Well, I have the regulator marked up to 650mV and 8.9A output. I've used the numbers with some of Bitmain's BM1384 numbers to get some updated extrapolations for performance data. The regulator's not running as efficient as I'd hoped (I was hoping to break 87% peak but the best I saw was 85%). The board-level efficiency data is assuming 100mW toward auxilary stuff, which may be an underestimate.

Actual regulator efficiency data based on this afternoon's measurements:


As you can see, the peak tends to be around 3.5A but efficiency between 3.5 and 8A is still pretty respectable.



This is an extrapolation of expected hashrates versus power required from the USB. Each line is peaked at the rated maximum hashrate for that voltage, as given by Bitmain's chip performance chart, and the power requirement is solved from the regulator input vs output power, Bitmain's chip-level efficiency numbers, and the expected 100mW maintenance power.



This is basically another visualization of the above, with the extrapolated hashrates divided out for board-level hashing efficiency (W/GH). The range is somewhat deceptive, as with the current driver the minimum operating frequency is 100MHz so a minimum hashrate of 5.5GH; it also has a minimum increment of 25MHz, so 1.375GH steps - 5.5, 6.875, 8.25, 9.625, and 11GH for these chart ranges.

We've got a meeting in the morning so I'm not sure when I'll get started on working over the numbers for 675mV through 800mV, but if I can keep getting the performance out of it that I'm seeing so far, I'm definitely going to approve the design and finish putting it in the Compac PCB for prototyping. I'm fairly certain the Compac regulator will be slightly more efficient and/or reliable than this prototype, as I'll be a bit more careful with the layout, and the output will be directly tied to the chip. One thing I noticed when testing was load regulation wasn't so great; going from 500mA to 8A output typically required an increase of 30mV of measured output voltage to stay in range for the measurement. I took my voltage measurement from the regulator board right at the output socket, but the Vsense line on the regulator chip is tied closer to the inductor so it's possible some voltage was dropped in the output trace (though I can't see it being 30mV). The chip actually has a sense pin for GND as well, which was tied locally, so it wasn't taking into account ground-plane drops either. The input voltage to the regulator board was measured below the USB jack, so that does take into consideration any terminal impedance losses and such.

Once I get the regulator marked the rest of the way up, I'll post some updated charts and then rig up a full mockup of the Compac to get actual expected performance data for board-level hashing efficiency. That'll probably take a couple days to get all the data, since for each voltage level there'd be potentially a dozen workable hashrates (and that's if we don't implement my generic clock code so we're stuck on 25MHz increments, min 100MHz) and I'll need to keep a watch on error rates.

Sorry, I don't have any photos of the test rig all set up. Maybe tomorrow. The only additions from already-seen stuff are the jerryrigged FET dummy load and my worked-over scope probes. It is kinda cool to see everything socketed together though.

I laid down plenty of solder in the parts that are going to be handling upward of 10A, especially the parts I want low resistance to keep down ripple voltage. The corner is cut on the regulator board because there's a capacitor on the BM1384 breakout board it needs to accommodate.

I do have a scope, two in fact though one's trigger functoin doesn't work. It's about 30 years old. I have a 50MHz Rigol. I'd like to get a 4-channel one of these days; it'd make the efficiency testing I'm doing right now a lot quicker. I'm having to swap probe pairs to measure each meter board. That does cut down on ground-loop errors though, which is good, but it's certainly not making the task fast.

I have completed 550mV and 575mV curves from 500mA up to about 8.5A out. Now just to get 600-800 in 25mV increments, at 500mA output steps. Something like 160 more data points, each one recording 2 numbers, switching probes, recording two more numbers, then doing 5 calculations on those numbers to get actual current, actual power and efficiency. I might be done today.

Also, the 550mV and 575mV both peaked at just shy of 83% at 3.5A output. If the higher voltages stretch that a bit (say, 86% at 625mV) that could put us pretty close to the goal of 8GH off 2.5W input.

Contact Oregon Mines to get ahold of Matt Carson, the guy in charge of Minersource. He's now Oregon Mines' VP of Sales I believe. There's more discussion around this in their Group Buys thread.

