The Texas Buck driver series, Q8 / Skyray King 2S/4S buck driver RELEASED!

467 posts / 0 new
Last post
Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Texas_Ace wrote:
Well we can play with the layout but the final version my goal is to have it as universal as possible. Meaning the fewest possible parts changes between setups to make it simpler for new guys to build and use these successfully. AKA, we need to keep the parts list simple as we can, so even if we give up a little efficiency, if it covers a wide range of uses, it is worth it. The layout I am using now is 100% by the example in the datasheet.

 

 I think that's important for having a baseline to go to smaller sizes.  It maybe won't be a big deal for an extra model at this size.  We're talking about 1 extra small fet but three less unusual resistors, so from an assymbly perspective, it's about the same, or actually fewer parts (but as one of them is more specialized, potentially harder to get).  The basic design is a little better for downsizing.  The current sensing nfet won't shrink (there aren't many models made), although you could replace it with a regular nfet and small resistor but still stuck with the small pfet to drive it, so for small models may as well have this (I mean your present) circuit.  

 

It's fine either way.  And if you aren't tempted by it, it gives me a project to pretend I may one day tackle to learn diptrace, and who knows, maybe I even will. I was kind of thinking about it before you restarted this, so I might still.

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas

Oh, I never said we can’t do it, I just wanted to keep this in mind.

The whole idea behind my driver projects is standardization. When I first started researching drivers on here it took me WAY too much effort to figure out what the difference between drivers was, what parts you needed ect. That is what led to me making any of these drivers and I don’t want to stray from that.

So Like the Texas Avenger drivers I want the Texas buck series to be the same basic setup for all the sizes, just with different components for the current/voltage but once again those are standard for all the sizes.

If these new components can scale to the smaller drivers just as easily as the stock setup, then I am all for them. I honestly just don’t have the time to research the details on the components right now.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

The existing pfet version should scale down a little better. So it's better for that.  I think of the nfet version as a Q8 special, non-series version. I edited my reply above about the driver too.  It's still possible maybe to consider a minimal layout change with an nfet.  I'd momentarily forgotten that indeed on of those drivers allows an nfet to be placed high-side.  I think it was huge or had some other issue, but I didn't search hard for other options for it yet either.  This wouldn't be kept in a scale down either though.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Unfortunatley though, probably many of the big components will need to change in any scaled down version. The inductor takes too much real estate.  And the diode could probably be significantly smaller. The sense resistors, well maybe can just use fewer, we'll see, but probably will want to tune the values better than just that.  The pfet might be able to stay the same.  It won't quite be like you just buy 5 of everything and make different drivers though.

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas

Yeah, those will have to change but that is also due to not needing as much current, so it will still follow a simple chart where people can figure out what they need by voltage and current.

It is mostly simplicity of design, aka the design is the same and uses the same type of components, only the values (and footprints in some cases) change.

It is not iron clad either way, it is simply the goal I am going for.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Sure.. but I thought buck flipping is like the national sport of Texas Smile  It's more dangerous than cow tipping for sure, but more fun?

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

... So I found the problem with the sensefet.  Oh well.  It has a current output of sorts for sensing, but it can't support that current with more than the drain to source voltage drop, typically half of that or so.

It's nothing more than a parallel current path effectively though a parallel fet.  No amps or anything built in.Aparently this is more stable than just measuring the Rdson voltage drop directly, but it's still a tiny drop, and needs an amp.

 

At that point, it seems might as well just use tiny sense resistors and an amp.  Sure that adds more loss than this, but still maybe 10 times less loss than what we have and much simpler to implement.  Which is to say, maybe still not so exciting, I don't know it's the next thing I'll look at.  There are aparently off the shelf IC's designed for that too, "current sense amplifiers" so next I'll find out what's wrong with those.  That at least is a directly removable solution when downsizing, just increase the resistor size and remove the amp.

