Monday morning. Arrive at the office with lots to do. Switch on the computer, and…

That’s not what I wanted to see. There’s lots of advice out there on the web about fixing this problem, but it turned out not to be so straightforward.

The first step was to get hold of a bootable Windows 10 installer. That entailed installing a desktop and web browser on my office Debian Linux machine which normally runs withou a monitor. Downloading the image from Microsoft was easy enough, though it’s 4.5GB in size so takes a while. I wrote it to a USB stick (which I had to go and buy, of course) and…nothing. It turns out that the Windows 10 .iso image is only suitable for writing to a DVD, and is not a ‘hybrid’ image which can work also on a USB stick. Of course, I didn’t have any blank DVDs lying around, did I?

Luckily Linux was able to mount the .iso image so I could see the files inside it. That meant it was possible to reformat the USB stick, creating a single partition formatted FAT32, and copy the files on to it.

This ran in to a problem because the main archive of Windows files, /sources/install.wim, is just over 4GB in size and thus can’t be written to a FAT32 filesystem in its entirety. I tried to fix this by formatting the stick in ExFAT or NTFS, but my PC’s motherboard (an Asus Z77-Wifi) can’t boot from those filesystems so that didn’t get anywhere. I suspect the right answer is to use UEFI boot, creating a FAT32 partition with the EFI folder on it, and a separate NTFS partition with the other files on. But I was in too much of a hurry for that and decided to just put up with a slightly truncated install.wim file.

Now I had a bootable Windows 10 recovery/install USB stick. Great! The advice from Microsoft is to use the ‘fix startup problems’ option available on the menu. So I did that, and it spent a while diagnosing, then proclaimed that it couldn’t fix my PC. Great. I tried the other options (returning to a system restore point, removing the latest updates) but they all denied that my PC had Windows 10 on it at all.

This is getting annoying. Time to break out the command line, which is probably the most useful tool in the Windows 10 recovery toolkit. The official advice from Microsoft is to do this:

bootrec /scanos
bootrec /rebuildbcd
bootrec /fixmbr
bootrec /fixboot

But the /rebuildbcd stage failed for me, with a ‘The requested system device cannot be found’ error. The next advice I found was to manually reconstruct the BCD (the Boot Configuration Data, which tells the boot process where to find the operating system) by doing:

bcdboot c:\windows

But that didn’t work either.

I had to dig deeper. Luckily I’m not the only person to have had this problem. This page at screwloose.com.au explained more advanced use of bcdboot, in which it’s possible to tell the tool where the Windows installation lives, and where the System partition which holds the boot information is. I had to rummage round with diskpart to find the right drive letters, because the layout of the partitions on my drive meant that they were not the same as during normal operation. However, I came up with the following:

bcdboot e:\windows /l en-gb /s g: /f all

which worked! It completed successfully! Reboot and…the error was still there. Looking more deeply with diskpart, it turned out that the 100MB partition (drive G:) that was supposed to be the startup partition was formatted NTFS and wasn’t marked as active. The screwloose page indicates that the startup partition needs to be formatted as FAT32 and marked as active. Clearly, formatting it FAT32 deleted all the data on it, so I ran the bcdboot command again to re-create it.

Success! After a shutdown, the PC booted and came up as if nothing had happened. I’m happy that it’s working again, but it’s still a mystery how the error occurred. Something must have got corrupted, but I can’t believe that the entire contents of the startup partition got erased and replaced with an NTFS filesystem.

This post contains no electronics or software, just woodwork.

Our house is full of traps for the inquisitive toddler. There’s cat food and litter, an office full of sharp tools, a fireplace, and most tricky of all, a staircase descending from the corner of the living room guarded only by a coffee-table-height shelf.

Fortunately, there is a handy range of wooden barriers available from various sellers on allegro.pl (search for ‘bramka’ or ‘barierka’ to get started). It’s a sort of modular barrier system, with a basic unit 91cm wide, though other widths are available. The sections screw together with plastic couplings that allow them to hinge. Wall mounting brackets and feet are available, as are sections with a child-proof gate built in. They’re cheap and simple enough that hacking them for specific jobs is easy and worthwhile.

