TEC and current driver chips for laser?

Oh, I am sure everyone in the trade understood. I like the LM324 a lot. Low quiescent current and the input CM range includes the negative rail. Very practical for battery powered gear.

Joerg wrote in news:pgH7h.4834$ snipped-for-privacy@newssvr14.news.prodigy.com:

Nicer than CA3140? If so, I must do some rethinking.

Joerg wrote in news:AeH7h.4833$ snipped-for-privacy@newssvr14.news.prodigy.com:

Regulating the regulator? :) I can see that might make the LM317 begin to look like an odd solution...

True. I guess if we can now get IC's with all this in one package it makes sense to get what we can. I think when I decided against the Wavelength Electronics module there weren't many dedicated IC for this, at least, I didn't know of any.

. .

True again, but I like the ease of use, the fact that for many tasks you can start with the same device whose character is predictable and reliable. Above all, unless you're in major production, the costs of laying tracks for discrete parts and soldering them all is far greater in both money and time, than using the LM317. I think of it like the electrons, protons, neutrons.. while we can reduce to quarks and get great power and understanding, things like the LM317 and the LF412 exist at that magical point where costs increase if you go either up or down scale, once context is taken into account.

Don't know about nice but the LM324 is certainly cheaper. By a factor of around three IIRC.

The modulator I built with it kind of did that. I was feeding the Vsense pin and that could probably be construed as "regulating" it. With most LD applications there is a need to keep the light output constant. It's also a much safer method not to exceed the LD limits when you start it up at grossly differing body temperatures. As long as you ramp up from a lower current.

IIRC the ADN2830 showed up arond 2003, others were a bit earlier. The downside is that these chips are expensive.

Plus with the LM317 you get a device that can stomach a few watts if heat-sinked, no need for an extra FET.

I think the LFCSP packages with the exposed paddle should be soldered to a square of metal which is connected to the ground plane by lots of vias. It is important that the paddle is really soldered properly. When soldering them by hand, I put some flux on the paddle of the IC and then tack down some pins of the device before soldering the ground connection from the back side of the PCB. I feed solder down one via and make sure it comes up the other vias, that way I can be pretty sure it has wetted the IC package. Other people do it in different ways and the assembly contractors seem to have no trouble with them, I think they use solder paste. The thermal resistance in the package would be pretty low because the paddle is just a square of solid copper with the die attached to one side and the PCB soldered to the other side. (or at least that applies to the ones that I have seen, though those were not really high-power devices.)

Chris

Joerg wrote in news:hkL7h.16717$ snipped-for-privacy@newssvr27.news.prodigy.net:

One liberating thing about light show devices is the output is aimed to get the strongest beam possible at any set of conditions. Thermistors in the current sensing network help here, by dramatically reducing the effects of large ambient temperature differences. Slow ramping is done with that memcap I put in parallel with the low ESR ceramic and the tantalum I mentioned earlier. It's almost a perfect circuit, as the LD limits the voltage to guarantee the memcap never sees too much. Everything assists everything else like a perfectly tied little Gordian Knot. >:) The downside is, getting ANY kind of modulation in there is a no-no, which is why I can't really advise on a mod for that. I got lazy and distracted to other projects, but I do intend to work on this again.

As an aside, there's another good feature about this LM317 current reg thing, the thermistor allows close-to-limit CW operation despite a large change in case temperature for the whole module. This means you can supply the unit with between 5 and maybe 25 V, at least 12 or 15. That is wasteful at high values, but has one ENORMOUS advantage, heat changes like that can change the focus of an acrylic aspheric lens, providing a perfect no- moving-parts method of very fine, accurate maintenance of focus or collimation, the latter being very important for light shows, so the cost is very low, and the performance, for that purpose, is very high, as good as it can get.

And no doubt subject to change... That's another reason I like the LM317. It's not going to force anything on me that I can't control. I doubt they will ever go out of fashion, wither. :)

Indeed. That focussing thing I mentioned above is a nice exploit of this capability.