I'm working up a one-FET dummy load right now, nothing fancy active control, just a multi-turn pot on the gate. Coupling that with the two power meters (source and load) I'll be able to get efficiency curves for the regulator pretty quickly and I should have some fresh extrapolations for the hashing efficiency later today. And yes, I'll take more electroporn pictures for y'all.

I'll admit, all the time wasted trying to iron out the old regulator circuit was fairly disheartening. But I'm much more confident about this one, partly because it's less complex and more modular - both of which things I really like in a design. I'm starting to wish I'd gone this route to begin with instead of looking for a monolithic solution. It does save quite a bit of space, but at the cost of plenty of suck, apparently. And there are almost no options for going below 0.6V natively in a monolithic buck regulator.

I received some busted S5s last week, and when I have some downtime on the Compac project (while waiting for prototype boards to come in) I'll probably be back onto my inline regulator project. If I can find an S5 board among the pile that still works, I'll do some testing on low-volt and see what in the design is the failure point. I suspect the clock signal is being compressed too much by the level shifter and when the voltages move around it's not hitting thresholds. If I can find a straightforward fix to get them going at lower voltage I'll post about it.

Got some more infornography for you today.

Here's the current two-chip rig, with chips in parallel. This is what's been running since Friday or so. I'm still guessing it's signal integrity causing issues, since with the two boards mounted directly (the shortest possible inter-chip signal path using our breakout boards) the errors dropped from 50% to 25% (based on hashrate efficiency).



And here's some new toys I built today.



The center stick is a test board of the new regulator design. I had trouble with it for a few minutes, but it was just the bootstrap pin not soldered properly. I took it up to about 7.8A without noticeable trouble (I wasn't observing output ripple though, just the average with a multimeter) so I'm already liking it better than the previous design.

The guy on the left is a new USB power meter. I haven't attached the second socket yet, as I was using it for impedance measurements. This guy takes in USB signal+power in the B socket. The currently-headered pair of pins right above the B socket allow for USB power to pass through. If that's not jumpered, the 4 pins adjacent (two +V, two GND) are the power input. The two pins past that (labeled A and G) are the current meter pair; voltage on the A pin (with respect to G, the ground) is approximately 100mV per amp of current. After calibration (linear regression on 12 data points from 50mA to 3A) the actual current is 0.99 the measured value. The second set of pins, by the USB A socket, is for measuring the output voltage at the sockets. The first socket (present) gets USB data, but the second will be power only. I set one up with two sockets because I want my regulator board pulling from USB to emulate actual source impedance conditions, and the signal socket runs the USB-UART adapter (which is also powering the PLL and IO LDOs on the breakout board).

The guy on the right is a stripped-down version of the above, except designed to socket right inbetween my regulator board's output and the BM1384 breakout board. It's only got two pin sets, one for voltage and one for current. A linear regression on its calibration (14 data points between 800mA and 10A) had up to 3% error in the low end (below 4A or so) and I wasn't really happy with it, but for some reason 0.97(Im^1.01) yields between a maximum error of 0.83% with an average error of 0.33% so I'm okay with that.
There's a CR2032 button cell socket on the other side, so this guy is a self-contained meter. Unfortunately with a Vcc of 3.0V, I can only measure up to a bit below 10A, since I'm using a 3mOhm sense resistor, DC gain of 100, and my output is divided by 3 to scale evenly to 100mV/A which means my maximum voltage range at the output pin is 0-1V.

Now that both meters are characterized and calibrated, I can hook them up to the new regulator board tomorrow with an adjustable dummy load and get some efficiency curves. If it behaves how it should, after that I'll tie in a single-chip BM1384 board, the USB-UART adapter and LED circuit and get an actual matrix of Compac W/GH performance data. Expect news tomorrow - hopefully good news, but all I can guarantee is there should be news. And also sandwiches.

If the regulator works how it should, I'll finish redesigning the Compac board and might post some sample layouts. I'll probably also put together a basic two-chip BM1384 breakout with both parallel and string IO/power options so I can better test chip cooperation for Amita and TypeZero designs. As soon as both of those boards are ironed out, we'll be sending off for prototype PCBs. If the Compac board tests well, we'll be looking to ship a few out to our selected testers for finagling. If they report back with favorable reviews, we can start looking into batching a big order.


