sure you do not use wire wound resistors, their inductive nature causes problems. So make sure it's a film type.
There are more ways to get to about 2.66 ohms though. As the whole combination dissipates a whopping 1.25 volts x 4.5 amps or 5.6 watts you have some choices to make. One would be to use 0.4 watt metal film resistors. 14 3.9 ohm resistors in paralel would give you the value needed, and be within their power rating as well.