Referee 1's review (Tim Holy):

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication, the manuscript has been revised in light of these comments and to address other editorial requirements.
*********

Review of the original submission:
The submitted manuscript by Albeanu and colleagues, "LED arrays as Cost Effective and Efficient Light Sources for Widefield Microscopy," describes the technical details of assembling an LED light source for fluorescence microscopy. While one can find brief descriptions of LED illumination sources on the web or in publications from other labs, and LEDs are already used in some commercial products, it could indeed be useful to have a permanent (journal) archive containing a more comprehensive description of LED light sources for people who want to "go it on their own." The systems built by the authors seem to be very well-designed and quite flexible. Therefore, this manuscript appears to be a good candidate for publication in PLoSOne.

The main issues with the manuscript are simply in making sure that the techniques are sufficiently well-described and documented to make this manuscript useful to others. Since this manuscript is most likely to be used by more technically-capable labs, overall the level of provided detail is quite adequate.

As a general point, there are a few places where the treatment of optics in particular needs a bit more detail and sophistication; fortunately, those should be easy to fix. There are only two issues of any real import:

1. The word "Kohler" seemingly does not appear anywhere in the manuscript, and this is an omission for a manuscript about illumination in widefield microscopy. This rears its head most strongly on p. 11: "the tube length and the lens should be chosen specifically so that the light beam fills the back aperture of the objective...if there is a concern that inhomogeneity inside the LED array can affect uniformity of lighting, a diffuser...can be used." Filling the back aperture, while a well-known criterion for laser-scanning microscopy, also has the less-often-stated implicit component: the beam should also be collimated, i.e. with the center of the waist positioned at "infinity." This arrangement causes the beam to be focused to a point in the sample plane (I'm assuming infinity-focused objectives, as modern objectives are); since a collimated laser beam can be thought of as also originating from a single point source (with an infinity-focused lens system used to produce the collimated beam), this corresponds to forming an image of the (effective) source in the focal plane of the sample.

But when performing widefield imaging, you definitely do not want to form an image of the illumination source in the focal plane of the sample. Instead, the right thing to do is Kohler illumination (or a close approximation). For epifluorescence, this corresponds to forming an image of the light source in the back focal plane of the objective, so that each point in the source gets mapped to all the points in the sample plane. This is the means for making sure that the light source does not produce inhomogenous illumination. One nice feature of LEDs---one that the authors don't mention---is that their emission is "thin" along the optic axis (i.e., effectively planar), so that you don't have some parts of the distributed source being more or less in focus.

One can alternatively think about it in terms closer to the author's description. For a given configuration, there are in fact two ways to arrange the focus so that the illumination approximately fills the back aperture: if the focused image of the source is closer to the LED than the objective (i.e., rays from a single point are diverging by the time they strike the objective), all might be well; but if the focused image would instead be produced inside the objective (i.e., rays from a single point are still converging when they strike the objective), then the illumination will probably be very inhomogeneous. The author's description does not point out the importance of these considerations, and so some people who might wish to build their own systems might be puzzled about why they follow the authors' prescription yet get very inhomogenous illumination.

The bottom line is that the focusing of the illumination need to be described in more detail, and perhaps with a bit more sensitivity to the centuries-old understanding of the problem of illumination in widefield microscopy. Furthermore, since the authors bring up the idea of the diffuser, they should also explicitly say which (if any) images in the paper were taken with a diffuser.

2. It's long been possible to make an LED light source that generates a lot of lumens: just hook up a whole bunch of LEDs. However, if I remember correctly, it's a theorem (see, e.g., Born & Wolf or books by Warren Smith) that what matters most for microscopy is the source flux, i.e., lumens per unit area of the emitter. That's what the modern generation of LEDs delivers in a way that their predecessors do not. (It's also why arc lamps have been popular despite their problems.) So the table on total lumens output, while easy to pull from manufacturer's data, somewhat misses the point when it comes to microscopy applications. It would be more useful if the authors could quantify brightness per unit source area, or at least acknowledge that it's flux rather than total output that fundamentally matters.

Smaller points:
p. 4: While there is discussion of edge-emitters and surface-emitters, there is no mention of what class the LEDs used in the study fall; in general, this is a distinction which isn't used later in the manuscript. (Note this intersects with the "light density" issue described above, since edge emitters have much higher flux for a given total output.)

p. 5: "Directionality of light" is a reasonable description, but "tightly focused" is a separate issue (a high NA lens can form a "tightly focused" image despite the fact that the angles of some of the rays are large). It's also not clear what "good optics" are---the quality of lenses in illumination systems does not have to be high, but the optical layout does have to be well thought-out.

p.6: "No unwanted heat" is not entirely accurate; Lumileds LEDs, after all, require one to bond the LED to a heat sink, and failing to do so shortens the lifetime of the LED and gives variable output due to temperature changes. It certainly is fair to say, however, that the heat output is far smaller in proportion to the light output when compared with halogen, mercury, or xenon sources.

p.6: "Do not require a warm-up period" At least in our experience, this can be a bit misleading because the light output of an LED is somewhat affected by temperature. Very stable illumination is indeed possible, but one gets the best results only when temperature is constant, or at least consistent from frame to frame. If the LED is being switched on and off to effectively shutter the image, it may be useful to take images on a regular schedule and ignore the first few frames so that the temperature cycles have the chance to move into a periodic pattern. Or perhaps the authors' heat-sinking is simply better than ours, and this just isn't an issue for them. At any rate, some mention of the temperature-dependence of the output could be useful to readers.

p. 10: "high powered LEDs can afford brighter illumination of the specimen than Xenon lamps": reveal here that the basis for the claim will be made explicit later (in Fig. 4 and corresponding text)---otherwise it's ambiguous whether this is based on theory, experiment, or made up out of thin air.

p. 11: The authors may want to acknoweledge that using a potentiometer rather than PWM does decrease the lifetime of the LED, but presumably the simplicity of the device makes this tradeoff worthwhile. This is a vey minor point, and the authors are free to follow or ignore this particular bit of advice as they see fit.

p. 12: were the other details of the LED coupling the same as in Fig. 2A? If so, how were the "same optics" used for coupling the Xenon lamp and the LED array?

Fig. 1B: I'm puzzled by the circuit diagram: there is a dangling 820Ohm resistor (the trace is missing, presumably), and the symbol for the potentiometer is unfamiliar (in addition to the arrow indicating adjustability, there seems to be a second arrow almost as if it functions as a diode?). I just may be unaware of this type of component or symbol, but perhaps some explanation in the figure caption might be useful for some readers.

Fig 1C: in the left panel, what is actually being measured, light output or voltage supplied across the LED? In what way does the left panel show something about how the light follows the supply? (Are there two traces here, right on top of one another? If so, make the underneath one thicker.) The panel on the right is much clearer. It might be useful to label the right part of this figure as panel D, since it's a separate paradigm.

Fig 4B: which trace is which? If we use the color scheme in Fig 4C, it looks like the LED trace is of lower intensity than the Xenon in 4B, but 4C suggests the opposite is true in general. (Although in 4B the other trace appears blue rather than black, so perhaps there is no way to determine the correspondence.)

Fig 5C: what wavelength LEDs were used for the two imaging conditions? I did not see a long-wavelength LED in the "parts list."

Review of the first revised manuscript:
The revised manuscript represents a thorough reworking of the original text, and authors have addressed all concerns in a very satisfactory manner. In particular, the manuscript is detailed and authoritative on the conceptual and technological issues relating to illumination, and will serve as a valuable reference for anyone wishing to use LEDs in microscopy applications.