Also, on a side note, Novak's getting really darn good at toner transfer etches. That regulator chip is a 16-pin 3x3mm QFN package, and the current sense amplifiers on the meter boards are SC70-6, which is about half the size of a SOT23 and twice as many pins.

I might point a BM1384 board at it if I can talk one into running at 6MHz instead of 100+

I did put out Blade kits, but nothing for the Block Erupter (except sending a handful of parts to one friend). I had at one time seventeen custom-overclocked to 460MH/s but I didn't make any kits for them. I think I still have two, one that works and one that smoked - I also killed the first one while trying to overclock.

I was thinking it might be interesting to rig up a BE with a software-adjustable voltage and clock but I don't think there's enough free pins on the micro. Maybe hardware-adjustable would be good enough.

If by "cannot do" you mean "choose not to do" - probably for reasons of selfishness or complacency - we agree. People are often inspired by stories of exceptional people, not realizing that the majority of those people are ordinary people who chose to do something exceptional and put forth the effort to follow through on that decision. Inspiration is useless unless the inspired choose to act.

Yep. Novak and I share a lot of opinions with grouchy old people, despite being in our twenties. It's pretty fun.

I tend to ask more from people than is normally expected, but I generally only expect between a half and a quarter what I'd expect from myself. There's a lot of the reason I don't answer every stupid question politely, because I figure answering the question once should be enough and anyone wondering should seek the answer first.

The things I despise most about American culture are summed up succinctly as "greed". Laziness is based on greed, a selfish desire to be sedentary at another's expense. Americans love money, but don't really understand the purpose of money. Simply having a lot of money is stupid if that money isn't working for anything. Money is a tool. Money is simply a means of exchange between goods and services, so amassed wealth is actually a detraction - it hurts everything else around you.
America loves cities. Our political, economic and social systems are set up to reward people that congregate in cities at the expense of rural areas. But rural areas are where almost literally all of the actual production happens, where all of the actual resources exist. Cities are huge consumers because the demand curve is shifted heavily by population density, and because the only resources produced in cities are management (which is to say, people charging inordinate sums of money to make decisions about things they never see) and maintenance services (which is to say, the oft-underappreciated toils required to keep management alive and comfortable). Cities are places for stagnant wealth to amass, and for some of that wealth to shift from non-producers to other non-producers, and leave all the producers out to get walked on.

I can't really claim to be a patriot, not like a lot of people I know in military and civil service. Heck I'm not even registered to vote. But one thing I can't stand is businesses leaving the country or jacking up prices so they can increase profit margins. Shifting your central offices to Canada so you can avoid paying US corporate tax? Bastard, you just lost my business. Shifting your manufacturing to China so you can avoid paying a fair wage for a reasonable workload? Thanks but no thanks. Asking your customers to pay you $100 per unit on a thing that only cost you $30 to make - $5 in materials, $5 in assembly labor and $20 in management wages - because you think your ability to wear a necktie and occasionally make selfish decisions earns you the income of ten thousand households a year? Screw that forever.

So yeah, there's good reason for large corporations to specialize, and outsource, and do all kinds of other stuff to protect their bottom lines. But in my mind, a great way to keep costs down is to pay MANAGEMENT a fair wage instead of worrying about making bosses millionaires while profiting from slavery. Find people willing to work hard, with an eye for quality, and let them do it. Then find people willing to stand in front of them (not above, in front) to keep them organized, and pay them about the same amount. And then find someone willing to stand in front of those guys (not above, in front) and keep them organized, and pay them about the same amount. Increasing wages by an order of magnitude with every step up the hierarchy is stupid. If delegation is in any way effective (which is to say, because you hired competent people capable of making decisions and being accountable for them), the workload at any tier of the hierarchy should be about the same as at any other level. Even if Novak and I's concern ever becomes a multi-million-dollar business, we'll probably never get paid a six-figure sum. If you make $1M a year and oversee 100 employees, you could reduce your wages to twice the US median family income (six times what I'm living off now) and still give a $9000 raise to every single employee without increasing costs one whit. Only greed prevents this from being standard practice.

