Product design blog

Recent posts

  • Raspberry why? 3 unfortunate things about the Raspberry Pi hardware design Are you using a Raspberry Pi B2 or thinking about getting one? Read on! Disclaimer: the idea behind the Pi is great, the community around it is amazing, the price ...
    Posted Sep 28, 2015, 3:43 AM by Georges Meinders
  • Upgrade to Windows 8.1 (You might live a few hours less) This is a little story about what happened to me when I obeyed Windows 8.0's suggestion (nagging) to upgrade to Windows 8.1 on my laptop. See it ...
    Posted Aug 8, 2014, 4:44 AM by Georges Meinders
  • So you want to get crypto coin mining? It sounds too good to be true: start Bitcoin mining and get rich doing nothing. There has a been a lot of fuzz in the media about Bitcoins lately and ...
    Posted Dec 22, 2013, 2:52 PM by Georges Meinders
  • First post    Megahard will be posting content on hardware design engineering related subjects. Expect an engineer's general brain farts, reviews of attended events, neat technology news, interesting science news, CAD tutorials ...
    Posted Aug 14, 2013, 8:53 PM by Georges Meinders
Showing posts 1 - 4 of 4. View more »

Raspberry why? 3 unfortunate things about the Raspberry Pi hardware design

posted Mar 30, 2015, 1:45 PM by Georges Meinders   [ updated Sep 28, 2015, 3:43 AM ]

Raspberry Pi
Are you using a Raspberry Pi B2 or thinking about getting one? Read on! Disclaimer: the idea behind the Pi is great, the community around it is amazing, the price is great, coding hardware with Python = cool, but the hardware design itself? ... meh... 

Here are some of the more common reasons why people run into trouble with the Raspberry Pi B2 due to some unfortunate hardware design choices. Also included are suggested fixes. Please comment if you have great ideas!

1. The USB power supply connector that is neither USB or power

Small USB connector for small powerBy design the RPi B2 is supposed to be fed all of its power through a Micro-B type USB connector. This type of connector implies that it respects the USB 2.0 standard for power (since it is a USB 2.0 type connector and all). This standard implies a maximum power of 2,5W per port (or 500mA current @ 5V). The RPi B2 actually draws way more than that in a lot of situations and it should even be able to draw way more (see point 2 below).  This has quite some confusing consequences for its users: 

  • Most standard USB 2.0 compatible wall adapters (such as for charging your smartphone) won't work, since they can't supply enough  power to the RPi B2 at a stable enough voltage. This will cause the RPi B2 to crash. 
  • Most standard (cheap) USB 2.0 cables can be problematic as well as the Micro B connector on the Pi itself; again these are designed to provide a stable 5V @ 500mA maximum. Not more than that. The consequence is that in some cases the voltage may drop when drawing over 500mA adding to unstable behavior
  • To make things even more unpredictable and unstable the RPi input power is limited by a poly fuse at it's input port. This polyfuse will also cause the voltage to drop as more power is drawn (anything over the 700mA hold current of the fuse). When the voltage drops enough the board will crash. 

2. The USB 2.0 ports that aren't actually USB 2.0 ports

RPi USB ports which aren't USB ports
The USB 2.0 client ports on the Pi should allow a 500mA current draw @ 5V per port. Well... guess what? They don't. Not even close. 

Besides the fact that the main power supply circuit of the Pi can't deliver enough power to power all USB client ports, each of the client ports are also current-limited by polyfuses. This means these ports are not actually USB 2.0 ports as defined by the standard, as implied by the looks of the connectors and as implied by the RPi hardware wiki. 

Again this is very confusing for users and will make the RPi very unstable for a lot of folks. The RPi will (seem to) crash randomly depending on: what USB devices you plug in, how many of them you plug in, for how long you plug them in, what the USB devices are doing, length of cables, quality of cables, what the Ethernet connection is doing and on whether the Pi is processing a lot. In other words: !!@#$!#$ wat? 
A workaround to this would be to use an externally powered USB 2.0 hub... But... Come on, wasn't the RPi supposed to be cool for it's small size, low cost and easy of use? Oh and remember, only use hubs that are on the "Raspberry Pi tested & working USB hubs"... hmmm...

