Topic: Why haven't we seen PDIP chips the traditional hobby DIY guy could use? (Read 1636 times)

To be specific, QFP-EP. Have as good a heat transfer as any QFN with belly pads, but a lot easier to verify placement and proper soldering. Leaded packages would be nice to see, folks wantin' to play with designs and such.

QFP...
What about heat transfer

Ah. I spent a lot of high school and my first few years of college programming, and then remembered how much more I enjoy hardware work instead so I shifted focus back to that. To me, building is much better than typing. I'll probably get into FPGA sometime in the future but for now I'm satisfied.

Grave mistake or not, it is going to have a programmable microcontroller (most likely 8-bit AVR) with firmware. USB 2.0 has enough bandwidth to support all that, but if we're working with Bitmain chips again it'll be running on bare unaddressed UART so we'll either set up a simple control set the chips ignore or something using accessory pins on the converter to talk to the micro. I2C temp sensor and voltage control and PWM fan speed. Coded in assembler, probably less than two hundred lines.

If we keep on Bitmain ASICs, they take in a fixed clock and each chip has an internal PLL multiplier so frequency setting is done there. But most everything else you mentioned is something I'm planning. Thanks for the advice.

Also, I tend to be more impressed when people choose to use screwdrivers and pencils than automate everything ever. Remember, I like simple.

I'm not going to get into a fight over who's the better engineer, because you're probably going to win - probably with both knowledge and experience. But to quote Abraham Lincoln, "I will do the very best I know how - the very best I can, and I mean to keep on doing so until the end." Which requires doing, so I'm gonna.

Fun, and a perception of play, is a very individual thing. The only 'fun' that I can recall that involved soldering irons was when the professors left the room, we over-applied the rosin-based flux and played really loud "Smoke on the water" by Deep Purple on the tape recorder and tried to come up with alternative lyrics to that song. I'll be perfectly happy to never need to touch the soldering iron again in my life and do all my work by supplying design plans to the workshops.

I'm glad to hear that you are considering next version.

But you are going to commit a grave mistake if its going to have a programmable microcontroller with firmware. USB 2.0 has enough bandwidth to support all the required circuitry in the completely dumb-terminal mode.

Unfortunately I don't know the specifics of the ASIC you are using, but the general design points are as follows:

1) absolutely no firmware or microcontrollers can be run on the same power supply as the mining chip !!!one!eleven!
2) simple, high bandwidth USB to serial converter chip with high noise immunity using SPI, I2C, plain serial with hardware flow control, etc.
3) pre-calibrated SPI/I2C thermometer IC touching the same heathsink
4) external clock generator with programmable frequency and duty cycle.
5) SPI/I2C interfaced programmable voltage converter
6) if the chip doesn't have a an on-die temp-sensing diode maybe there is a way to rig up a temperature sensor by doing e.g. IO-pad leakage to time converter with external RC components and measure chip temperature indirectly by measuring time in the host software.
7) obviously, if there is an on-die diode then just hook it up the external thermometer IC.

I'm trying to. But I sense that you are probably not aware what can a single person do with e.g. a copy of National Instruments LabView Student Edition when that person doesn't need to practice screwdriver artistry and work around firmware limitations/bugs/faults but can simply do a proper laboratory workflow by setting up operational points (voltage,clock parameters) and measure the outputs (temperature,hashing speed & error rates).

One additional comment about the lame design choice that Spondoolies made: they used on-chip temperature-to-digital converter macro instead of a cheaper and simpler diode with external thermometer IC. The internal SSN (simultaneous switching noise) on their chip made their temperature readouts rather unreliable. And they knew very well that the SHA256 mining chip is very close to the theoretical maximum toggle rate of any practical digital circuit.

I mean no offense to your opinions, but I've had a heck of a lot more fun designing the Compac than I would have had working on an FPGA board. I like to work with hardware instead of pretending to work with hardware, and programming a thing to think it's the hardware I want is not nearly as fun as getting iron burns.

ntegrating a microcontroller with digital voltage control and temp sensing will go on the next model up. Don't dis the manual voltage adjustment until you consider that pretty much no other stick miners have any voltage adjustment inherent to the design, manual or otherwise. Also consider that the whole thing was designed, tested and built (and financed) by one guy - the additional driver and firmware code and hardware to integrate all that stuff into a single-chip stick miner not only adds substantially to my workload, but also the price.

On the Spondoolies comment, I like that they put in a lot of work making their machines very adjustable and very sensor'd with readouts on just about everything. But the design topology itself is pretty much the exact opposite of what I'd like to see in a miner (and so is the price) - and yeah, the inordinately long air paths is suboptimal to say the least. Most of their miners I look at as "a showcase of how not to design hardware", but they're by no means alone. Nobody can do everything right, so everyone is doing something wrong - and some are doing a lot of things wrong.

Mining ASICs that were easier to install on boards (and verify correct installation) would be nice. I'd like to see pinned parts. SMD are a lot easier to work with in bulk than PDIP, but BGA and cut-up QFN (for example, BM1382) aren't exactly user-friendly. Give me a QFP any day.

http://www.ztex.de/ is opening for me just fine. Ask them or search around on the forum for the used ones. They won't be cheap, because they are actually very well designed.

The cheap ones are e.g. Enterpoint's Cairnsmore with their horrible ringing on the clock signals that almost killed the whole business.

Are those guys still selling?

Substitute "crutches" for "solutions".

The mining hardware market is a showcase of how not to design hardware. You can find pretty much every possible dumb fault in various hardware miners. By "dumb" I mean "something that would be nearly trivial financially to fix and productize correctly".

ngzhang was the closest to an ideal with his dual FPGA board. But he routed only TxD and RxD from the FTDI serial chip to the Xilinx Spartan, despite FTDI supporting 8 or 12 signals. How much time they then spend trying to fix and work around this problem at the protocol level!

