Defcon:Blog Keklkakl blog blag

30Dec/170

Fake electronic components, a hazard of online shopping

This adventure started with a plan, and a bit of drafted design, to make a homebrew computer based on a Z80 CPU, and related components. But, I didn't actually have a Z80 chip yet! So, being cheap, I got the idea that going to eBay to find the needed components would be a great idea. Boy was I wrong.

After looking over my ideas, I searched eBay, and came up with a list of orders for central parts: A pair of Z80C0020 CPUs for $3.35, a Z84C4208 SIO/2 for $7.90, and five HM628128LP RAM chips for $6.0. All items from sellers providing free international shipping! The money left my account in december of '15.

As with all orders that include free international shipping, delivery time was a few weeks, so the project was put on a back-burner. When the parts arrived in early february, I was already well into other projects. So I gave them a quick look, everything seemed in order, and I packed them in my parts bin to wait. For nearly a full year.

When winter came around in 2016 and Christmas was approaching, it was time to start a winter-project. The Z80 homebrew project was resurrected, the components were dug out from the parts bin, and the first test-circuit was wired up on a set of breadboards purchased at bottom-price sometime during summer. A test-setup for this CPU is extremely simple. In the Z80/8080 instruction sets, the opcode 0x00, 0b00000000 is a NOP, so simply making all data-bus pins be tied to ground makes the CPU NOP its way through all 65536 addresses! So: behold the test-setup..

Z80 CPU Test circuit

Z80 CPU Test circuit

I checked, and checked again that I had the pinouts and connections right, made sure my bench power-supply was set to 4.99Volts, and examined my documentation to see that I should expect about 5-30mA current, LEDs not included. Next I connected my VCC from the power-supply to the setup, and immediately ripped the wired back off. The powersupply screamed 800mA at me! What could be causing my simple test circuit to draw 10-20 times as much current than I expected?

My first suspect was of course my wiring. So I checked that again. Pin-for-pin it was correct. The second suspect was the cheap breadboards, and the doublesided tape that stuck them to a wooden board. To eliminate those, I moved the circuit to a "known good" set of breadboards, part of an electronics trainer. After again checking the wiring, power was applied, and again the circuit pulled 6-800mA, and none of the LEDs lit up.

Next suspect was the Z80 chip itself. It could easily have been zapped by static during shipping, so I was happy I had ordered two, and slapped the second one on. This time, the power-supply reported a draw of 10mA, with a dim glow from one of the LEDs. Weird. I pondered for a moment, and just a moment long enough... Suddenly the current consumption rose to some high value I can't remember, and the famous, magical blue smoke appeared out of the Z80. Two for two...

As I now didn't have a CPU to use in the project, I decided to look up what the Z80 would cost from the reputable sources. After seeing that Mouser, DigiKey and RS Online all had the Z84C0010PEG in stock, I selected to order two from DigiKey for a total of $12, plus just about $18 for shipping. So, close to 10 times more than I paid for the two duds from eBay. But unlike the first two that took around 25 days to arrive and were non-functional, DigiKey had the parts in my mailbox some 50 hours after ordering, transported from the UK to Norway. And when I inserted one of them in the test circuit, it worked flawlessly, and pulled just under 28mA.

Z84C0010PEG, genuine

With a sneaking suspicion that less than $10 for the SIO/2 delivered might be a very optimistic price, I decided to test that one with only the supply voltage applied to the VCC/GND pins. I was most certainly correct. Just under an amp was being pulled, and the chip got very hot indeed (but did not release any magical smoke). Was I surprised? Well, the first time, with the CPU I was, even if the price had been very low indeed. But when the SIO/2 turned out to be a dud as well, I was not surprised.

The cheapskate in me had received a lesson: Always buy active semiconductor components from a proper, reputable source!

There's two types of semiconductor fakery going on: Clones of actual components, and outright fake duds. In the first group, you have things like the FTDI232/PL2303 fakes, where functional (but possibly crippled) semiconductor copies or clones are being manufactured and sold, as being the real deal. Some of these clones actually make their way into the mainstream supply chain of the official parts, sometimes because the original manufacturer "moonlights" poor-quality cheaper editions of the original.

The second type relies on online trade, cheapskates like me, or naive buyers. In this group, a seller looking to make a quick buck targets one or more reasonably popular but usually somewhat obscure components. They gets a hold of a bunch of very cheap components in the same package (like the DIP40 of a Z80). The cheap component will typically be a random IC that sells very cheaply, factory defects that can be picked from scrap heaps, or even old parts pulled from "recycling". With the package, they then clear away any markings on the package, and silk-print the markings of their "target". This convincing lookalike is then put up for sale, usually with a price placed around 1/2 of the "per one" price of sellers like DigiKey or Mouser.

