Every time the phone was plugged into the computer iTunes would say that there was an update for the carrier settings, but it never applied and the menu entries remained blank. O2 support were only slightly helpful, taking me through the settings which appear on their website but not fixing the Visual Voicemail issues or the prompt in iTunes.

There are threads on the O2 site here and here and a useful one on AVForums which gives the fix.

Unfortunately the fix is to completely reset the iPhone and start from scratch. Restoring from a backup simply brings the old (broken) settings back. I couldn’t find a way to only synchronise contacts back to the phone, though iTunes suggests it is possible.

Follow these steps to repair it:

1. erase the phone from the menu: Settings -> General -> Reset -> Erase All Content and Settings
2. connect the phone to iTunes and restore iOS4
3. set the device up as a new phone, not from a backup!

Browsing through the carrier settings menu should now show the correct values for APN (idata.o2.co.uk). Importantly, the Visual Voicemail settings should also be in the menu.

Now to setup all my apps, account sand contacts from scratch..

If you use it, make sure you use the latest version of libnids (currently v1.24) as it fixes a bunch of important bugs. This currently isn’t in Debian/Ubuntu and the maintainer has disappeared, so compile from source for now. Red Hat based distributions get it easy it seems, as somebody has already compiled it for RHEL4 or later.

Big thanks go to Simon from the company who responded to my queries very quickly.

The best bit? It’s pure Perl, therefore also uses real Perl regular expressions. Yes, there might be grep --perl-regexp, but nobody bothers compiling that in. Plus ack has some other neat features.

See more at the ack website.

Copper layers

Silkscreen

This board uses 0803 SMD parts, but boy are they tiny.  These are some of the larger surface mount components, I definitely would not try the smaller ones.  After applying paste by hand to the board (I used this stuff) I carefully placed the components using a pair of tweezers I bought specially.

Most board houses will apply a solder mask to manufactured PCBs, this is a special layer on top of the board that covers up everything except the pads.  This means a thin line of paste can be run along all the pins of the chip, there is no need to try and put an individual dot on every pad!  As the paste heats up and turns into solder the flux will evaporate and surface tension pulls the solder paste back onto each pad.

The original board, paste and tweezers can be seen in the image below.

SMD board tools

After assembly came the fun bit, cooking time!  I kept an old frying pan specially for this, into which I placed the board.  Cooking time was no more than a few minutes, it is very visible when the paste starts to reflow and turn into molten solder.  As soon as that happened I turned off the gas hob and left it to cool.

Cooking an SMD PCB!

Note that the oven thermometer really doesn’t work, it had barely reached 50°C by the time the board was finished.  A non contact IR thermometer apparently works much better and should be considered if you want to roughly follow a specific reflow profile.

The final result is shown below, after the header pins were hand soldered.

The finished SMD PCB

Did it work first time?  Well, mostly.  The PIC powered up, the LED blinked on and it talked to PICKit2 on the PC.  After a little examination I found that two of the pads on the chip were not properly soldered, which played havoc with some code I was writing.

Next time I think I would use more paste, I was quite frugal with the syringe and had to fix the two bad joints by hand.  I also probably wouldn’t design all future PCBs as SMD, populating and reflowing the board is easy but the parts are fairly expensive compared to standard “through hole” alternatives.

One more picture now of it in use.  This little thing really speeds up prototyping time, no need to connect a pull-up resistor, power and ICSP lines manually.  If only writing code was so easy..

SMD breakout in use with a graphic LCD

For more information, watch this useful video on Youtube and read the excellent SMD How To from Sparkfun.

Many people, including those who charge for wedding pictures, seem to think that an entry level Canon or Nikon digital SLR with an extra lens is all that is required.  Now I’m not equipment bashing, some of them take fantastic pictures.  But that “some” is probably less than 0.5%; it is really sad just how many photos are really just average, overprocessed snapshots.

To explain my point, I’ll use two pictures I have taken very recently.  Here’s the first one:

Example one: A cat

Look at that amazing bokeh in the background!  Look at the way the cute little face is super sharp and the rest of her isn’t!  But really, it’s just a cat.  Unless you know the cat you probably don’t care.  This is the sort of image which becomes incredibly easy with a SLR and 15 minutes practise.  And guess what, that doesn’t make you a professional photographer.

Verdict?  Average.

Here’s the second image:

Example two: A bolt

Look at the awesome colours!  Feel the rustic vibe!  Oooh at the focus once again as your eyes travel around the image looking for.. well.. something interesting.  It’s not art.  It’s not good stock photography.  It’s just a rather average bolt attached to a gate that somebody happened to paint a rather hateful colour.

Verdict?  Incredibly average.

Now before I am criticised for taking the fun out of photography, that isn’t my point at all.  I took the first picture because the cat is owned by two people I know, who cooed over it like proud parents. The cat hated the camera too, which was a bonus.

The second was an opportunistic shot as I wanted to see just how close my new lens would focus in macro mode. Quite close apparently!  I’m rather happy of how it came out as I took it hand-held (albiet with IS), but that’s a relatively geeky photo-obsessive thing to worry about.  Unless you are terrible with a camera, most anything except super macro will look sharp at 800 pixels wide on a computer screen.

However, neither photo required any more skill than setting the camera to Program mode and hitting a button.  Neither has a unique selling point, neither couldn’t be emulated by somebody with a minimum of teaching.  In actual fact, neither of the above images have any adjustments (exposure or otherwise) and that’s something I try to aim for.