I have pretty stiff opinions about how to do things, and most of them are based around not being a greedy dick. Please don't clutter my project discussion thread (wherein we demonstrate how effective we are at not being greedy dicks) with suggestions that I should ever do (or even endorse) otherwise.

Profit margins and excessive management wages also put upward pressure on costs. In the last 50 years, adjusting for inflation, the average worker wage in the US has not changed at all while average CEO wage has increased by an order of magnitude. When there's a job that can be done by anyone with basic dexterity, there's no reason to ship your materials twelve time zones away when the guy down the street from you is looking for work. My political leanings have always bent toward isolationist, and the 5:1 trade defecit with China doesn't help convince me I'm wrong. Then again, I also don't like unions - bashing people over the head to get what you want is only required when what you want is undeserved, or when your boss is a jackass, and then there's all the artificial limitations you impose on yourself when you sign away your individual rights to an organization.

So yeah, imaginary lines on a map are generally meaningless. But when I KNOW a guy that doesn't have a job and is about to get kicked out of his house, how @!#$ing dare I hire someone 8000 miles away to do work he could be doing just as well, without communication hassle of time zones, or added transport costs, or supervision costs to maintain basic quality control? How @!#$ing dare I export the economy of the imaginary construct of my country, my state, my county, my town to some hole on the other side of the world just because I want to save a nickel per unit and pocket most of that nickel for myself? Hell no. I'll pay my neighbor to do good work before I'll pay anyone else. If the cost is too high, either I'm paying him more than the job should be worth or, God forbid, I'm paying myself too high a fraction of the revenue I'm not laboring to create. Domestic labor only causes undue pressure on cost when too many of the people involved are greedy.

Don't worry Meech, I'm currently reclined on the couch, pantsless, book in hand. Doesn't get much more breakful than that on a Friday night.

Novak was just commenting earlier today that Friday afternoons are sorta our weekends. After a full week of work we tend to slack off on Friday afternoons (he took off today about 2PM) and then come Saturday we end up coming back in to work because, hey, we already slacked off on Friday. I spent pretty much all of the post-cheeseburger evening reading webcomics and haven't done anything productive since 4PM. This weekend will be pretty slow for me; with the exception of some basic Amita testing and rough PCB redesign, I have not much minerwork to do until fresh regulator parts arrive Monday.

I choose the best of both worlds, and tend to write mean things in long, well-thought-out posts.

I'm hoping for a regulator efficiency such that we can run a Compac stock at 150MHz (8.25GH) off a 500mA 5V port. I also want the Compac to be capable of taking in 1.5A of 5V comfortably. I've only got one powered hub to test with, but it's one I rebuilt with beefed-up 5V lines, a couple extra capacitors and a Molex for power.

Also, I wired one board straight to another board in an effort to cut down on potential wiring issues causing comm noise or clock jitter that might be hosing up my hashes. I'm still getting HW errors (both boards worked flawlessly as single boards), but clocked at 150MHz I should see 16.5GH; cgminer is reporting about 13 and the 3-hour average on Eligius is 12GH. That makes it look like half the shares returned by the second board are hosed. Better than all the shares, which I was seeing previously. Maybe tomorrow I'll work on chaining up and putting two boards in series. I wonder how the level shifters will help knock out any wiring concerns. If I still get moderate cooperation between the chips I'll probably send off for a modified two-chip breakout board when I send off for Compac prototype PCBs to see about improving performance and getting layout ideas for the Amita board and TypeZero.

But right now it's Friday evening and I'm full up on cheeseburgers and taters and about eight glasses of sweet tea so, gonna relax a bit instead.

He probably didn't mean yours specifically, but a general "you". A lot of folks have cheap 7-port powered hubs which would be okay for stock settings but nowhere near what these sticks should be capable of.

The most steady current I saw was about 1.4A, but mind you that's for two chips on a regulator which should be 5-10% less efficient than I expect for the final.

Thanks. Yeah I need to track down the source of the hardware errors in parallel before working on stringing two boards, but hopefully I have something stable before the new regulator parts come in. If that's the case, and the test board regulator works as well as it should, we could be ordering Compac prototype boards (and possibly Amita boards) end of next week.