 

 All this though is just for ideas to make what might already be the best driver here (depending on use anyway), even better.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

I did find a very nice and very expensive current sense amplifier designed for almost exactly this purpose.

https://datasheets.maximintegrated.com/en/ds/MAX9643.pdf

But it's not quiiiiite drop in.  After staring at the lm3409 sense circuit I confirmed that yes, it needs a Vsns- to be at least 1.25V above ground.  That can probably be done by referencing the amp ground to the MCU Vcc for instance, but let's come back to it after finishing up the baseline design.

 

So, thinking about iadj a little.   A couple of things.  Right now, the iadj voltage is 4.2 V times the iadj PWM duty cycle.   Except I think that 4.2V is actual battery voltage?  You are using a single battery Vcc design here right?  This is unfortunate.  It's going to make output fall off as voltage sags, just as it would in a direct driver!... because the battery is directly driving the iadj pin.  That's pretty obnoxious for a buck driver.  So you might want to bring back out the ldo.

 

Edit: I got confused on PWM RC optimization.  I'll come back to it. 

 

Still.. an RC of 1 would have a nice touch that all mode switches will be soft.

 

Another point though, right now full scale is at a duty factor of about 25%.   If we use a voltage divider we could make 100% be full scale. But is 50% still 50%?  I have to scratch my head on that.  

 

Finally, did you actually hook up uvlo?   Do we want voltage protection there or in the attiny?  Having it redundant seems unnecessary and could cause confusion.  Plugging Vin straight to uvlo will just override it, and you get rid of two resistors too.

 

 

 

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas

I had a few mins to update the design some. Getting closer to a final layout, thick traces all around and everything is moved in to allow for 3mm edge clearance (4mm would be better but just not going to happen without cramming things too close to get thick traces.

The pads for LVP are there but do not have to be used. Still got some work on the bottom side to figure out along with some other items of intrest but getting it worked out as I have time.

If you can come up with a better option for basically any of it that is scalable down to the smaller drivers using the same design and offers notable improvements without a ton of extra cost ect I am fully open to it.

I dont have much time to digest the rest of what you said right now, maybe later this weekend.

Here is the latest layout:

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Nice, I definitely want to give some of the details a very thorough once over still, not about major layout changes in this round, but just focussing on the details like the iadj voltage, input caps, etc.  But I don't want to trickle bits out now.  I'm pretty well along on a rework of some numbers, but I might need a couple or even three weeks, because I have almost no spare time for awhile now. 

 

One hint.. I'd only looked at input cap current (and had some formula copy error, did say I needed to review those) but I hadn't looked at input cap voltage at all yet. Some news there.  I'll also give a perspective on the issue of where to charge coff.  

 

Anyway, do reread my last post.  A few little tweaks there that I think need considering even in a by-the-book design, and this iadj usage isn't quite by the book. 

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

What does the jumper do?  Does that bypass the uvlo resistor?  Rather it looks like it just disconnects it. 

 

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas

Opps, I noticed I mistakenly deleted a trace in the schematic and it led to it being erased in the PCB. The jumper traces you noted. I replaced them so it should be complete now.

In answer to your question, the jumper is simply to allow the ground pane to access the bottom of the buck IC for both heat sinking and EMF interference reasons. This is what they recommend in the Datasheet and it makes sense.

You will also note that I have ground/copper pours between as many traces as possible to keep EM interference to a minimum hopefully.

Also what kind of size / spacing should we aim for with the inductor throw holes? I suppose I can mount them offset to where the springs would be, not perfect but should work in a pinch.

I added an LDO as well, kept meaning to do this but never got around to it, things are getting crowded in a hurry!

The bottom side has now been roughed out, obviously I have not solder masked the area but you get the idea.

I will be placing a fairly large order with Arrow in the next week with a discount code. If we can get the components sorted by then it is possible I could order a set of them and if Leroy orders a set of boards maybe I can buy one of them from him and build one of these for testing.

It all depends on budget though, got a lot of money tied up in various projects right now and need to free up some more for a Group but for 219C 90+ CRI LED’s I will be handling before long.

I need people to buy my stuff! but that means I need to get an official listing up for the skyray kings lol.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Well, no problem, we're still polishing things off.  I have/had a couple of bugs too.  

 

Updated:

I hadn't thought much about the diode in the iadj input.  I first wrote this and I thought I'd overlooked something there but not really.