3. The Micro SD card reader that maybe reads SD cards... or destroys them...  

One would assume that if an SD memory card is put into an SD card reader this would actually work. Not with the Raspberry Pi... Before using or buying an SD-card for the Pi, be sure to first check the tested & supported SD-card list

But even if you do use a correct card you will eventually get into trouble when running the Pi for extended amounts of time (minutes to months). Here's why: 
  • If you remove power from the RPi or remove the SD card while it is being written to, there's a big chance the OS-image will become corrupted and will fail to boot. See point 1 and 2: this will happen (a lot) if experimenting with power-draw, (USB-)devices, GPIO and/or *cables everywhere*. 
  • The SD card is flash memory. Flash memory has a limited amount of write cycles per memory cell before it will fail. The RPi has no correction mechanisms for this. If you leave the RPi on long enough and actually use the SD card to write to during it's operation, it will fail eventually. All by itself. Withing months, according to the internet. There are workarounds though to limit the amount of "unnesesary" writes to the SD card and extend the life of your SD card a bit. Does that make RPi a really stable platform? Not really in my opinion... 

Other smaller problems

The above are some of the main problems with the RPi B2 (and other RPi's too). These are different problems however in that they don't depend on the hardware "design fails" as much, but are more general misfortunes: 

Fixing the Raspberry Pi B2 hardware design

All of the problems described above could be fixed in future hardware designs. Better yet: they can be fixed without increasing the price of the RPi. Examples of fixes: 
  1. Add pads (or a power-happy connector) for an external power supply + use a poly fuse with a much higher hold current at the input power circuit (or use multiple smaller ones in parallel) for the entire board. Specify a slightly higher input voltage than 5V to compensate for the poly fuse voltage drop(s). Place the voltage regulator circuit behind the poly fuse. Of course traces would need to be adjusted to support the higher currents. 
  2. Replace the USB port poly fuses by poly fuses with much higher hold currents and/or place some of them in parallel. 
  3. Use a type of non-error prone system memory, such as eMMC. And/or... add an SATA interface if price allows. Just... any memory that is not prone to fail after a bit of use.  


Upgrade to Windows 8.1 (You might live a few hours less)

posted Aug 8, 2014, 4:30 AM by Georges Meinders   [ updated Aug 8, 2014, 4:44 AM ]

This is a little story about what happened to me when I obeyed Windows 8.0's suggestion (nagging) to upgrade to Windows 8.1 on my laptop. See it as a warning. 
I thought it would just be another update, but it turned out to be Microsoft's way of robbing me of a few hours of my life for no good reason. 

This is what happened to me and a lot of other innocent people who decided to casually update Windows 8.0 to 8.1: 
  1. Agree to the nagging implying that it's probably a good idea to upgrade. All you need to do is to press a (trap) button. 
  2. Installation takes place. It takes about the time of completely re-installing windows (long).
  3. Windows changed a little. Nothing really noticeable to me, since I use Classic Shell. This tool removes all that is new (and bad) about Windows 8 and recovers all that was good about Windows 7. 
  4. Actually, Windows changed a lot. It became completely useless. Internet stopped working. Completely. After Googling (from my other PC) I found out a lot of people were having similar issues. Some had driver problems, some changed their DNS from auto discovery to a manual (often Google) DNS and some... couldn't figure out what the problem was at all. Like me. People suggested going as far as downgrading back to Windows 8. Which seemed the only option for me too. 
  5. Try to downgrade to Windows 8.0. I had to insert the original installation media again. But Windows didn't want the media on my external drive (guess I had to mount that to make it seem like a DVD). So I inserted the physical DVD that came with my laptop. That was actually the first time I unpacked that DVD.
  6. One hour later Windows was downgraded, internet worked again and all my applications had been deleted. Why? WHYYYYYYYY?! Also, Windows was now in Danish. Now I do love and respect the Danish people (mainly because they build space rockets for fun), but I don't really understand the language.
That's when I decided to type this little blog post and help explain to people what upgrading to Windows 8.1 really means. It means a gamble which results in throwing away a couple of hours of your life and be very frustrated and angry or having... Windows 8.1 which is essentially Windows 8.0 without the nagging to upgrade to 8.1 obviously. Is it worth the gamble to you? 

So you want to get crypto coin mining?

posted Dec 21, 2013, 8:33 AM by Georges Meinders   [ updated Dec 22, 2013, 2:52 PM ]

It sounds too good to be true: start Bitcoin mining and get rich doing nothing. There has a been a lot of fuzz in the media about Bitcoins lately and this may have sparked the desire in you to get crypto mining. This blog is meant as a quick guide to get you up to speed, giving you a realistic view on what crypto currency mining is about. You'll also learn about some common mistakes to avoid which you won't find in the sea of shallow media hyped articles out there.

