the driver specs are: output voltage 30-36 [V], 3000 [mA] ... The 10 [W] chips run at 9-11 [V], 900 [mA] ... so is this possible
No.
Not the way I understand the scenario.
You'd have to run three 10W LEDs in series to match the voltage but the constant-current driver then has the wrong current.
You really need to buy or build drivers to specifically match the LEDs or LED-modules you are driving.
You are going to want to make a current driven set-up.
As high power LEDs get hot and/or age they will change their characteristics. You are going to want to account for that and the easiest way is driving a 4.2A (or, advisably 4A with brand Unknown eBay type) and let the voltage be what it is.
The cheap Chinese drivers usually drive a current about 10% below the normal advised LED current for that type and dimension. The best driver I have seen (German built) was 97% efficient given a narrow input voltage band and exactly the specified LED. My own measurements of cheaper models come between 70% worst case and 91% best case. This differs from type to type and greatly with input voltage.
Many cheap modules also have an input pin somewhere on the main chip that allows 0-2V or PWM (3.3V or 5V) dimming, if you know what to look for. In case you want to add that later.
I have not had the need to import and/or use 100W 24V COB drivers yet, so no specifics on that exact type.
If you use an Atmel to drive it with PWM, again make sure you use a current measurement for your feedback parameter.
As for driving them with AC, there are undoubtedly modules that do exactly that, but I don't know any specifics.
If you build something yourself for 4 lamps, that are well protected from touching and moisture (this includes the LED's metal base plates and such) you could potentially regulate the current with an Arduino through 4 lamps in series with a voltage taken directly from a rectified AC.
This, however is risky and requires a lot of careful experimentation and testing with a low voltage, low current power source to avoid damage (48V, 3A and one LED PWM'ed, for example, no risk of death and if you accidentally short the PWM transistor the supply will limit to 3A, leaving your lamp in tact).
And officially you need to perform power factor correction, but if it's for hobby / one-off there's worse devices in the world that used to be mass produced, so don't worry too much.
If you are working with just the LEDs, keep a VERY close eye on temperatures as well, they need a lot of help getting rid of heat and just the LED's back surface will absolutely not be enough for more than a watt or 2.
Best Answer
I've been playing with some Luxeon COBs wit a 1.2A test current and 2.4A max.
It's all about thermal management. You cannot pump these guys with max current without extraordinary thermal management.
On your part, the most important spec in the datasheet is the thermal resistance from junction to case (thermal pad). On the Luxeon COBs I have the thermal resistance is only 0.29° C/W.
You cannot connect these in parallel. I have tried. Even with matched forward voltages, (42.0V and 42.2V) it did not work.
I also tried 5 ohm resistors, I have thousands of 5.1 ohm 5 watt resistors left over from a telecom product I used to manufacture. I had to connect 4 them in a 20 Watt series parallel network so they didn't get so hot.
Connecting the COBs in parallel what happens is due to their thermal sensitivity there becomes a mismatch in forward voltage and current. It happened over a 12 hour period and the thermal runaway go to the point where one got all the current and the other almost none.
This COB has been to hell and back as shown here. This has gotten so hot the twice solder on the power pad liquefied. That happened at about 1.5 Amp while mounted on a 0.125" thick, 1" wide, foot long copper bar. That bar is screwed down to a 0.125" x 0.5" copper bar that is soldered to a 0.5" diameter copper water pipe.
With no fan and nothing flowing through the copper pipe the max current would be less than 0.5 Amp. At 500mA the temperature of the copper bar rose to 73.4º C in half an hour. Had I allowed it top go, the temperature would have continued to rise.
I always use a constant current source. My favorite is the Meanwell HLG series which range in price from $30 to $200, ranging from 60 W to 540 Watt.
If you build your own the BOM for a driver is under $2 for a buck step down.
For slightly less ($1.40 qty 10) you can use an LM3466 LED string load balancer. I will be trying this myself in the next day or two. This week I bought the TI eval board for this part AN-2182 LM3466 Demonstration Board Reference Design.
With COBs there would only be one COB per string. The COB is essentially a string of 12 LEDs (3v x 12).
I doubt you will be able to push them over 1 Amp without them burning up or liquid cool. So a HLG-540-42B would be sufficient or 2 x HLG-240. I recommend the B model because it has a dimmer as well as voltage and current adjustment.
Or if you want to go green. You can try this: