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

One thing I like on the 1206 is you can buy the same 10uF cap as for Cin and Cout if you want. You can use 5 maybe 6 of them then and it becomes almost free to go ahead and order 10. The $0.30 I like has a one week lead time but there are others for 60 or 70 cents I think available instantly too.

So, if you want to save a good bit of space, I keep coming back to that jumper and I had an idea. You can connect any steady voltage over 1.24 to pin 1 of the the IC to enable it. Which means any battery plane other than the ground plane is fine. I'm pretty sure the one directly behind it on the board is not a ground. So stick a via through to connect pin 1 to the battery pad on the other side and you remove two resistors and a big jumper, save 3 components to spec, purchase, install, and you can shove the IC waaaay out to the left and gain all kinds of space. Cf can slide out around rsense then, the Rsense can come down, the IC can come down or up a bit, a ton of flexibility opens up.

I like not needing extra resistor values anyway, ones that will need to change for 2S vs 4S battery configurations (or 1S), add more permutations to everything.

Then I might ask C1 to get bigger too... why? Just because we'll have about 5 or 6 10uF 1206 caps already, might as well buy 10 for $2 and use 7. I'm trying to simplify the parts list, but not a big deal either way obviously.

Question, what's the deal with the switch? I thought the Q8 has a switch, so shouldn't it connect somehwere? The switch pads have a cap specced, not a switch. Related then are we using 220K and 47k or 22k and 4.7k?

The pad behind the IC is a ground pad sadly and there is not an easy way to get to a positive pad without a jumper somewhere or running around the ground ring. Plus if I did have it connect to anything besides where it is now the layout would have to change for 2s2p and 4s1p setups.

The switch pads are just a pad, It is easier to set it as a resistor in the schematic so that it doesn’t fight you having it in place plus it makes sure you don’t forget it.

The MCU side of things is just like the rest of the TA drivers, it could even be used with a clicky switch if wanted. So for an e-switch it would be 220k/47k. For a clicky I actually like 19.1k/4.7k a bit better, it makes the OTC a bit more reliable.

Well it would only have to change if where it connected to became a ground. There's at least one pad that's always hot right? Looked possible to get a trace from there. Ok, well I'm probably missing something. Sure seems like a bunch of space just for a voltage.

Just noticing C1 can move left and jump the pwm trace downward. That might allow a bigger C1 (maybe not), but also opens a trace that you could run out through roff under the mcu. Loses some of the space gains.

We're already right at 3.0V for the voltage sense on the mcu for full battery, 3.06 for 4.35V batteries, so going less than 22k/4.7k cuts out voltage monitoring on the top end unless I go up to a 3.3V ldo. The 3V ldo helps the soft mode transition a little, but 3.3 isn't a big difference either.

I don't fully understand the off time clock issues/voodoo.

In 4s mode the pad directly under the buck IC is ground, in 2s mode it is hot. The only pad that is always hot is the pad behind the sense resistors. Although the jumper resistor really doesn’t take up much usable space. I still could not put anything on that side of the IC without cutting off the ground plane running under it, which is the entire reason for that jumper in the first place. It could be moved out but it would not free up as much space as you think.

C1 doesn’t need to be any larger, the normal 0805 10uf amps we use work fine for it. Although having it jump the trace is not a bad idea anyways, the ground path to it is kind thin as is (well it is actually non-existent right now due to a minor movement during the last edit).

I actually forgot about the multi-cell useage. The voltage divider will actually have to be completely recalculated for the higher voltage, I think people use 36k? for 4s? I forgot.

I meant to say bring the C1 trace out under the U1, not mcu. It's a way to get voltage there. C1 doesn't need to be bigger but we'll have 5 to 6 1206 10uf caps already so will be ordering 10 anyway. Can probably just squish one in though. I've already got 0805's so it's fine with me in any case, just thinking to simplify.

Yeah, I think R1 can be at minimum 22K but I don't know. It depends on

1) battery config

2) ldo voltage

3) software and mcu capability

I think I know mcu capability, but I don't know the software side of things as far as calibration.

For different configs we need to change:

Ruv1 (if it's used at all) , R1, Rsense, Roff.

This is what got me back on this track.

Not sure I understand, you are saying bring the trace from the via that feeds power to the MCU around the ground ring of the driver to the buck IC?