Generic picture of Fake Z80

So, how does one spot the fake? Most of the time, simply looking at the price including shipping should be enough. If a component costs $6 plus shipping from Mouser, and sells for $2 including free international shipping, just walk away. It's going to be a fake. Also, if a seller is using the same stock, but not manufacturer, photo of the component as every other seller with the same device at the same price you're probably in muddy waters. Unfortunately it's not always that simple. Often times, the price is probable, and the photo looks convincing. And when you get the component in hand, the relabeling has been so well done you don't spot it. It really is a buyer beware situation.

With all these fake components going for cheap on eBay, does that mean that it is never OK to buy semiconductors there? I think it sometimes is OK. If the part has a realistic price, a plausible description, maybe even a unique photo taken by the seller, the probability of a good part goes up. As long as the seller is not located in China. Look for European or American sellers, Japan is typically also a "good source". In my experience, buying things from China is usually a decent experience EXCEPT for semiconductor parts. I've once received an IC, transistor or FET from China that was what I ordered...

Considering that a "realistic price" is typically the same as the price for a new component (or even higher) bought from DigiKey or Mouser, when would these "rules" for eBay-ing semiconductors apply? The answer to this is very simple: Non-active, obsolete parts. Sometimes you need an old part that's no longer in production. Trying to find a "good" seller of these is sometimes impossible, and if possible they are usually priced waaay expensive. So the best option for getting a decent price is going to eBay.

One example of mine of this is the AY-3-8910/YM2149. I have wanted to experiment with one of the "olde-time" 8-bit audio-chips, and looking over the various soundchip options I decided I wanted to try out the General Instruments AY-3-8910 or one of it's variants. Unfortunately the all of the AY-3-89xx chips, in GI or Microchip name, as well as the YM2149F have been out of production for several years. The only option of getting a hold of one is to get NOS or Pulled parts. I located a not-eBay seller (unfortunately I've lost the link) of genuine NOS AY-3-8912 chips. Unfortunately, paying $50 per chip was not something I felt like doing for a bit of experimentation. My limit was at $25. Going to eBay instead I located a chap in Germany that wanted $5 plus $1 shipping for one YM2149F. I also located a pair of sellers in China, one that wanted $8, free shipping for a Microchip AY-3-8910 and one that wanted $7 plus $2 shipping for an AY-3-8912. I ordered those three. When they arrived, I found out that the 8910 one was clearly a fake, while the 8912 and the 2149 both worked just fine!

There are a LOT of fake semiconductors on eBay and AliExpress. Generally, cheap, free-shipping components sold on these sites from mainland China are guaranteed to be fake. If the component you are looking for is still in active production, buy it from a proper electronics distributor like RS Components, DigiKey or Mouser. If the component you are looking for is obsolete/out of production, take care, research probable prices, and be critical to sellers location.

This post is part of a series documenting my journey in making my own homebrew computer, the posts can be found on this link.

7Dec/170

Scribbling ideas for a homebrew computer design

This is part two of a series, documenting my journey in making my own homebrew computer (previous post here). As the title suggests, it looks at my process of selecting a design. But first, as I am a computer history nerd, a little computer history ...

The world of computing as we know it wasn't in my opinion started by the big behemoths of the early computing age. Yes, they created the field of computing, and through steady capability increase combined with reduced sizes made the systems we have today possible. But it required the invention of the microprocessor to make the idea of personal computing a realistic idea. Before that, it seemed computers was something for big businesses, universities and research institutes. But with the creation of chips, originally designed for flight computers in jetfighters and for calculators at accountant desks, hobbyists started playing with the idea of creating a computer of their own! Early on, the homebrew computer systems was the way to get a computer in a private home. Early examples of commercial personal computers, like the Altair 8800 and Processor Technology Sol-20 were in reality business-backed homebrew systems. In fact, the first of these commercial systems were shipped as blank PCBs and bags of components. Even the famous Apple I computer is an example of this. It is this enabling of computers in the average Joe's hands that enabled the explosion of computing that has lead to the technology enabling we have today.

Sol Terminal Computer / Sol-20

Sol Terminal Computer / Sol-20

So, to design thoughts and ideas. I wanted to create a computer that fits in the era of homebrew. This means to me, a fairly low-end system, designed by an individual or a small group of people, using time-period components, to be built in the home. If my design is able to be produced as a kit, that fits right in with the "standard" of the day. System designs were prototyped in basements and cellars, to later be sold as kits for other people to brew at home, and modify and expand. It was natural for me to select an 8-bit platform as my target, after all the 8-bit "revolution" was at the core of homebrew. As mentioned, I wanted to use time-period components, but I also wanted to be able to create my system without having to resort to "New-old-stock" or cannibalizing of parts to get a hold of parts that are no longer manufactured. This put some hefty restrictions on my options, and also meant I had to allow myself to use a few bits of "modern" technology.