Should I buy hardware to start Bitcoin mining now? People are getting rich! 

No. Obviously a lot of people have become rich Bitcoin mining or trading, some will say it was luck some will say they knew Bitcoin was going to go up like it did. The recipe for success in Bitcoin mining has been simple (as always in Hindsight): start mining/buying early when nobody believes Bitcoin will have value in the future and wait until Bitcoin becomes popular and expensive. Right now I'd say forget about Bitcoin mining with your own hardware; this has to do with the algorithm (SHA-256) Bitcoin uses that makes it possible to develop Bitcoin mining hardware which doubles the hashing/mining speed every month on average. You may have heard of FPGA and ASIC terminology used to describe fast Bitcoin mining hardware. Because of the way Bitcoin is designed this cheaper/faster mining for everyone results in a steep monthly increase in Bitcoin mining difficulty (see chart to the right); this makes it hard to break-even on the hardware you purchased let alone make a profit. 
It's a rat-race in which I don't think can be won anymore: the first person who has fast hardware makes a lot of money for a while until everybody has fast hardware and new faster hardware comes out. Basically the scenarios which allow you to become rich if you start mining Bitcoin with your own hardware now: 
  1. Bitcoin goes up in price like crazy in the future or 
  2. you order the newest Bitcoin mining specific hardware and you make sure you're one of the first people in the world to use it (close to impossible). 
That's an extremely high-risk investment in my opinion.

Does that mean there's no money to be made in Bitcoin anymore? 

No. You can still make money by trading Bitcoins or buying hashing power from could hashing platforms. However, I wouldn't recommend it personally. The only way to make (a bit) of money is to buy low and hope for an equal or preferably higher Bitcoin price in the future. People who bought hashing power (or outright Bitcoins or other Bitcoin securities) at the high of 1200 USD/BTC will not be making profit now that BTC is at around 600 USD at the time of writing. That being said, as always in trading with patience the BTC price could sky-rocket again. There are people out there saying Bitcoin will go to 40.000 USD/BTC. Unfortunately they do not specify whether that will happen because of hyperinflation of the USD or hyperdeflation of the BTC ;-) 

So, Mr. Megahard what do you recommend as an investment now? 

Scrypt mining! So the main problem in BTC mining is the exponential growth of the difficulty. Lets try to put that in normal English again: with the same hardware you'll be mining 2 times less "value" than whatever you mined in a previous month (assuming Bitcoin doesn't change in value). This is essentially due to a "hashing algorithm" called SHA-256 which Bitcoin uses. However a lot of other crypto currencies exist which use different algorithms. Most notably currently is an algorithm called "Scrypt" which is what is used in the popular Litecoin currency. Mining a currency using this algorithm is therefore often referred to as "Scrypt mining". This algorithm basically makes it very hard for hardware engineers (like me) to develop more efficient mining hardware which beats mass produced PC video cards in mining performance. There are currently no FPGA or ASIC implementations on the market for Scrypt mining which can beat videocards in terms of price, energy usage, hash rates or availability. 

Anoncoin Difficulty Chart
Because in Scrypt mining everybody uses standard mass produced PC hardware, everybody is mining along with similar investment costs per mining speed (expressed as "hash rate"). This results in a difficulty factor for Scrypt mining which has not increased exponentially since the existence of Scrypt currencies. The difficulty factor and value of some Scrypt currencies (like Anoncoin) have remained quite stable, while that of others like the popular Litecoin have increased linearly with popularity and buying power, yet not exponentially like Bitcoin did. This tells you that when Scrypt mining with certain hardware, the amount of buying power you generate remains quite stable over time compared to Bitcoin mining with Bitcoin mining hardware. This makes investing in Scrypt mining hardware a little less risky and less stressful. Less stressful because it doesn't really matter all that much whether you start mining today or in a month from now (which in Bitcoin mining can easily make the difference between profit or loss).
 