Here’s a set across the end of the living room, protecting the fireplace and cat flap to the garden. These are almost freestanding, but tied to the bannisters at one end with string and blu-tacked to the floor by the window.

The steps up to the office are guarded by these two sections. I added the cat flap in one of them so that the cats can get to their food and litter tray.

The stairs down from the living room are guarded by another hacked set of barriers, cut to length and with an aperture provided for our eldest son’s favourite bookshelves.

I used the leftover upper and lower rails from the segment I’d cut down in length to make a rail to hold the ends of the uprights above the bookshelves. There’s a join in the middle done with one of the plastic hinge pieces and a not inconsiderable amount of wood glue.

The barrier is held up using zip ties through plastic corner blocks screwed to the underside of the wooden shelf top, so the screw holes won’t show when we eventually remove it. Felt pads on its feet stop it scratching the wooden floor.

In the workshop I have a couple of these Tascam M-1B line mixers. They’re a very useful box: they have eight inputs, each with a level control and pan pot, which are mixed together and fed into a master volume control. There’s also a pair of headphone outputs with their own volume control driven by a headphone amplifier with enough voltage swing to drive 600 ohm headphones properly. Into the mixer I typically feed the sound output of my PC, other audio sources such as a cassette player (yes!), and a spare cable lying around on the workbench ready to connect to whatever I’m working on, or my phone. The output is connected to the workshop speakers. With this setup I can hear all the sources at once without fiddling around, switching anything or unplugging anything.

Tascam M-1B Line Mixer

The workshop I set up a couple of years ago turned out to have space for a new audio source: a turntable. Great! I can listen to vinyl (not ‘vinyls’, please). I had a spare turntable on the shelf, so I put it in place. The trouble is, the turntable has a magnetic cartridge which can’t be connected straight to the line inputs on my trusty mixer. It needed amplification and RIAA equalisation. I could have just gone and bought an off-the-shelf preamplifier, but where’s the fun in that? Looking at the back of the Tascam mixer, there’s a blanking plate which looks perfect for adding a preamplifier, neatly built in. No worries about trailing cables or yet another power supply to plug in. Now to find a suitable circuit.

A quick web search pulled up an application note from National Semiconductor for their LM833 audio op-amp which shows a simple preamplifier circuit. It then goes on to describes the circuit’s deficiencies and how a two-stage design can improve on it, but I decided to stick with the simple version.

I didn’t have any LM833 op-amps in the spares box, but did have some NE5532s, which are not only an excellent audio op-amp but live two to a package. Finding the passive components in the spares box also proved to be a challenge. Getting the right component values is one thing, but finding two identical ones of each proved to be impossible. However, all was not lost, so I thought. The RIAA equalisation curve needs three time constants, of 75µs, 318µs and 3180µs. I found pairs of resistors and capacitors that gave the same time constants as R1C1 and R2C2 even if their values weren’t what the original design specified. Here’s a circuit diagram with my substitutions marked in red.

I built up a small piece of matrix board and tried it out. It worked, but didn’t sound quite right: there was a noticeable lack of bass. I decided to check the frequency response. The application note shows a useful table of the RIAA response at various spot frequencies, so I had something to compare my results with. I used the ARTA STEPS program with a PC sound card to measure the frequency response of my circuit. It took a bit of fiddling with the levels to get good results, since the phono preamplifier has a lot of gain, but the result looked like this.

The curve looks convincingly like the RIAA curve, but it should be +17dB at 50Hz (relative to 1kHz). In fact it’s more like +13dB in my version, so I’ve lost 4dB of bass. At the other end of the spectrum it’s more accurate: at 20kHz it’s at almost exactly -20dB as it should be. (Note, when taking readings from the plot, that the level at 1kHz is about -3dB. It’s the relative levels that are important).

What I’d neglected in my component value calculations was that the relative values of the two RC networks are also important. The National Semiconductor application note actually goes on to explain more about this, but I hadn’t read any further than the circuit diagram. Rather than delving in to the maths and risking discovering that I needed more components I didn’t have, I started playing with the values I actually did have. Here’s a plot of the before-and-after empirical results.

