WirelessMoves – Page 150 – My Thoughts on the Evolution of Wireless, FOSS and the Past, Present and Future of Computing

Learning How Computers Work – Comparing Approaches of the 1980s

In a previous post I've been musing about how in the 1980s, microcomputer learning kits appeared on the market to teach kids and adults how computers work. Note that this is different from learning how to work with computers or even how to program them in a higher level language as in both cases the inner workings of a computer is abstracted. As I'm fascinated by the approach I've also wondered in the post if such kits are still available today or if learning about how computers work has effectively moved on to only be a topic at university courses. Here's what I've found out so far:

Even after some more research I still came up empty handed and it seems such learning and tinkering kits have disappeared today. What I have found, however, is a number of 1980s microcomputer learning kits each with a slightly different approach to the topic. As I am not only interested in the historical perspective but also wonder how this approach could be revived with current hardware and perhaps some self-written software I had a closer look at the tutorials which can be found in PDF format on websites of vintage computing enthusiasts to see how the different kits went about teaching computing basics. Interestingly enough, the approaches are quite different:

The Busch 2090 4-bit Microcomputer Kit

Released in 1981, the Busch 2090 is the oldest of the three microcomputer learning kits I could find. It's based on a 4-bit microcomputer. As the other two, the kit is part of a larger electronics experiment program so it works well together with other electronics kits of the company to expand experiments into the physical world by triggering outputs to LEDs and loudspeakers and receiving input from different kind of self assembled circuits. I've blogged about it in a previous post and you can find some more details see here. (Sorry for this and many other links in this post being in German but it seems these kits were mainly popular in the German speaking world. Google Translate is a helpful tool to translate the pages into your language on the fly.)

Programs are written in a 4-bit pseudo machine code. Each pseudo instruction is always 3 hexadecimal digits long and all operations are explained in the decimal, binary and hexadecimal system. While most pseudo instructions closely resemble actual machine instructions some also perform more complex tasks such as displaying the contents of the accumulator or other registers on the display or reading key input. What I liked best about the teaching approach of the tutorial is the jovial style and a mascot that demonstrates important principles in a funny and easy understandable way as shown in the first image on the left

Another great approach of the tutorial is that almost all program listings contain the memory address for each instruction, the instruction code, the mnemonic for it and a description for each operation as shown in the third image on the left. This makes understanding and memorizing the codes and mnemonics very easy.

I find that the writing style and level of detail makes this kit appealing for kids from 12 to 99. The descriptions of how a microprocessor works and how it communicates with memory and peripherals are very simple but still precise. The tutorial has around 80 pages, so it's not a daunting task to go through it even with the attention span of a 12 year old and I found it to be a perfect balance between beeing too simple and too complex for people that have never touched the subject before.

The Philips 6400 Microcomputer Master Lab

Two years later, in 1983, Philips released it's microcomputer learning kit, the Philips 6400 microcomputer masterlab. While the general approach is the same as in the Busch 2090 there are some significant differences in the practical implementation. Philips used an 8-bit processor for its kit and instead of pseudo machine code instructions, real machine code was used for user programs. I very much like this approach as this is as close to the bare metal as possible. It makes programming a bit more complicated but from the tutorial I take it that the kit was addressed at a somewhat older audience than the Busch computer, perhaps 16 years and up, so this is not much of an issue. Less humor and more facts I would say.

The level of detail is much deeper compared to the Busch kit. At 160 pages the tutorial is twice the size and technical descriptions go much further and deeper. There are good technical descriptions of how two numbers are added in the hexadecimal system, how this is done in the binary system and finally of how this is implemented in hardware with Boolean logic. How subtraction is done by adding the two's complement of a number to another is also introduced as well as multiplication and division operations. The kit also introduces the reader reader to function calls, the stack and stack pointer, discusses Boolean logic (AND, OR, NAND, NOR, XOR), bit masking, jump conditions and quite a number of other bare metal operations. For somebody in the final years of high school who has a knack for mathematics, this level of detail and complexity is probably just right.

My only gripe with the tutorial is that at the beginning program listings are not explained at all and the user has to type in numbers without really knowing what they do or what they stand for. This gets better as more details are introduced as shown in the image on the left but I still felt when reading the tutorial that the program listings should have contained more explanations on an instruction to instruction basis for an easier understanding.

The Kosmos CP1

The tutorial I have of this kit is from 1983 so it also seems to have been brought to the market two years after the Busch 2090. Like the Busch computer, the Kosmos CP1 also seems to address a younger audience as the introduction to computing is quite on a high level of abstraction. Like the Busch approach, the kit uses pseudo-machine instructions, that are, however, much more abstracted than the Busch pseudo-instructions that stick much closer to real machine instructions.