This brings me directly to the choice of a CPU for my system. I talked about how the 6502 was the first CPU that I got to know in-depth in my previous post. Unfortunately (or fortunately, take a pick), I early on decided that the 6502 would not be the CPU for my system. While the CPU is still available, in a CMOS-ified version with added instructions (W65C02S6TPG-14), it would require some supporting components that are no longer available (other than NOS) to make a full computer that I would be happy with. With that realization, I started looking at what other microprocessors were used in significant number of designs. Candidates seemed to be: RCA CDP1802, Motorola 6800, Intel 8088, Intel 8086 and Zilog Z80. Of those, the only one that has an active, still in production version that's not a derivation is the Z80. As a bonus for choosing that one, it comes in a bread- and perf-board friendly 40Pin DIP, and has significantly simpler interfacing than the other options, by having full address- and databuses, simple single-phase clock and a trivial voltage-supply requirement of simply 5Volt. The Z80 was also designed to be relatively compatible with the 8080, so a lot of existing software runs on both CPUs.

Annotated photo of Z80 die, by Ken Shirriff

Landing on the Z80 gave me additional things to explore. Being a retrocomputing-nerd, I have for a long time wanted a system that could run CP/M. Control Program for Microcomputers is a very significant part of Personal Computing History, and a great number of systems, both commercial and homebrew, either had CP/M ported, or were designed specifically for it. With the Altair and IMSAI computers kick-starting a form of standardization with the S100 bus, CP/M became more or less the standard operating system for general personal computers, and thus had a majority share of the system software marked. It was only beat by the multitudes of different ROM BASIC versions summed up. CP/M was designed for select 8-bit Intel 8080 development systems, but due to its design it was easily ported to compatible machines, and with the compatibility of the Z80, quite a lot of systems were designed with just that CPU. While exploring what it would take to let me design a CP/M friendly system, I discovered Grant Searle's brilliant set of designs. On taking in the content of his site, I was first impressed by the amount and quality of work mr Searle has done to make his systems available and understandable. Second, I was blown away by how simple a CP/M system could be. So my path was early on set; I would take inspiration from those designs, and get CP/M running!

With a path staked, my plan became to try implementing the "Fully functional Z80 CP/M machine using only 9 chips" from Grant Searle on a breadboard or perfboard before proceeding to attempt a design of my own, to make sure that my understanding of the complexities involved was sane. However, I allowed myself to start sketching ideas for what I wanted to accomplish with my own design while waiting for parts. The journey of trying to get the first system operational will be documented in separate articles, this one is about the sketching...

My homebrew hybrid prototype of Grant Searle and Spencer Owens's designs

One of my early design choices was between simplicity or expandability. I had looked at the designs in the Retrobrew Computers community, and generally thought they were trying to be somewhat more "modern" than what I was trying to achieve. I wanted something that was a bit simpler, but at the same time I thought that having the option of extending the system would be great. This led to quite a few nights reading up on various bus architectures from the homebrew era, thinking about using one of the "standard" options like S100[1][2] or STD80[1][2]. One big issue with using one of the full-featured old standard bus designs was that it would require a lot more complexity than I wanted. So I decided not to go down that route, and rather go for a more raw CPU signal extension bus.

While examining what I would need to put on such an expansion interface, I became aware of Spencer Owen's RC2014. His first version of the RC2014 was basically Grant Searle's simple Z80 BASIC system, split into multiple cards, one for each major component, connected to each other on a simple 40-pin perfboard-friendly bus-approach. I really liked the concept, but saw a few challenges in it. It used single-row 2.54mm spaced header pins, and omitted a few signals that I knew would be needed later. From experience, I felt that using double-row connectors would lead to more solid mechanical properties, as well as making the bus mechanically smaller. Spencer's bus does not "fit" on a 100mm wide PCB, a real issue as prices for PCBs (and EDA/CAD if not using KiCAD or EasyEDA) has a definite crossing point at exactly 100mm... A double-row physical layout would also very easily translate into a card-edge implementation if one would like a truly retro feel... There will be a (possibly short) article about the result, with pinout and concept released to the public domain. I call my idea/approach the Z50Bus.

With a bus inspired by RC2014, it would also be possible to take advantage of the many cool designs that Scott M Baker has made. Take a look at http://www.smbaker.com/category/electronics-projects/, he has some really sleek additions and extensions based on the fairly simple RC2014. After seeing what smbaker has been able to do, I was sure that I wanted a simple, yet complete CPU bus, and that way be able to reduce the amount of "onboard" features.

Originally, I really really wanted to have a graphics system as a direct part of my design. Quite a lot of time was spent thinking about using memory-mapped or register/io-space based with their pros and cons, debating using a prefab Video Display Processor like the VIC, VDP or Yamaha V-series, and researching indexed video, colour-register based approaches and planar graphics. After seeing FAP, the FPGA Assisted Processor, I came to the conclusion that using an FPGA would be very similar to the time-period correct ULA, PLA and CPLD approaches that were actually used on Home Computers of the 1979-1985 period, just faster and reconfigurable. Going that route should allow a core hardware design to implement many different graphics subsystems, allowing experimentation.