If it were my company we'd do what we said we were going to do. There's enough profit to be made from being trustworthy and reliable, no need to squeak margins by screwing people over. It sucks how much greed there is in the bitcoin economy (not just from manufacturers - nigh every participant is steeped in it), but greed is a basic tenet of human nature and few people ever see past it.

So, update. I designed two fresh regulators, each using sturdy external FETs. One is using a driver chip with a convenient package but a lowerbound of 0.62V; the other has a sorta sucky package but it also has internal compensation and an adjustable reference down to 0.5V so I have it calculated for an adjustable range of 0.55 to 0.80V which should give people plenty of room to play.

We have parts coming to prototype these regulators, and I'll start with the wider-range one because I'd rather have that lowerbound available. I probably won't be able to play with them until Monday.

I set up two boards in parallel since I have them running below a 30A regulator right now. There's plenty of current headroom. I'm seeing fairly frequent hardware errors, which could be due to heat or crappy inter-board wiring. It's currently sitting at 150MHz, which should be 16.5GH but I'm seeing about 14.5 probably because of errors. I'm going to play with chip addressing a bit and see if it even cares what order chips are in; so far I don't think it does.

Tomorrow I'll tweak things a bit and see if I can't get them running well in parallel at 175MHz (19.25GH or 200MHz (22GH) before I start messing with putting them in a string. Hopefully by cheeseburger time tomorrow I'll have two boards strung, which should shift the regulator efficiency up 5-10%. Very excited about that.

Today's Porn:

The two green boards are the BM1384 breakout boards. The unwieldy length of multicolored cabling between them is the fairly jankety inter-board comms. I stole one of Novak's USB adapters, which is currently running off my powered metered USB port. The dark board off the right-side breakout is a corner off a dead Garden blade, rehashed slightly as an adjustable-output 30A converter that's drawing power through a USB jack plugged into a dummy socket off my bench power supply. The fan is swiped from a retired AM Cube, just in case. I'm pretty sure it's the wiring causing issues though, probably a pretty noisy environment to be running 1.8V 115200baud UART lines through without shielding.

The crappy getup will be mining at Eligius on the burger donation address (1BURGERAXHH6Yi6LRybRJK7ybEm5m5HwTr) in case anyone wants to see how it behaves overnight. If I have time to get a series string of chips going tomorrow, we'll have the first taste of the Amita, and I'm pretty stoked about that.

Oh hey yeah, I sent you an email about some buck chip options and ideas from the last six hours or so of research and calculation. Think about that for a while and I'm gonna fall asleep.

But yeah the test chip, I think it was at 125MHz and it was definitely submitting shares at about 580mV at one point. Awesome.

Yep. I did some quick calculations and it should only take about 220uF to handle the burst transient tolerably before the regulator can kick in. We had something like 360uF on the output and it still bottomed pretty good. Has to be overcurrent protection kicking in (at about 11A). The new design I'm working on right now (literally) I can set the overcurrent pretty much wherever I want, so I'm gonna make sure it'll be good for those kinds of burst transients. If we start testing higher voltages and frequencies and find that the burst requirement gets worse, I can just take the overcurrent a bit higher with a change of one part. The FETs will be good for a heck of a lot more current than this thing will ever need. We're aiming to make it capable of maxing out a 1.5A port in steady state, which it should hopefully find around 300MHz (16.5GH). I mean, it'd be pretty great if it could do more than that. Since we can do pretty much whatever the heck we want, I'm gonna give it a full range of 600-800mV so you can clock it up to 400MHz if you can keep it cool enough. Whatever.

Yeah, the schfifty-three we put on there I think has... 3x 68uF output caps? And it behaved admirably even at starting the chip on about 640mV 250MHz. I didn't test the chip higher than that but I probably will tomorrow. It really wants a higher voltage to run at 300MHz stable anyway. I'm kinda surprised it started at 640. Our homemade regulator (well, except for the inductor) will be about as beefy as a schfifty-three circuit (by which I mean the TPS53355DQP) which has been the favorite of miners for a while now - S1, S2, S3, AM Cube, AM Tube, BTCGarden AMV1, New R-Box and a few other things use 'em without any problems. But they're like five bucks per chip. I can build an equivalent non-monolithic circuit for under $2 in silicon, with better power dissipation since it's less dense.