Having gone through the other two tutorials before I focused on this one I have to admit that I found the pseudo-machine code of this approach a bit difficult to understand and I wonder why they ventured so far away from the bare metal instructions. Perhaps this was done to get around the hexadecimal system. Yes, there's not a single hex number throughout the whole tutorial, the computer is entirely programed with decimal numbers. Discovering the hexadecimal system later on must be somewhat of a culture shock. The image on the left shows how a program for the CP1 looks like. Each command is well described but the commands themselves are a bit too abstracted from the real world for my taste. The tutorial has 145 pages and is split into two parts. The first 60
pages are dedicated to introducing the commands and how a computer
works and the second part is dedicated to program examples that are
discussed in detail. Quite a number of the examples let the computer interact with the real world via electronic input and output circuits that can be plugged together on an extension board. While the other two kits include such extension boards, some resistors, transistors, LEDs, etc., I am not sure they were included in the CP1 kit or if was assumed that the reader has already bought other electronics experiment kits from Kosmos before.

What All Have in Common And Why Simplicity Trumps The Full Keyboard + Screen Approach

One thing all three kits have in common is a 7-segment display with six digits (Busch 2090 and Kosmos CP1) or eight digits (Philips 6400). As can be seen in the images the kits are completely operated and programed via numerical input pads and a number of special keys for operations such as performing a reset, for selecting and running a program, for entering programming mode, etc.). This might seem a bit archaic from today's point of view but since these kits are there to introduce the bare metal innards of computers I would still do it like that today. It might be too cumbersome to write long programs with only a numerical keypad to type in raw machine instruction codes but the point of these kits is to teach how a computer works on a low level. Therefore, interacting with it on this level is very appropriate and serves the 'bare metal' approach rather than being a restriction.

In addtion, all kits are very hardware centric so experiments are not limited to generating an output on a 7 segment display and receiving input from the numeric keypad. Instead, experiments can make use of digital input and output ports to which electronic circuits that are put together by the user on an extension board can either deliver input or receive an output to trigger lights or generate a sound.

My Preferences For A Modern Microcomputer Learning Kit

If I were to create such a kit again today for kids age 12 and up and also an adult audience, I think I would take an approach between the Busch 2090 and CP1 on the one hand and the Philips approach on the other. I like the simple and fun approach to the topic of the Busch 2090 tutorial but instead of using pseudo machine instructions I would make use of real machine instructions as is done by the Philips kit. That would mean to introduce function calls a bit earlier in the tutorial to do some more complicated things as showing register contents on the 7-segment display or receiving keyboard input but I think that is a compromise that would be worth it. One could even have two tutorials then, one for a younger and one for a somewhat more technically knowledgeable audience. Or one could create an easy to understand tutorial for both audiences and have a second one with more technical details that builds on the first one for those that have been hooked.

While I think that the audience for such kits was small back in the 1980's and might perhaps be even smaller today it would still be fun to work on this for pure self enjoyment and for giving something to those who'd like to find out how a computer really works. Too bad the day only has 24 hours…

I Just Noticed We Had 40 MHz LTE Channels in Europe, Counting the American Way

So far, pretty much everyone in the industry measured channel bandwidths in what either the downlink or the uplink channel provides. A UMTS channel thus always had a bandwidth of 5 MHz despite twice the spectrum being used, 5 MHz for the downlink and another 5 MHz for the uplink. Sometimes this was also described as 2 x 5 MHz. But it seems some in the US are now adding uplink and downlink together as for example in this Gigaom article. A small 5 MHz LTE carrier now has a bandwidth of 10 MHz. Sounds nice when comparing it to other US network operators that in their words also use 10 MHz LTE carriers, which are, of course 2 x 10 MHz. So by those standards network operators in Europe using the 1800 or 2600 MHz bands already have 40 MHz LTE deployed! That sounds nice! After 4G, Real 4G, True 4G and LTE-Advanced it's the latest smoke bomb to confuse the world and make yourself look better than you really are. Sigh…

Raising the Shields – Part 7: Auto-Delete Cookies When The Browser Closes

Most users today are very happy that web services recognize them when they come back to use a service over many weeks or even months as they don't have to identify themselves each time they visit a site. While this is undoubtedly convenient it creates a number of severe privacy issues:

On sites like my favorite online shopping portal I like to browse for things anonymously. If I keep being recognized then the shopping portal can record all searches and results I have clicked on. To me that feels like I if I was observed by a dozen cameras in a store and the store then analyzes the recordings and keeps the results indefinitely. No thanks.