In the ASIC group the Spondoolies would be the closest to the nice platform for experimentation. But they completely fubared the airflow design, heating some chips in the loop with the exhaust air of the others. How dumb is that?

The way I'm serving the market of experimenters is by warning them away, unless they want to learn about the pathologies of electronic design. Buy and disassemble an old radio or TV, or even an old computer.

There's almost nothing in the Bitcoin hardware milieu that I could with the clear conscience suggest as a learning and educational toy for the interested student. ztex.de is the only exception, but they never really catered to the demands of the Bitcoin market.

Or perhaps it enables it by allowing people to add their own solutions   But people have to be willing to put in the effort, and I just rarely see that sort of desire.

You could volunteer to cater to that market

You... you're going places....

The painfulness of driver code is mostly Con Kolivas, Kano's & Luke-Jr fault. They intentionally made their miners frighteningly convoluted "bags of drivers". The early BitFury miner was a thing of beauty and elegance despite having to straighten out the jumbled data due to the metal misrouting in his original chips.

In my opinion lack of voltage control and temperature measurement (not just overtemp warning) is what nearly completely removes the learning and experimenting potential of those devices.

It is one thing to "make them work" and the other thing "to make them work and make them worth studying and learning".

The 2nd market is there, but isn't getting served. Such a shame...

Then what are you waiting for? Get an FPGA kit, it will completely cure your jones for DIY. All the toolchains I've seen support schematic capture as one of the design input method, so you aren't saddled with coding in Verilog or VHDL, and go back to the familiar concept wiring the gates.

For the sake of old times buy yourself a can of rosin in the store that serves musicians (it is used to maintain bows for violin, cellos, etc.) and slowly melt it to get the whole sensual experience of the old style electronic DIY.

I don't know how true that is, given the very existence of that project (GekkoScience Compac, for those wondering - on sale in 8 minutes).
Even going back, there were plenty of BitFury- and Avalon-based designs being made both small and large scale.  Even some later chips wound up in third party designs (from Spondoolies-Tech to HashFast's, even BFL).

In terms of end-users, this is true.  Short of the occasional overclock that was pretty common, I think I've probably done the most tinkering on small scale miners.  But then, I wouldn't really expect most people to start getting creative with existing Bitcoin mining designs, given that most of the audience is just people who want to mine with it.

Bit of an annoyance to write the matching driver code, though  And while adding a temp sensor on the heat sink (I don't think the BM1384 has an on-die one) could allow for warnings, it would have to be further interfaced again with driver code in order to actually clock things up/down based on temperature.  The Avalon Nano is a nice example of a StickMiner that has all that already on-board though


I'm glad to say that the Compac strikes a bit of a middle-ground, but overall the market has evolved to barely serve any people at all, preferring instead to sell cloud mining contracts.  Inevitable, but unfortunate nonetheless.

As for FPGAs - I agree, and they're still useful when one gets bored of Bitcoin mining

Aaaah the good ole iron burns... whatnot using your blistery fingers to try to fix a prototype....

But yeah, i've been intending to get an FPGA since i used them on the university and was wooed by VHDL and Verilog.... it has been like 12 years since i last touched an FPGA 

I read your review of some upcoming USB stick miner and it shows that the Bitcoin community completely missed the ball on creative experimenting.

It wouldn't increase the cost much to include the proper software-controllable voltage regulator instead of a lousy trimpot. Same with adding a pre-calibrated heath-sink temperature sensor IC. Does the included ASIC chip even have an on-chip diode that can be used to measure the die temperature?

The marked evolved to serve "compulsive miner" people, not the "learning and experimenting" people.

Dude, if you are longing for the breadboard days you just need to get yourself an FPGA development board. The cheapest ones are under $100 and for sure they can mine coins. What do you think is my avatar?

But you will have to move with the times. Search this forum archives for the mentions of FPGA and you will find plenty of interesting resources.

Unless you are one of those who are really longing for the soldering iron burns, the FPGA kit is all you need to make you happy. With FPGAs you won't even need the expensive logic analyzers, the development toolchains have a Chipscope, which is basically a software logic analyzer connected over JTAG/serial/USB interface.

 

There are SMD proto boards, including some that make soldering SMD parts fairly easy by having a pre-soldered track that you just push up to the side of packages like QFNs.. bit costly though.


If but not for the lack of getting old BitFury chips, the NanoFury project would be just about perfect for that


Probably more Adafruit's domain - they accept Bitcoin and have done some Bitcoin-related articles.  SparkFun not so much.  But again, they'd have to be able to source the chips - unless you'd be okay with just a bare board, but then that's something that can be whipped up in no time and sent off to one of the board fabs (OSHPark is nice and hobbyist-friendly with quality boards).  I do wonder how stable that would be, though.. routed out to various pin headers? mmm.

That's too bad, DIP packages are what most of us can handle on a hobby project... you know, protoboards and such... would be fun to have a way to experiment with mining chips on everything, or even a diy miner kit... but i digress....

It's just that i see sidehack having all the fun, and i want some cake too hehehehe

On a side note, i guess a BM1384 breakout board would be a product Sparkfun would sell nicely

Not just that, though - the packages are just physically large, even SOIC.  Example from my Block Erupter USB teardown:

That tiny rectangle in the center is the metal substrate for an octal d-type flipflop.

Granted, a Bitcoin mining ASIC would have more silicon, but look at all the space the lead package has to take up.

Then there's possible issues with parasitics, thermal management, and.. well, QFN is about as low as a modern mining ASIC should probably go.

Yes true the days are all gone. These days its all about space and efficently. And with the right tools you can solder your own BGA SMD components.