Probably. The tune works off not just the total hashrate but the health of individual chips. If a weak chip is getting starved and stops returning nonces, the software will notice, drop the target a bit, restart the string and ramp up to the new target. VH did that specifically because something like a 90-92% target could still be met even if one chip isn't hashing at all, which would cause that chip unnecessary damage in the long run and kill the whole string.

I'll probably start that flashlight thread tomorrow, but earlier today I had an idea. Phil, that low-brightness model you wanted, it'd be a shame to turn an LED capable of 800lm down to 25lm but I already have a reel of white LEDs I built my shop lights out of that max around 120lm (test is 36lm) which would work perfectly for a low-intensity model. Just did some checking for compact lower-current regulators and the winner's one I'm already testing with so I could use an existing driver pill design, replace a couple parts and have the same control system dialing in 0-240mA (0-120lm) instead of 0-2.5A (0-800lm) for the big one. Would save multiple dollars off the parts cost and make more sense for the end product. I can play around with the specifics for that anytime this week.

That sounds uncomfortably like theft and my ethics don't allow for deceptions, profitable or otherwise.

I really only hit up here and Speculation. Don't care about much beyond hardware. Might be worth starting a self-moderated topic (if that option is available) and transferring the talk over there, because y'all have some good ideas.

Right now I'm just interested in pocket-size, but if it proves robust and I can find a decent source for housings (I have only basic machining infrastructure, enough to mill and tap heatsinks and cut sheet aluminum for the hubs and miners) pretty much anything can be made. Matching a 10 million candle light with LEDs will probably be cost-prohibitive however.

The flashlight I'm working on now, I had already bought one I about 80% liked years ago and I recently discovered I can still buy the same housing in bulk so that's the starting point. The housing was fine but the electronics were not that great. The light I found that was basically perfect went out of production about six years ago so a similar housing might have to be a custom job.

Phil, 25lm is definitely doable. The hardware is built for way overkill but the actual brightness level is set in software so a quick change to the program (which I could do here in like five minutes, or anyone with a PICKit and some wire could do at home) can set it anywhere you want. The test setup I have now is a lower-current approximation of the final product, and the intensity starts to ripple a bit around I think 40lm (a consequence of being optimized for 200-600lm) but that's still pretty low.

I will, just because I know some people will want it, end up programming a multi-mode version, but I'll set the button delay to something like half a second with no persistent memory so it should be really hard to "accidentally" cycle into a different mode. If a two-mode was set up for low first, then high, Phil, one light could do both your needs.

USB charging would not be a built-in function. You'd have to pull the battery and put it in a separate charger. But I'll definitely look into building a USB-powered charger. It's gotta be incredibly foolproof though, since as Samsung demonstrated not long ago, lithium batteries can be pretty dangerous when mistreated.



So, we're noticeably off topic. Where's a better place to move that line of discussion so we don't glut up the Terminus support thread with non-Terminus non-support chatter?

I wouldn't sell you something that easy to break. Goes against my principles.

Yep, sure can.

I talked to a contact last fall who does good business with Canaan - he runs an I believe multi-megawatt datacenter entirely full of Canaan gear. Told me they would sell chips and even offered design assistance, but the minimum buy-in was $10k, where I was only looking out lay out around $2k to get started with another outfit I'd been talking to for a while. I figured I'd keep Canaan as a backup plan.

It is not USB rechargeable, but I bet I could build a USB-powered recharger for the swappable batteryies without a lot of trouble since the cell voltage peaks at 4.2 volts. The same type system I use for dynamic current regulation inside the flashlight could be easily ported over to a CC/CV lithium charger.

I know we're way off topic but that there is a good idea.

And because it's GekkoScience, it will have none of what that guy just suggested.


(like, the brightness would be set by changing some firmware written in assembler and then re-flashing the board)

The only thing far enough along to talk about is an LED flashlight. I like the lithium-battery aluminum "tactical" flashlights that fit in your pocket, but I don't like how hard it is to find one without all the stupid strobe modes and jagged edges and stuff. So I'm making one that just, you know, turns on and off.

Course because it's GekkoScience, I'm overbuilding the electronics to handle putting out around 800lm even though the default is 300. And because it's GekkoScience, I'm making it hackable such that the brightness is software-adjustable and it has the capacity to do multi-mode as well. If I can find a good lensed housing I'll make one with an adjustable beam too.