For now that is. For now, because companies are reportedly working on FPGA & ASIC Scrypt mining rigs which consume less energy than off-the-shelve video cards. While there are without a doubt people out there who are trying to build and sell FPGA Scrypt mining machines, at the time of writing there are no FPGA scrypt mining machines on the market and the companies working on them have no track record and thus no proof of actually being able to sell FPGA Scrypt mining machines. On top of that scammers are popping up claiming to sell FPGA Scrypt mining machines or cloud hashing. This doesn't help the credibility of FPGA Scrypt mining rigs commercial feasibility and that brings me to warning #1: don't believe people who claim to sell working and efficient FPGA Scrypt mining machines right now! If you're interested in knowing more about the status of Scrypt specific mining hardware, check out the Litecoin forums for an example of some guys working on ASIC hardware, which they'll initially sell hashing power off of. While they seem credible; keep your guard up, it's a weird business known to attract a lot of scammers. 
        The way things look right now (December 2013) personal mining rigs can actually be quite a good investment with a break-even time of around 4 to 8 months (depending on your investment and the energy prices where you live). I make this statement with a few important assumptions which are the following:
  1. The buying power/value of the currency you'll mine needs to remain stable or needs to go up in the future. If the currency you mine goes up in the future the break-even point will come much more quickly. 
  2. Another assumption is that companies will need at least 4 to 8 months more before they release FPGA/ASIC fast Scrypt mining rigs. This is because the increase in your buying power with a static hash rate needs to remain stable over time. The buying power increase of whatever you mine will decline if other people can quickly get more hash/sec for less money (which is again what happened to Bitcoin mining).
As with every investment: there's a risk of making a loss. You won't loose the computer you invested in, but you could still end up with a loss caused by the energy bill necessary for mining.

Don't take my word for it, calculate your own profit!

If you want to know what you'll be earning and spending on electricity bills while mining, some nifty calculators exist on the web. They take the current real-world mining parameters such as difficulty, block reward, energy prices and your hash rate (hardware performance) as inputs and give you net income as output. Keep in mind that these parameters DO NOT remain stable over time and are not predictable: value of coins change, energy prices change and mining difficulty factors change over time. You'll have to make a prediction of how you think these parameters will change in the future and decide to start mining or not based on that.

Interesting websites to check out when calculating your future riches:
  • http://coinpolice.com/coin-comparison/ --> A website which will show you which crypto currency is currently the most attractive to mine. This will allow you to hop from currency to currency to get the most profit possible. Again, keep in mind that a currency which is profitable at a certain moment can become worthless quickly (or skyrocket like crazy!). 
  • http://www.coinwarz.com/cryptocurrency --> Profit calculators for various crypto currencies, historical difficulty charts and currency exchange rate charts.
  • https://bitcointalk.org --> Forums where you can find a lot of practical info. Read on to get an idea of what practical info I'm talking about, this is where I got a lot of input for this blogpost. 
  • https://www.cryptsy.com/ --> Check out historical exchange rates between A LOT of crypto currencies. Some people chose to mine in a very unknown (easy to mine) currency in hopes of the currency gaining popularity and thus value relative to other currencies. This can be a fun, yet risky game to play! 
  • https://litecoin.info/Mining_hardware_comparison --> Scrypt mining hardware comparison. This will give you an idea for the hardware you need and the hash rate you'll get from that hardware. IMPORTANT: a lot of values there are maximum values and some may even be invented by mining nerds to "look cool". If you just go out and buy the cheapest card listed there without paying attention to the brand you'll most probably come out very disappointed! Video card brand, hardware revision and even bios version are extremely important factors in mining. Please read on for more info on this. 
  • https://localbitcoins.com/ --> If you're not interested in mining but would like to trade crypto currencies, I found the easiest way to get into the game is this website. It allows you to buy Bitcoins in fiat cash (EUR/USD/etc) or by wire transfer which you can easily trade for other currencies on cryptsy.

Sounds good! I want to get my own Scrypt mining rig RIGHT NOW! How do I get started? 

The short answer: get yourself a mining rig, plug it in and GET MONEY... and... don't forget to read the long answer first. As always the devil is in the details and getting the right rig is the devilish detail in this case. Sorry.