However, I have chosen to not implement the actual graphics system initially! Most homebrew computers of the early microprocessor age didn't actually have graphics capabilities. Most didn't even support a keyboard. Generally the approach in the early days of homebrew computing was to use a teletype terminal, or a video display terminal. So I decided, if a set of serial ports was enough to get the homebrew computing started, a set of serial ports would be enough to get my system started as well! After all, I had already included an expansion interface, so adding graphics and keyboard I/O later on would simply be just that; an expansion. However, to be able to use memory-mapped video later, I would have to design my memory-map and RAM-interfacing with that in mind. The memory map, and the reasoning behind it is topic for a separate article.

Similar to the decision design for graphics as an expansion, I decided that storage media like floppy disk support is to be an expansion. I intend to create a floppy disk interface, for an actual floppy drive and/or a Lotharek HXC. Also on the roadmap is trying to go real retro, with a cassette interface. But I changed my mind about one form of storage connection. Even if IDE appeared on the market late in the 80's, it is too embarrassingly trivial to implement the 8bit microcomputer compatible mode of IDE. With a sane I/O interface like the Z80 has, all that is required to get working 8-bit mode IDE is a simple address selector, and two OR-gates and a few resistors. Plop a Compact Flash card adapter onto that, and you have an interface to solid state storage "for free". The fact that Grant Searle's CP/M computer uses that exact approach meant that is was an obvious choice for me to include it.

Hand-drawn early version of the system design

So, my basic design was centered around a Z80 CPU, with a memory map that would work with CP/M and a future graphics expansion, and serial ports for console and communication. To make it possible to have reconfigurable serial port speeds, and also general purpose timer triggered events I figured I needed to include some form of timer controller as an on-board peripheral. It would also be nice if interfacing with the "outside world" was easy, so adding one or more configurable parallel I/O ports made sense. The more I looked at that spec, the more obvious it became that I was more or less describing the standard Zilog "stack" (minus the DMAC): Z80 CPU as the processor, Z80 SIO for serial ports, Z80 CTC for the timer/counters and the Z80 PIO for two parallel ports. From this, it became natural to also design the system to take full advantage of the somewhat special, but cool Z80 Interrupt Mode 2.

IM2 is designed with a very powerful daisy-chain form of interrupt priority and arbitration. In IM2, devices are daisy-chained using their Interrupt Enable In (IEI) and Interrupt Enable Out (IEO) pins, with priority being resolved by their order in the chain. Additionally, interrupt handling is vectored, with the CPU providing upper 8 bits and the interrupting device providing the lower 8 bits of the relevant Interrupt Service Routine to be executed. The whole IM2 system looked very nice to me, so it is a central design element. It introduces challenges when trying to use non-Zilog peripherals unfortunately, but Luke Maurits has given me some ideas to solve that later from his LM-512 Homebrew computer.

As part of my process of building my "learning computers", in the forms of Grant Searle's CP/M computer and Spencer Owen's RC2014, I figured that following an approach close to those made sense. Using 74LS138 for address decoding of on-board peripherals was a safe and natural approach. To enable a memory map for CP/M plus expansion for graphics, I decided to add a "write-only configuration register" by including a 74LS273, placed at the IO address that Grant's monitor ROM uses to enable/disable onboard ROM. I decided that I did not want the complexity of catering for a large memory space, onboard memory was set to 64kbyte onboard RAM, 16k onboard ROM, with a bit of banking for a max configuration of 176k ( 16k*8+48k) RAM and 64k (16k*4) ROM in an expanded system using memory banking via the configuration register. This configuration would allow use of CP/M, expansion with graphics and storage later, but it restricts the system to not be compatible with Fuzix by Alan Cox et al.

To close off a by now rather lengthy post, a summary of design conclusions is in place

  • Z80 CPU
  • Zilog SIO/2 serial port controller with dual serial ports
  • Zilog CTC for four channels of Counter/Timer, 2 of them dedicated to the SIO/2
  • Zilog PIO for two Parallel ports
  • Expandable system through Z50Bus
  • Memory map designed to allow for graphics expansion
  • CP/M compatibility
  • ROM part of memory from address $0000 disable-able (needed for CP/M and more)
  • For fun, a dedicated ROM Cartridge connector
  • Z80 Interrupt Mode 2 in the core
  • Compatibility with software from Grant Searle (for easy system bring-up)

The last point in the list turned out to be very useful, as it allowed me to tackle a bunch of problems I encountered, without having to also write my own system ROM monitor. It also let me simply load up CP/M onto a CF card and start playing Zork!