That is a not something I like doing unless there are no other options. If you scratch the masking it will short out. Plus it also doesn’t allow the ground pour to contact the body and thus transfer heat effectively.

The only other option is to run another trace on the bottom side of the driver to a via neat the buck IC. Not ideal but possible. Although I am still not sure how much usable space that would free up. That space used by the jumper is not really usable anyways, the only thing it does is allow the IC to be pushed a bit closer to the edge.

For battery side, I don't understand all the issues there with battery configs and you do so that's fine. I thought bringing it off the midline like the mcu via might work, but I don't know. I thought scratching is a risk on that side (because of bashing and spinning batteries).

The last one I was talking about was splitting it off from the hot side of C1 on the component side, and winding it between LED+ and the switch, under the around Roff and under u1 to pin 1, so not around the outside.

Ok, maybe it doesn't save as much as I think. I thought Cf could then push out behind the Rsense, minor, but all together then U1 can move out quite a bit, or out even more and also up a bit, or out a bit less and down a whole bunch (I'm removing both UV resistors too in my idea because we just don't need them, so that's two resistors, a jumper and cap gone from that space and more use of the curvature). Looked to me like in the U1 down-and-out option you have a ton of room to move rsesne down (opening some space or options top right), and probably to move roff and CF2 up then at the same time adding clearance there.

Anyway, just ideas in case they're helpful, just ignore them if they're not. If there's enough room as is, then maybe no need for any of it either.

For now though I am speccing Ruv1/2. So what voltage cuttoff do people like? 2.7V and 0.2V hysteresis ok?

I'm trying to find options to do everything with E24 series resistors, fewer values being bonus points. I'm not sure how important that is. We have to work out R1/R2 though. I have no idea how calibration works, and I'm not sure I know how the OTC works either. I don't know where standard values exist, or if they would be setup for the same mcu voltage as us anyway, which I think matters for battery sensing.

If we are removing the low voltage resistors then that is a bit different and I might be able to figure out a way to do it (mostly because that would make it a worthwhile amount of space savings for the effort).

I think the mis-understood idea from you of using a via from the sense resistors is the best idea. The batteries can not be inserted like a normal SRK with this driver, they MUST be inserted from the rear or they wills hort out.

Still not ideal but a viable option.

Leaving low voltage control to the MCU is best like we said before, that way it can be adjusted to each persons likes later. I like mine a bit higher at around ~3.1v step down and ~2.75V cutoff.

It's a pretty big savings. Worthwhile? Depends what if anything it's needed for and how much effort it is. It does create a new problem, figuring out what best to use it for, and I don't even know the answer. bigger diode? More input caps? just more clearance and copper between things? I should learn diptrace a little someday to at least have appreciation of the effort side.

But how is mcu voltage cuttoff adjust? In software recompile? In software UI? By changing R1 and R2? Where can get up to speed about R1 R2 and R5, low voltage and OTC?

At the very least it preps the driver for small footprints later. I will give it a go at some point.

The LVP in the ATtiny is based around a simple voltage divider. The internal reference is 1.1v IIRC, so it simply needs to have the output of that voltage divider be less then that. For all the details on the ATtiny I would look at the data sheet, good luck, it is a long one.

MTN lists 36k/4.7k for 3s/4s usage IIRC. Easiest way is to simply find what others have used and go with that. The fine tuning is done in the software calibration anyways as long as the voltage on the input pin is less then 1.1V.

OTC is not needed for the e-switch anyways and it would not effect the buck side of the driver. It is simply a 1uf cap that the MCU monitors the voltage of. When the power is turned off the cap slowly drains over the course of a few seconds. The MCU checks the voltage when it turns back on and if it is above X value it is a short click, Below X and it is a med click, below Y value and it is a long click.

Pretty simple really, once again it is calibrated in the software so it doesn’t have to be very precise, just consistent.

Ok, well the basics are the same as what I figured. I missed the 1.1V though. It seems there are three references possible, 1.1, 2.56, vcc, or an external reference, so my information was based on vcc.

So 16.8*4.7/(36+4.7)= 1.9

At 2.5V per cell, Even 10*...= 1.15

So I'm guessing MTN is using the 2.56 internal reference, not the 1.1.