Except for Safari and perhaps Firefox in the future, browsers allow the use of third-party cookies on web pages. This way, advertisers can track a user's path through the Internet because each time a web site is visited that links to some content of the advertiser, the same cookie is sent back thus creating a trail of where the user went for the advertising company.

The third-party cookie mechanism also allows popular social networking services to keep track of where their users go when they leave their website.

All these things are totally unacceptable to me. Fortunately there's an easy fix for this. In the web browser a few simple settings protect users from such schemes:

Disallow 3rd party cookies ("Accept 3rd party cookies = never"). It's done by the Safari browser by default and I've used it for many months now without any bad side effects.

Configure the browser to delete all cookies when it is closed ("Keep until: I close Firefox"). This way, Amazon and other web services I use do not recognize me again once I restart the browser. The downside is of course that I have to log-in again for personalized services. But with the browser's autocomplete username and password feature it is only a minor inconvenience.

On a few websites such as my blogging service I would like to remain logged in despite browser restarts. For this, there's a cookie whitelist in Firefox (click on the "Exception" button in the Privacy Tab in Preferences). Only cookies on the whitelist survive browser restarts if the option "Keep Until" is set as described in the previous bullet point.

The images below show the settings I have made in Firefox to ensure as much privacy for myself as possible for normal web surfing. For advanced privacy needs I then use a TORified browser as described in a previous post.

This has already been part 7 of my "Raising the Shields" series and it will probably not be the last. In case you have missed some of them and are interested, here's a link that shows them all in sequential order.

Bluetooth Revival With PC Connectivity

Over the years my use of Bluetooth for various purposes has significantly diminished. With the advent of Wi-Fi tethering that I wished for back in 2006 when the first phones with Wi-Fi connectivity were pioneered by Nokia (see here and here) and a few years later realized on the Android platform, my Bluetooth use became occasional at best. I tried Bluetooth headsets every now and then but was always disappointed with their range and I just don't listen often enough to music on my smartphone to have made it worthwhile to buy a Bluetooth enabled stereo headset. Also for transferring pictures or files from mobile devices to my PC or vice versa, Bluetooth was mostly not an option because I always needed a dongle on the PC side. The time this takes was usually the same as removing the SD card from the smartphone and inserting it into the PC or plugging in a cable to access the pictures from the simulated SD card provided by the generic SD card driver that Android provides. My new PC, however, now provides for the first time a built-in Bluetooth transmitter and due to no longer having to plug-in anything, I have found myself using Bluetooth again for quickly moving files and images back and forth if there is only one or two files or images involved in the transfer. A late renaissance of the technology as Wi-Fi direct still doesn't seem to take hold (see here, here and here).

Probing Layer 1 – Part 5: DVB-T Signals

Venturing down a bit on the frequency scale I've now also taken a closer look at DVB-T television signals with my DVB-T USB receiver stick and SDR-Sharp. The image on the left shows the right edge of an 8 MHz DVB-T signal that encodes 4 television stations. Like in the LTE signals I wrote about in a previous post, there are also vertical stripes that can be seen in the waterfall diagram. The stripes are much narrower, however, likely due to the 1 or 4 kHz carriers used in the C-OFDM modulation of DVB-T compared to the 15 kHz OFDM carriers used in the LTE downlink. Interesting to note is that parts of the channel on the frequency axis are broadcast with more power than others. Not sure why, my understanding of DVB-T signals is very limited.

I was also quite baffled with how little of the spectrum assigned to terrestrial television is actually used today. In Germany, DVB-T can be broadcast in the 177.5 – 226.5 MHz range (i.e. a total bandwidth of 49 MHz) and between 474 and 786 MHz (i.e. a bandwidth of 312 MHz). That's 361 MHz in total or enough for 45 DVB-T channels. Despite DVB-T being a single frequency technology in which neighboring transmitters can use the same channel, not all can be used simultaneously as different TV stations are broadcast in different parts of the country. But even if only every second channel is used, that spectrum could still hold an impressive 90 TV channels. In practice much less is broadcast today and when scanning through the spectrum in Cologne, most channels were empty.