If the cap never gets over 1.24V the diode is never an issue in the circuit and can be ignored.  As it is now, that's the case.  As it is now we saturate way below 100% PWM too (not necessarily ideal, but just lose mode resolution). And as it is now, the LDO is needed.  What thought does come is that maybe there is a way to leverage the 1.24V reference where the LDO isn't needed. I'm not seeing it though without other parts anyway. 

 

Of course one alternative to the LDO could be a zener diode on the mcu iadj output, but the LDO works fine.

  

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Ok, I see what you mean about the jumper now.  Well for now that will work.  One can always choose Ruv2/Ruv1 small so the low voltage protection never kicks in.  It just seems like if you're doing low votlage protection in the mcu, not point having it both places, and could remove two resistors.

 

The jump seems like it should be avoidable, but then I realized it probably helps heat sink the resistors anyway.

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas

That is true, we could just use the MCU for the LVP, although since the pads are in place that is easy enough to simply not populate one of them later as well.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Texas_Ace wrote:
That is true, we could just use the MCU for the LVP, although since the pads are in place that is easy enough to simply not populate one of them later as well.

 

Yeah, I guess leaving off Ruv1 does the trick.  Still need an Ruv2, but not a big deal. It provides an option I suppose if the mcu thing doesn't work well for some reason.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

On the other hand, it only needs 1.24V as I reall to be high.  If there's a way to route Vcc Mcu to it from the other side without cutting off other grounds, that could eliminate the jumper as well as the two resistors. Or it least turn the jumper into just an optional heat sink lol, just thinking about possibilities to remove parts, but cheap parts in this case.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

You're worried about size of ground trace on the left of U2 (there is a thin trace through Ruv1 and Ruv2 even without the jumper) but on the upper right of U2 you've pinched the ground without much reason.  Hmm, maybe I'm getting it, you expect that side to actually deliver heat instead of removing it?

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

As far as through holes.  On digikey bourns is the main brand that pops for less than 10mOhm greater than 22uH > 15A.    They have some in about 30mm spacing (yikes) and some in about 12mm spacing.  For the most part 12mm is a vertical inductor mount.   1.7mm pin diameter.  I didn't realize how big the pin spacing is on the flat ones.  Probably would need one pin southeast of LED+ and one between Q1 and D1, with whatever interference that causes on the other side.

 

Here's one of the flat kind:

http://www.digikey.com/short/3bcd55

 

However there is also this:

http://www.digikey.com/short/3bcd55

http://media.digikey.com/pdf/Data%20Sheets/Wurth%20Electronics%20PDFs/S14100034.pdf

 

2mm pin diameter, 11mm spacing (close enough to 12, I'm sure they'll bend) and it's a horizontal off-center mount.   That actually looks pretty good for us.  But you have to look at the picture to understand hole placement. I'm not sure about that 5db attenuation at 1Mhz.  Seems to be related to the self resonant peak.  I can't pretend I've understood that issue well, which is why I've preferred inductors that just don't get funny until a few MHz, but these big ones just don't have the detail spec sheets.

 

Not clear if the bourns inductors are better.  They just don't show as much detail.

 

One could I suppose solder some pins to the sides of one of these low resistance 22mm's we were looking at:

http://www.digikey.com/short/3bcdfz

but that's not very user friendly.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

IADJ is licked

 

I started this PWM the iadj pin thing, so I'm finishing it.  I worked out by hand and time slice integration/simulation, the situation with iadj, with the two agreeing.  Bacgkround, the iadj pin can only take up to 1.24 V input  to control current output of the light.  Any more gets cut to 1.24 anyway by an internal diode.

 

Now, imagine a resistor R1 on the left in series with a capacitor C1 on the right, and they meet at Vout in the middle.  A resistor R2 is in parallel with the cap, both connecting to ground on the right.  R1 connects to a PWM voltage V1 on the left.

 

This happens to be just like your circuit except R2 is infinity, no problem.  (my R1 is your Rf2)

 

The result is the maybe not surprising (I'd call it less obvious than it appears, when you start proving):    Vout= V1* D * [R2/(R1+R2)]

If R2 is infinity this is just      Vout=V1*D * 1

D is duty cycle.