I got great enjoyment from going out with my camera that day and I learned a bit too.  But nowhere near enough to justify calling myself a “photographer” or worse still charging for wedding photography and delivering poor results.  Many people only get married once and that’s a darn good reason not to ruin their day.

By all means post your photos online and take criticism, it’s the best way to improve.  Relish the fact that any digital SLR will take pictures that most compact cameras simply can’t, and that will put you above most of the rest of your friends and family.  I love looking back at my early pictures and laughing at all the silly mistakes I used to make, whilst I’m happily oblivious to the ones I am still making.

To wrap up slightly more positively, a link to the flickr page of a friend, Angela.  She has been taking some lovely photos recently which deserve recognition, including a few very cute ones of some kittens.

I needed this for rsync, to let a remote backup server slowly catch up if large amounts of data were added to the live server during the day. A useful post on the rsync mailing list discusses an rsync patch but also the timeout command.

After installing (the Debian package is simply timeout) it is as easy as running with the number of seconds to run for:

It is also possible to specify the signal which will be sent to a program, which is useful if you do not want to simply send SIGKILL. I used SIGHUP in the hope that rsync would have a chance to exit gracefully:

A full list of signals and their numeric values can be found in man 1 kill.

A wrapper script is also available from Johannes Buchner.

By the end of this post there will be a PIC uC talking to Hyperterminal on the PC using less than 50 lines of code.

Pros & cons of using the BoostC library

There main advantage to using this library to implement serial communication is that it sets up all registers and control bits for you (including TRIS for the TX and RX pins). This means limited digging around in the datasheet, you can code without worrying about the underlying hardware.

However it may be a disadvantage too. The limited control over the hardware means that certain things are fixed. For example, 8-bit transmission is the default. It might be possible to change these by modifying the PIC registers directly but I haven’t tried this yet. Leave a comment if you know it works!

Requirements

A few things are necessary to make this work. Firstly a PIC with hardware USART support, in this post I’ll be using the PIC 18F1320 as an example. The hardware USART is listed under “peripherals” in datasheets.

In order to talk to a PC we also need some form of level convertor from logic levels (+5v/0v) to RS-232 (maximum voltage range +12/-12v).  I use a MAX232 breakout board that I built myself, but this board or this board from Sparkfun should work just fine.

Connecting it up

The purpose of this post isn’t to discuss connectivity, but briefly the PIC TX and RX pins need to be connected to the level shifter, as do +5v and GND. The serial pins on the PIC can be found in diagrams in the datasheet. Remember which pins they are as they are needed later.

Find the right values

The BoostC library is incredibly easy to use but there are a three values to work out first. Remember the choices so they can be used in the code later.

bit_time

The first value is bit_time, which is the number of CPU instruction cycles per bit (note that the header file says this is used by the software USART, but code will not compile without it so it is included here for reference).

Take the clock speed in hz, divide by four and then divide by your desired bit rate. I recommend starting slow (e.g. 1200 baud) and trying to faster when it works, so for an 8Mhz clock and 1200 baud we get:

Baud rate generator mode

The second thing to work out is whether we need high or low speed mode and what divisor to use. Fortunately these are fairly simple as they are explained in the datasheet.

If using a slow baud rate with a fast oscillator, the datasheet recommends using the high baud rate generator. The datasheet also suggests that it might help to reduce baud rate error. Enabling this means using a mode of 1. 

For this example I used the 8Mhz internal oscillator (fairly low speed) and did not use the high speed generator, so I set mode to 0.

Baud rate generator divisor

Finding the divisor number is as easy as reading the datasheet to find the “SPBRG value”. There are tables with the combinations of oscillator speed and baud rate showing the correct value. Be careful to use the correct table for whether you plan to work with the high or low speed generator!

For 1200 baud at 8Mhz using the low speed generator, the value is 103.

Using the code

Below is example code which shows how to use the information described above. It was adapted from the file serial_test.c which comes with BoostC.

For some reason the BoostC examples use the hexadecimal locations of the correct registers. It is possible to use their names instead, e.g. TRISB/LATB/PORTB as I have below.

In the code below it is necessary to change 4 things:

• the ports and pin numbers for TX and RX (e.g. if TX is RA6, use 6)
• the bit_time value calculated above
• the call to uart_init, which will need the high/low speed flag and the divisor value
• the clock frequency, if 8Mhz is not used

Code example

Setting up a PC client

As I mentioned above the settings for serial communication are fixed by the library.  You will need to set 1200 bits per second, 8 data bits, no parity, 1 stop bit and Xon/Xoff flow control.  This is often shortened to “1200 8-N-1″.

If you use Hyperterminal, all settings are on the properties page for the relevant serial port.

Conclusion

And finally, some proof that it actually works!  If you find this post useful, or you need further help, please leave a comment below.

Update

Mike (K8LH) has posted a great little utility on the Microchip forums that can help work out the correct values, if you don’t fancy digging through datasheets.

This causes the router to request DNS server addresses when the PPP session is negotiated (when it “dials up” to ADSL).

After reconnecting the PPP session (either reload or shutdown the interface and then bring it back) it is possible to check that DNS servers have been obtained using sh host:

If the router is running a DNS server it is possible to direct DHCP clients at the router IP instead, forcing them to use these same settings.