360 MHz is quite a sizable chunk of spectrum and as the popularity of DVB-T and terrestrial TV broadcasting is on the decline I can see why there are moves to re-assign the 694-790 MHz range for wireless Internet connectivity, i.e. for use with LTE and perhaps other wireless technologies in the future. The Wikipedia article linked above indicates that the additional band could be brought into operation by 2026. This would give network operators access to an additional 96 MHz of spectrum that, with a duplex gap of 11 MHz, would offer 42 MHz for downlink and 42 MHz for uplink data transmission. That's a bit more than the current digital dividend spectrum in the 800 MHz band bundled in LTE band 20. Here, 30 MHz of bandwidth is available in each direction and used by three network operators in Germany today.

The New Nexus 7 Tablet (2013) Supports AT&T, Verizon and T-Mobile LTE In One Device

While in Europe, GSM, UMTS and LTE are used by all network operators, the US wireless landscape has always been much more diverse. This meant and still means that there always had and have to be several device variants to support different networks. But with the advent of LTE and advances in chip technology this may be about to change.

When Google recently introduced the 2013 version of the Nexus 7 tablet, it upgraded the cellular hardware to support 7 LTE bands. For details see Google's page on the Nexus 7 and AnandTech's mini review over here which has a somewhat different frequency listing. Apart from the stunning number of supported LTE bands it also supports the five major UMTS frequency bands.

The LTE band combination for the US is especially interesting, as band 13 is included for Verizon's LTE, band 17 for AT&T's LTE and band 4 for T-Mobile's LTE. This might very well become the future trend and will finally allow US consumers the same flexibility as in Europe to buy a device independent of the network operator or even change network operators over the life cycle of the device.

I'm not quite sure how Sprint fits into this equation!? From what I can tell they have LTE up and running in a 1900 MHz band and have taken over Clearwire's 2500 MHz TDD assets. While the 2500 MHz TDD band is not supported by the device I have no information which band Sprint uses with it's 1900 MHz assets. If you have some more information on that, please leave a comment.

Also interesting is the absence of CDMA support in the tablet. Probably not surprising for a tablet, as mobile telephony, if implemented on the user interface at all, is not a prime use case for a tablet. Also, unlike UMTS with it's great data rates, CDMA EvDo only offers limited speeds which are undesirable in a data heavy product as well. So why bother?

40% of UMTS Band 1 is Unused

While in the US, network operators were perhaps struggling with the amount of spectrum they had for their 3G services and thus rushed to jump onto the LTE bandwagon, Europe continues to enjoy very good data rates over 3G UMTS to this day (for details have a look here for example) in addition to the massive additional capacity now available on other frequency bands with LTE. So I was wondering a bit how much for the UMTS 2100 MHz band 1 spectrum is actually used today.

Cologne is one of the bigger cities in Germany so it is fair to assume that it is also a place in which the highest number of UMTS carriers are needed to satisfy demand. In total, band 1 can host 12 individual UMTS 5 MHz carriers. In practice, however, SDR-Sharp and my DVB-T stick show quite clearly that only 7 of those are used today. In other words, 40% of the bandwidth available for UMTS in this band are still unused.

It's interesting to also look at how many 5 MHz slots each of the four network operators has in Germany in the prime UMTS band. The distribution is as follows:

Operator 1: 2 channels
Operator 2: 3 channels
Operator 3: 3 channels
Operator 4: 4 channels

Operator 1 has both channels on air and thus, LTE in other bands is the only way to increase available capacity.

Operator 2 and 3 also have two channels on air and in addition have deployed 10 MHz in band 20 for LTE. If necessary, they could still extend their UMTS capacity with one extra channel.

Operator 4 is only using one of its four channels so far! That's in line with that operator always trailing all other operators in speed tests by quite a bit. As that operator does not have spectrum for LTE in the 800 MHz band I would not be surprised if they started with LTE in the 2100 MHz band with a 10 or 15 MHz carrier.

Probing Layer 1 – Part 4: LTE Up- and Downlink Observed

After having taken a look at GSM and UMTS signals on the physical layer, this post is about how LTE signals are visualized by SDR-Sharp.

In Germany we are in the comfortable position to have three network operators offering LTE services in the 800 and 1800 MHz bands so it is not difficult to find an LTE carrier signal. Like for UMTS, it's not possible to show the full 10 MHz (800 MHz band) or 20 MHz (1800 MHz band) LTE channel as the hardware and software are only capable of showing around 2 MHz at a time. But one can observe the edges of the signal or any particular part in between. The first picture on the left shows part of an LTE signal on the 1800 MHz band. All LTE signals I have observed feature the vertical stripes. While the stripes make sense due to the OFDM modulation using many individual 15 kHz carriers that for a 10 or 20 MHz channel I am not quite sure why there are strips that are marked in yellow while other parts are blue (i.e. a lower signal energy). I first thought that perhaps this might have something to do with the reference signals but they are evenly distributed through the carrier and are not only present in particular spots!? Another particularly interesting thing in the image is the reddish bars in parts which is actual data transmission occurring while I downloaded a web page. In other words, it's pretty easy to see on layer 1 how loaded a cell is.