 

Great so far.  So we can make 100% max range 1.24 V if Vin is 5V by making R2=0.33*R1 in the usual voltage divider way (requires adding pads for R2, and I think it's best to add it, can always not use it)

 

Or we can just leave it alone and just use a duty factor of 25% for max output.

 

But what about ripple voltage?  Ripple is obviously zero at duty cyle of 1.  So if we set the max to 1.24, we get zero ripple at max power.  That's kind of nice.

 

On the other hand, ripple (p2p) as a fraction is,  (don't read the equation, just skip to the words past it)

dVout/Vout =  (1-D)*(R1+R2)/(R1*R2) *1/(fC)

 

This is a max at (near) 0 voltage output, and if R2 is infinite max fractional ripple is just:

1/R1 * 1/(fC)

 

The ratio of max fractional ripple as R2 is decreased from infinity is just the inverse of the ratio of the max output voltage.

 

So for 1.24V output range, the max ripple (which is at minimum output) is 4 times higher than if setup for a 5V output range! (that's too bad)

 

So for 1.24V output range, we get better control ripple (0) at the max light ouput, but 4 times worse ripple at lowest output (near zero), compared to using the full 5V output range.  I'm not exactly sure where they cross.  Easy enough to work out, just didn't do it. 

 

This is a little significant because for 1.24V output range, and  R1 10kohm and cap of 1uF, the max ripple is 2% at 20khz PWM, not soooo tiny, and it's 1% at 50% outpu  and .5% at 75% output etc.

 

Of course we care most about ripple at max output so I still like setting max to 1.24 by adding another 3.3kOhm resistor.

 

 

I like it anyway, because it just seems more sensible and probably more matched with existing direct drive software to have 100% PWM actually be the max output, not 25% PWM being the max output.

 

So let's add another resistor.

 

 

 

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas
Flintrock wrote:

You’re worried about size of ground trace on the left of U2 (there is a thin trace through Ruv1 and Ruv2 even without the jumper) but on the upper right of U2 you’ve pinched the ground without much reason.  Hmm, maybe I’m getting it, you expect that side to actually deliver heat instead of removing it?

Correct, I wanted it electrically connected there to help remove any EM interference but I except the diode, FET and inductor to be significantly hotter then the IC thus no reason to feed the heat that way. Plus I want the heat to radiate out towards the flashlight body, not in to the center.

So basically add another voltage divider, thats not too bad. Got that done.

I tried adding in some through holes but they are just not going to work without risking a short with the springs on the bottom. The pads on the SMD inductor are the correct distance apart, you could always just solder it directly to the pads and epoxy it into place if you really needed to but I highly doubt anyone building this will try to cheap out on that when the rest of the driver is still so expensive. Anything besides an SMD inductor will not fit in the Q8 anyways.

Latest version with added Rf3:

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Great.  The only things I have left for this model are reviewing the input caps again and Coff charging.  My input cap loss numbers were wrong  (just the percentage form, but that's what I was looking at) but voltage swing is now a bit of an issue, especially at 4:1 and a little at 2:1.  With more wiggle room on ESR now, I'll want to review again if I can find bigger caps that are suitable. Ti has a document about this and they use 82uF for a 10A supply.

 

For Coff, it's a matter that setting it up for a good frequency at one voltage will probably make considerably unoptimal frequencies at other output configurations, but at worst it means you need to select new RC for different applications.  

 

Both issues will work the way they are, but may be significantly improvable too. But I won't wrap up my notes on them for a few days at least. 

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas

Thats fine, I will start trying to find the large components from Arrow and we can work out the values for the caps and resistors later, those are easy.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Well yes and no.. values above 1 uf at moderate voltage have size constraints.  I haven't looked beyond ceramic.  The coff thing would also be a rewire, I don't expect you to be convinced on that until I convince you, and it can be a beta revision (this being alpha) if it ever happens.  Again this all will work as is, and can be tested as is if it's go time.  