Hello Chris,

No doubt this will be fine in production but for prototyping this is a bear. Also, I like stuff to run at lower temps than those LFCSP. I am just about ready to go the usual route by doing it all in discrete/jelly-bean again.

Actually, the NTC's have their own inputs, even their own ADC. The pot has an input of its own at the cpu internal ADC.

Rene

Just in case the LM317 is too close to the lower rail, you can always have a series resistor together with the laserdiode.

Rene

Rene Tschaggelar wrote in news:45603a99 snipped-for-privacy@news.bluewin.ch:

I don't understand this. If it was a voltage regulation, it would arguably be too far, needing a negative on the sense pin to get it down to zero, it couldn't get 'too low' otherwise. All of which means nothing because it's current, not voltage, that must be controlled.

I actually do use a resistor in series in my drivers, I forgot to mention it here, not sure why. I use a 1 or 2 ohm series resistor before the parallel caps across the diode, to make them filter out spikes more easily.

Hello Rene,

That would be a good spot to connect the external prototype circuitry to. A tie into the thermistor node is always a white knuckle ride. If it shorts for some reason the TEC might attempt to create a new ice age or a smelter, depending on which way it goes because the uC would have no way of knowing what happened.

Hello Rene,

Usually it's not too close because the LD drops a good 2V so you could even use a BJT to control the Vsense node. Just in case the five cents for a BSS123 are too much money ;-)

One challenge that is left is to monitor the LD current. That would require a diff amp structure and a wee series resistor. Often MPD level monitoring will be enough though.

IIUC you'd still have almost a volt from Vsense (ADJ) to GND. Enough to hang a transistor to rob some current.

If you had one between cap and LD you could modulate :-)

Joerg wrote in news:ZM%7h.8349$ snipped-for-privacy@newssvr13.news.prodigy.com:

I remember that there's a circuit for current control right to xero that suggests a 1.25 bandgap reference to solve that, but if you're only modulating down to 90% of Imax I think it won't matter. That Adj pin will always be north of the LD load, for want of a proper way to say that..

Nice, I don't know how much filtering I'd lose with that, but it's a nice simple starting point. I was hoping to do it by modulating the sense input though, if it allows a single active part while still offering enough filtering to be safe. Either way, it seems I'd risk something nasty getting through, so maybe separating the supply and mod into two systems is the only way to control things easily. But if I can do it all with one LM317, I will.

That boils down to the question of which pathologies a LM317 without much output filtering could cause. Since it is a rather simple device maybe SPICE plus some lab bench jerking of the device can answer that.

One method that has worked for me in the past is electronic protection (in this case probably some kind of precise crowbar on the LD side) plus a cap that makes sure it can keep things muffled well enough until that protection has definitely kicked in.

Do you know how to measure die temperature? [I can explain the procedure if you don't.] I suspect things are better than you think regarding heat. Parts that are hot to the touch are still quite reasonable designs. That is, the designer takes temperature into account. The real problem would be some problem caused by the heat of this chip as it effects the rest of the system.

Well, I've got a good 20 years under the belt now and parts with temperatures too hot to the touch had higher failure rates than the others. Also, there are applications where there are strict limits on how hot something can get anywhere in the system (medical etc.). Even where there aren't I prefer things to remain in a range where touching doesn't lead to a blister ;-)

I've designed linear power chips They get hot. No big deal IF the design is good. I can't believe AD doing a schlock job. You don't need heat for a part to fail. There are plenty of failure mechanisms that don't involve the chip getting warm to the touch.

The next time you are on an airplane, watch the wing wiggle a bit. You are looking at something that is engineered properly. If the wing was stiff enough not to wiggle, it would be much heavier. It is the same case with your chip. If it ran cool, it would have to be bigger, have a slug in the package, etc.

I would bet AD has burn in data to show their design is good. It is that kind of a company.