The second image on the left shows an LTE uplink transmission from my UE. Again, keep in mind that the image only shows a fraction of the total bandwidth used for uplink transmissions. The transmission starts in the lower part of the waterfall diagram and was again recorded while downloading a web page. Once the page was fully loaded one can see very nicely how uplink transmissions become spurious and at some later point, not shown in the image, cease completely. A further detail worth mentioning is the absence of the vertical strips compared to downlink transmissions. This is because in the uplink direction SC-FDMA modulation is used.

Incredible insight gained with a 20 Euro DVB-T receiver and the power of open source SDR-Sharp!

Raising the Shields – Part 6: The Onion Router (TOR) on Android

In the previous post in my series on how to protect your privacy online I've looked at the TOR software which hides your location from prying eyes and encrypts all data right until the TOR exit node. I was delighted how easy it has become in recent years to install and use the package so the next logical step was to see how usable the Android variant of the package has become.

Installation of TOR on Android is straight forwarded. All that is required even on non-rooted devices is to install Orbot for the basic TOR functionality and Orweb, the web browser that uses Orbot. Both are available from app stores or can be installed directly from here and here.

On non-rooted devices only the Orweb browser uses the TOR functionality provided by Orbot so all other applications continue to work as before and are obviously as unprotected as before. This can be both good and bad depending on the usage scenario. If TOR is only to be used for specific privacy sensitive web browsing sessions then this is the best setup as all security measures are automatically taken care of such as Java script being disabled, use of a generic browser ID when contacting web sites to prevent browser fingerprinting, no use of persistent cookies and no Flash. While this is perfect for privacy, such settings are likely to stand in the way for many everyday usage scenarios that are perhaps a bit less privacy sensitive. For these, the standard browser can be used as before as no settings are changed and Internet access is as direct as before.

If Orbot is installed on a rooted Android device it can be used as an Internet proxy just like on the PC and all programs that are capable of standard proxy functionality can use a TOR tunnel. This way, mobile email, instant messaging and other applications can be protected as well.

When starting Orweb, it offers the user to also start the Orbot background app if it hasn't been started before. After a couple of seconds the secure TOR tunnel is established and while using TOR has somewhat of a a speed impact I found it entirely usable. Unfortunately there are some usability restrictions introduced by the privacy model. There's no browsing history in the browser from which a previously visited website can be quickly reached and there are also no bookmarks. While I can understand both measures I think there should really be a bookmark feature even if it is a compromise, perhaps with a warning to the user that using bookmarks could reveal privacy sensitive information when the device is stolen. A solution to the problem could perhaps be bookmarks stored somewhere on the web anonymously and reachable via a short bit.ly link?

To sum up, I found TOR a great privacy protection tool for mobile use. It's easy to install and use but as on the PC, privacy and security means there are some usability limitations. I therefore prefer using TOR like on the PC in a two browser setup, one for my standard web browsing needs and another one for stuff where I would like to guard my privacy.

Probing Layer 1 – Part 3: Further Thoughts on GSM 900 Use in Cities

Two years ago I used a mobile phone that could show me the GSM frequency channels to find out whether all GSM network operators in Germany make use of the GSM 900 MHz band. While the original two GSM network operators obviously made good use of their beachfront property, the other two network operators that came later and at first only had 1800 MHz spectrum made little to no use of the 900 MHz spectrum. With my SDR-Sharp + DVB-T Stick Layer 1 utility I now revisited the topic to have a look at the complete band owned by the original 1800 MHz carriers in the 900 MHz band to see if I had overlooked something or if something had changed in the meantime.

The result is pretty much the same as two years ago. One of the two operators makes a bit of use, I could observe at least one or two carriers in their 5 MHz part of the spectrum in Cologne (i.e. in the city). This is very little compared to the two original GSM network operators that have many many cells on air in the same amount of spectrum.

The second original 1800 MHz operator still doesn't seem to make any use of GSM 900 in Cologne. I could only see very faint 200 kHz signals and I am not quite sure what they are. Perhaps echos of other carriers nearby shown here by my tracing hardware which has its limitations? Or perhaps cells used outside of Cologne that are still visible here?

In any case, both make so little use of it that I wonder if one day one of them might just start with either a UMTS 900 or an LTE 900 carrier in this part of the spectrum.