I was kinda hoping bitshopper would show up to address Europe stock questions. He's had an order with us that's finally getting shipped so he'll be back in stock of things pretty soon, but I don't know the state of his order backlog so that might not have helped much.

As for the new shop, it's still pretty disorganized and we still have a lot of stuff in storage but the new place is fully functional. It's also a lot better insulated, which is great since we hit zero degrees (approaching -20C) the last couple days. We finally got gravel in the driveway, to replace all the mud and ruts, and my living quarters upstairs are almost finished. Some equipment to help stabilize the internet connection should arrive next week, and the kitchen still needs a fair bit of work. But what matters for you guys is the workbenches, machines and testing stations are all up and running. Goods have been rolling out all year to date. I'm even finding some time for new designs, some of which are not bitcoin-related since I pretty much work on whatever I want.

The barrel jack is rated for 8A. Might be good for more than that, but I can't officially recommend putting more current than that through it.

A 4.5A brick will work, but you'll be limited on speed to down around 500, maybe 600GH on a low voltage setting.

The max amount of current will depend on your frequency and voltage setting, and to some extent how hot the unit is. I've not had one go much over 10A on my test bench. The 6-pin connector will survive 20A if your cables are good quality, and that's way more juice than the miner will ever ask for.

Focus on home-friendly miners, you say? That sounds familiar.

So you think they're gonna start building miners with under 2000W power draw again? Have they ever gone backward in power draw, since the S3?

I stepped out of my workroom about a year and a half ago and smelled smoke. Usually when I smell smoke it's coming in from outside since there's a state lot down the road where they burn brush sometimes. So I was checking out the big door and when I turned around I noticed smoke coming over the wall from the hosting room.

Had a guy complaining that his miner was getting too hot and wanted it moved to a cooler shelf, so I moved it onto a shelf with only one other machine. But that miner's middle board burned up and ended up torching both miners and part of the cold-aisle isolation backing before I got it put out. Fortunately we kept a big extinguisher at the end of the shelf, and it only took a few extra seconds to kill the power just in case.

Had that happened when I was gone overnight instead of during the day, the whole place might have burned down.


More recently I noticed miners and email went down while I was out of town. When I got back in, I went right out to the shop to check on things. Thought maybe power was out, or the shop had burned down for real this time. Well just almost everything was still powered up, but the outlet the battery backup upstream of the modem was plugged into had burned itself out. Whoever wired that part of the building used the "stick the wires in this self-catching hole" option on the outlet, instead of doing it right and using the screw clamps, and the connection went sour, heated up and melted down the socket. Good thing it was in a metal conduit box. So I replaced that and got things back up running pretty quick.


That's about the biggest issues I've run into.

Building an FPGA miner goes way beyond what I currently have the time and desire to do. Building a board with an FPGA on it is fairly trivial, but making the FPGA into a miner requires skills I don't yet have, and making it operate with any reasonably efficient logic implementation would take ten times more of those skills.

If you want alts, talk to FutureBit.

Again, nope. I don't really care about altcoins so I don't build gear for them.

I'm also a fan of education, so if I can "encourage" people to learn a little bit more about the systems they want to use, rather than doing all the work for them, thereby increasing their overall understanding and hopefully improving general competence, well that's good too.

No. I build things that I think should exist but don't, and I still believe hobby and home-scale miners should be a thing. If I stop making them, they won't get made.

Those features? No. I don't have much interest in building tiny self-contained computers that also mine. If I get to bigger things than the Terminus there'll be better provision for a "bundled controller" but not for the Terminus series.

Jstefanop can put more R&D time into the complexity of his Futurebit software and hardware because he doesn't also build them, far as I know. Leastways he doesn't do all of it. My stuff is entirely in-house, including board population and metalwork for the heatsinks and housings, so I have to coordinate the production line and build all the tooling as well as the design work, which means keeping things simple. Plus "keep things simple" is a core component of my design philosophy.

Running on Linux is pretty easy. I've set up enough testing stations that, except for the dependency list, I have just about every other step memorized.
The only trick to it is making sure you don't put a zero where a capital O should be in the one line.