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas

Remind me what the coff thing was, just too busy lately to put my head into this like I should. You have been a big help, there is no way I would have had time this year to research parts selection enough to really get this project going. Still not sure how you find some of these things.

Heck I was just looking for a simple compact op-amp for a possible linear driver and was coming up with almost nothing.

It is not a major rush or anything, I am just making an order from arrow anyways and with the discount code I have it should make it reasonable to order a set of parts to give this a try. So I figured I would grab it while I can. Not worth making an order just for these parts if I have to pay shipping.

Same reason I was looking for an op-amp.

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

Well 95% of it has been useless lol.  It's like the monkey with a keyboard thing, but I like to be satisfied that the options have been explored.

  

Coff is the offtime clock (buck cycle off-time, not the traditional OTC).  Frequency depends on that and duty cycle (on time) which is Vout/Vin so it depends on Vo no matter what you do.  But they also charge it off Vo which modifies that Vo dependence.  I'm not sure that modification is useful for our purpose.  Charging it off Vin or maybe now even off the more stable ldo,  might give a more all-purpose setup for us.  But I'll have to generate all the numbers, mostly done, but still, no time is no time.

 

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

It hits me that if we really want to nail the Iadj thing, we have have to consider the output resistance of the mcu (both high and low if different).  I think it's specced at 5ma output (this might be entirely wrong/bad memory), but I don't know if it's limited to that by 1kohm ish internal resistance  or if it just means we shouldn't draw more than that off of it.  Using 10k ish output, of course that's a 10% correction on the voltage divider.  That's not a design change though, just details of the setup to work out/refine either in tweaking iadj resistors and or rsense.  Probably just things that will get fiddled with in field testing.  

I guess this could be an argument for not using the second resistor (but we still don't have to use it).

 

Edit looks like output and input impedance are about 20 to 40ohms depening on Vcc, so no problem.

Texas_Ace
Texas_Ace's picture
Offline
Last seen: 6 hours 49 min ago
Joined: 03/24/2016 - 07:44
Posts: 8492
Location: Everything is brighter in Texas
Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

First a minor thing: I have actually polished off some thoughts on Iadj, but no time to write it up.  Basically the diode bias current is going to cause a voltage offset at high resistances, preventing to reach zero, but you get ripple using low resistances.  None of it is terrible, but may we may need to live with some compromise between the two.  It's mostly unrelated to the two resistor issue. 

 

Now the bad news.  For the first time I looked closer at the mosfet.  That mosfet has 625nC of gate charge at 10V.  I hadn't really noticed that before.  That seemed a bit giant to me, and well, it kind of is.   That's 63nf of gate capacitance! 

 

Both the potential and charging energy is lost every cycle so just P=CV^2*f  (no 1/2).  Switching voltage is 6V, fixed in the IC (Vin-Vcc).

so 63E-9*6^2*1E6 = 2.268 W!  (I hope I made a mistake here, but I don't think so)

 

That doesn't sound sooo aweful, but that's happening all the time.  So at 2W power output, you've got 2.3W of switching loss. No fancy coff correction helps this.  Coff, and frequency only are impacted by voltages (and are under our control anyway).  This may finally be the reason PWM from full power doesn't work out so bad after all!

 

There are other mosfet losses that I've barely thought about. It seems very complicated actually and I likely won't ever find time or energy to deal with it all, but anyway this gives us a measure of at least part of it, and some relevant spec to improve.

 

I'll present the case about Coff when I get time to present it, but it's a triviality compared to this I think. Although the two will be very linked.  This may drive the need to care more about how set the frequency. 

 

Time to look carefully at what fets are available again.  I read a doc, maybe Ti, exactly warning against overspecced large, low rdson fets, for this reason.  We might need to look for something a little less beefy.  However, I still probably don't much free time for a bit.  

 

 

 

Flintrock
Offline
Last seen: 1 year 8 months ago
Joined: 09/10/2016 - 20:29
Posts: 1544

For comparison quickly the SIR404dp nfet has only 97 nC of gate charge.  nfets are said to be better.  Starting to see why.  I might be giving that high side nfet driver another look, but not before looking for more suitable pfets.

Pages