Picking a video card for Scrypt mining

For this bit of blogpost I'm assuming you're capable of building your own PC. If not, you may want to look into Webshops which specialize in selling Scrypt mining rigs. 
        The most important part of a Scrypt coin mining rig is the video card. The rest of the PC doesn't really matter: you can buy cheap on all other parts, just make sure your power supply is large enough to power all the video cards you're planning on plugging in.
        You can get an idea of hash rate performance of video cards at litecoin.info (or Google your currency of preference + "mining hardware"). For some reason AMD Radeon cards perform way better than any Nvidia cards, so forget about mining with an Nvidia card unless of course you already own one anyway.  A popular model used in mining right now is the: Radeon HD 7950. 
        DO NOT just buy the cheapest most available AMD Radeon video card you come across: not all brands of a certain model are suitable for Scrypt mining. No video card currently on the market has been specifically designed towards Scrypt mining, this means that you'll have to select a video card which "accidentally" happens to have been designed right for the Scrypt mining job. Use forums like bitcointalk to check whether the video card you're thinking off has any problems. Another good way to find out whether a certain brand is any good is googling for something like: "Radeon HD xxxx *brandname* bad hash rates scrypt mining". The consequences of picking a bad video card brand are as follows: very high videocard temperatures causing your PC to freeze or restart itself when mining, very high energy usage per hash and horrible hashing rates. If you're a nerd like me additional disadvantage of a "bad card" would be that there is no way to "tweak" your card to get higher hashing rates, lower temperatures and/or lower energy usage. In other words: choosing a bad brand makes the difference between a potentially viable investment or a useless horrible investment for mining AKA a loss. 

Video card brands to absolutely avoid at all cost are: XFX, Powercolor and 3D club. In any case; always check thoroughly whether the card you want to buy is suitable for mining. You'll no doubt notice that the videocards suitable for Scrypt mining are also the ones that are often sold out. This is due to the high demand in those cards for Scrypt mining

Get mining

Once you built your PC you can start mining in your OS of choice. There are guides out there that explain how to mine under Linux if you're into that. Mining software can be found in the hardware overviews
Hint 1: be aware that the popular CGMiner software only supports Scrypt mining up to version 3.7.2 (the newer versions dropped support and will do "nothing" when you try to Scrypt mine). 

When you've got all your hardware and software ready to go, you'll want to register at a mining pool of your choice (just Google one and sign up). You'll create a "worker" on the mining pool website of choice. You'll need the worker name + password as parameters to the mining software, the pool website should have a little guide explaining you how to input these parameters into the mining software. 

Hint 2: either use mining software with a graphical user interface like GUIminer in combination with CGminer or create a *.bat file where you store all the parameters you input to the program. This will make starting and restarting the software with all your parameters of choice much easier than inputting everything by hand every time you start the miner software. 

Serious mining rigs (more than 2 video cards)

SCRYPT MINING RIG 

As you get into Scrypt mining you'll notice people are going quite crazy building huge rigs to try and make a living off of mining alone (or trying strike it rich).  3-card rigs are pretty common, but how do you actually fit 3 cards in a standard PC-case? Well... you can't really; standard ATX cases don't have enough room to plug 3 or more of those large video cards into them. To do so, you'll need something called PCI-E risers. 

PCE-E risers

The risers basically come in 2 types: 1X PCI-E to 16X PCI-E (powered and unpowered) as well as  16X PCI-E to 16X PCI-E. Scrypt mining is not data intensive, which is why the 1X PCI-E will work just fine. Keep in mind that there can be video card detection issues when using the 1x PCI-E riser cables. Check out this website for more info on this. The easiest way is basically to go for 16x to 16x risers. 
        Don't be surprised if these risers are hard to get right now. Like with the availability of video cards this is due to the popularity of Scrypt mining, I myself can actually not seem to get my hands on them where I'm currently based (The Netherlands).

Conclusion

Some call it a pyramid scheme, some say it's the currency of the future, I say: get informed and decide for yourself. Personally I feel Scrypt-based currencies are currently a nice way of at least (probably) breaking even on a high-end nice gaming-PC, with a good chance of making profit or even getting rich like the early Bitcoin buyers and miners did. At the very least it's an exciting gamble! And if you decide to make your living out of it, that would be an extremely exciting (scary?) gamble.

I say, happy mining, happy Christmas and a happy new year! And contact me with any comments or questions you may have!


huge GPU mining rig
And here's another example of a crazy mining rig (source: ecoinomist.com)

First post

posted Aug 14, 2013, 8:52 PM by Georges Meinders   [ updated Aug 14, 2013, 8:53 PM ]

 My best Mark Zuckerberg impression
 Megahard will be posting content on hardware design engineering related subjects.

Expect an engineer's general brain farts, reviews of attended events, neat technology news, interesting science news, CAD tutorials and cool workplace videos. I hope this will eventually spark some nice discussions on here!

What you see on the left is me (the writer).


1-4 of 4