The green curve is the modified version. The level at 50Hz is now +17dB relative to the level at 1kHz, as it should be. Below 50Hz there’s a bit of a rolloff, but that’s not something I’m worried about – my speakers don’t go down that low and it makes for a useful rumble filter. The modified circuit is shown here.

I had to change both RC networks to get a good result. The time constant of one of them (11k/6n8) is still 75µs, but the other (180k/22n) is more like 4000µs than 3180µs. I suspect it’s interacting with the time constant of R0C0 to give roughly the right result.

It fits neatly in to the slot in the back panel of the mixer. The toggle switch connects the output of the preamplifier to inputs 1 and 2. Switching it off leaves those inputs available for normal line-level use if desired. There’s also an earth terminal.

Time for some testing. What were the skies like when you were young?

This is my Canon S100 digital camera. After long, somewhat arduous service, the shutter button started playing up: it would focus but not take the photo. Eventually I got a mobile phone with a better camera and quietly forgot about the Canon. The S100 has been extensively frozen, boiled, soaked, rattled and trodden on, so I figured that the shutter button problem was just due to the abuse. It doesn’t owe me anything. However, it’s still useful for some things the phone camera won’t do: it fits a tripod, and has good long exposure and manual control facilities including a handy neutral density filter.

Trying to use it to take some tricky oscilloscope photos the other day, the shutter button got really annoying. I searched on line for the problem and discovered that it’s quite common. That sounded like an interesting problem to me. Here’s my attempt at fixing it.

Before we start, don’t try this unless you’re used to dealing with annoyingly tiny things and have the tools to match: jeweller’s screwdrivers, tweezers, a very fine soldering iron and either amazing close-up vision or a good magnifier.

The camera comes apart surprisingly easily. There are six little black screws on the outside: two on each end and two on the bottom, close to the tripod mount. Take those screws out. The back comes off very easily – take care not to lose any of the buttons though.

The front also comes off easily, though there’s a little FFC connector to the setting ring round the lens. Unplug it carefully. It doesn’t have a locking bar so the flex PCB just pushes in and out. There’s also a foam gasket round the lens and the plastic cover for the HDMI and USB sockets. They’ll fall on the floor.

The shutter button and other controls are underneath the top cover. There are two more screws to remove: one on the front near the lens and one on one end. They’re different lengths so take note of where each came from. Before removing the top, there’s one more FFC to unplug. This one connects to the microphones. Again it has no locking bar so just pull gently. It’s the middle of the three FFCs in this photo.

With the top cover off, the controls themselves are revealed. On the left here, the function selector rotary switch, then the shutter button and zoom switch, then the power button, and on the right the big square thing is the GPS antenna.

If you put the battery back in, the camera actually works fine in this dismantled state, so it’s easy to try experiments. I prodded and poked the shutter button. Sometimes it worked, sometimes it didn’t. I doused the button with my favourite contact cleaner (DeoxIT D5) to no avail. I looked on the web for replacement buttons: “double action tactile switch” seemed to be the best thing to search for. There are lots out there which might fit, but they’d take a while to get hold of and would be really fiddly to solder, especially with the zoom switch so close by.

I wondered about a mechanical solution, extending the plastic rod from the outside shutter button that presses the switch to give it more poke, but sometimes the switch just refused to work no matter how hard I prodded it with a screwdriver, so that probably wasn’t worthwhile.

Given that it wasn’t worth spending a lot of time on what’s effectively a life-expired camera, an alternative engineering solution was called for. The first stage of the shutter button, the half-press to lock the focus and exposure, still works reliably. What if I just wired it in parallel with the shutter release contact? Then there would be no way to lock the exposure and focus, but at least the camera would take photos.

Some fiddly soldering with a fine iron bit and some polyurethane-insulated mod wire later, the two opposite corner contacts on the button are connected…

…and the S100 lives again. It’s not a full fix, call it a workaround, hack or bodge, but it’s good enough for me to take pictures. Oh yes, and assembly, as somebody once said, is the reverse of disassembly.