I guess we should target this for particular software (bistro?) and better get to the bottom of where that needs to be set. Fine tuning is fine, but 2.56 and 1.9 and 1.1 are all pretty different. Software calibration means before flashing or some kind of user interface adjustment?

http://www.mtnelectronics.com/index.php?route=product/product&product_id=610

Indeed they claim 19.1K for 1S and 36K for 2S-4s. That comes in at 0.83V at the mcu pin for 4.2V 1S and 1.94V at the pin for 16.8V. So in at least some of those driver variants they must be using either the 2.56V reference, or Vcc, not 1.1.

This stuff smells of weird historical baggage. 19.1? What? So you can have any of these different voltages and someone chose 19.1K? As if the OTC is going to be noticeably different if it's 18, 20, or even 22K instead? Very strange. Must be someone's favorite number.

Anyway, I guess we'll go with 36K and just assume someone will work out the software? It does seem to me that 36 can work, with the right voltage reference selected in the software.

It is historical baggage. But, I don’t think it’s weird. Driver design started around here as tiny mods to existing drivers, and worked its way up to what you see going on today. A lot of us don’t really know what we’re doing. We just know (from “history” :wink: ) what works and we do that. Some people have switched to other values like 22K because those work better in their designs. So, it’s moving, changing, accelerating. But, it takes time when we’re just poking at things with sharp sticks. We’re grateful when somebody like you comes along and gives us a push in the right direction. :smiley:

Sure, I didn't meant to criticize. I've got a bunch of work with weird historical baggage. Most of it never really had a good reason, but it had a reasonable way of coming to be. I suspect 22 became popular because it's a standard value, and yeah, maybe someone liked a little slower OTC.

Well with 36K presumably settled, I've got everything specced. I just need to add all the resistors to a shopping cart and write it up. Will do soon.

I honestly do not know what reference is used but I would plan on using the 1.1V reference since the firmware is already setup for that IIRC. I am guessing 4x the resistance should be roughly correct for 4x the voltage.

The 19.1x vs 22k doesn’t make a big difference but it is noticeable as about an extra .5-1 second for the long click to activate. It can be recalibrated for the 22k naturally but it ends up right at the very edge of the range.

19.1k puts it closer to the upper middle of the range which is nicer to work with.

All calibration is done before compiling the firmware.

We built the first FET drivers based on the Qlite 3.04A design and board. It used 19.1 and 4.7K resistors for the LVP. We still use those, for the most part, as they’re commonly available and many of us that bought parts bought by the 100’s for price reduction. So now when we build new drivers, we use the components at hand that are proven to work. No need to rethink something that’s been working for years.

I and some others started using a 22K so we could use an 8.4V 2 cell system with the Zener modification to 6V drivers, mainly, then, the MT-G2. Along came the XHP-70 and 50, and things are taking off. Mostly though, the off time cap works well with the 19.1/4.7K setup, so it hasn’t been modified for consistency reasons.

I’ve built HUNDREDS of these drivers, have quite a few more FET boards and a fair stock of most components. So that’s how it comes to be historical baggage. Comfychair started putting an Vishay 70N02 MOSFET on the Qlite board, replacing the 7135 chips, even flipping the huge MOSFET upside down to make it work. I was amongst those that implemented this at the time, and a lot of others picked it up. Wight started designing new boards, along with MattAus and several others, I found the SIR800DP through hours of eyeballing charts on DigiKey and Mouser, and here we are.

For the record, we had to keep numbers in the ATTiny13A firmware under 255, so the 22K was about the highest value we could use at 245 to keep the LVP with 8.4V cells. I would use 124/112 settings with an 3V emitter, 245/228 with 8.4V set-ups.

It might be of some importance to note that Wight designed but didn’t really build the drivers. He’d put the new designs up on OshPark and let us flesh em out. I went through quite a few revisions this way. And MattAus got so burned out designing for the hot steady flow of requests that he dropped out of the forum altogether. He was a critical component to these FET drivers when it all took off. Wight has been gone as well, for the most part. So our teams that designed and tested have fallen apart under the stress of the rapid growth and of course, they had jobs. MattAus is an Aerospace Engineer working at an airport, not sure about Wight.

Yes, some of the builds didn’t work well, like the Knucklehead. I still have a box of components from building those, to no avail.

I guess some people around here think I got 13000 posts by saying things like hey, cool light! hahaha
We went back and forth at an amazing pace, for quite a while, just like you guys are doing in this thread.
That, and of course, I talk too much.