Hi-Fi Preamp - PSU and Mechanical Details:

The power supply is essentially very straightforward. The mains enters via a filtered IEC inlet and powers two toroidal transformers. These connect to the power supply PCB, which generates two separate ±15V supplies for the main analogue PCB and the surround PCB. The +5V logic supply and the +12V relay supply is also generated here, and there is a mains-fail detect circuit which controls the output muting relays.

There is no mains power switch, as the unit is intended to be left powered up continuously. While in 'standby', all circuits remain powered, apart from the VFD display. Fortunately, the preamp doesn't consume much power, so it shouldn't adversely affect my electricity bill!

The analogue supplies are regulated by LM317T/337T's - chosen for their improved regulation and lower noise output compared to 7x15 types. Additionally, should it become necessary, they provide the option to vary the supply voltages. The 12V +RELAY supply is regulated by an LM2940, which is a low-dropout 12V regulator - this means that a mains dip shouldn't cause the record-out relays to drop out and affect a recording (assuming the rest of the equipment holds up!). I felt that it was definitely worth regulating this rail - apart from the above reason, it ensures that the relay coils are fed with a clean supply, eliminating the possibility of hum and noise being induced into the audio circuits. Finally, a standard 7805 deals with the logic supply.

The regulators use the case as a heatsink - they are bolted onto a nicely-machined length of aluminium which is in turn bolted to the bottom panel. The PSU PCB serves one final purposes - to separate the serial control connections from the 34-way ribbon cable and pass these through to the control PCB via the black connector on the right.

Top view of the PSU

No surface-mount stuff here!

Although it would have been possible to derive the logic and relay supplies from the same transformer used for the analogue supplies, it would have resulted in increased dissipation due to the greater voltage drop across the regulators. The VFD takes between 150 and 200mA, which is significant. Also, purists will assume that this will automatically affect the audio signal, and that isn't totally impossible. The main advantage of separate transformers, or separate windings on a single custom-made transformer in the real world, is ground management. The issue of avoiding earth loops and minimising digital breakthrough is complicated to say the least!

Ideally, the analogue and digital grounds should only be connected at one point. The PGA2310 datasheet suggests that this is close to the IC, so this happens on the main analogue PCB. Having the PIC on a separate board helps, and when the PGA2310s aren't being addressed there is no activity on their serial control bus. A common trick to achieve further isolation is to use opto-couplers to electrically isolate the digital and analogue section - I didn't feel that was required here as I couldn't detect or provoke any digital breakthrough on the PGA2310 test PCB.

The mains-fail detect circuit is based on good old-fashioned transistors. This is because you can predict exactly how they will behave during power-up and power-down, which is not necessarily the case with CMOS or TTL logic! The main function is to control the un-mute relays, as discussed on the control systems page, but there is also an active-low PWR_FAIL output that is passed to the control PCB - at the moment this hasn't been implemented in software, but I plan to create a power-fail interrupt that causes the PIC to write the current programme state to EEPROM during the brief hold-up time before the power supply capacitors discharge. Another option that I have is to use the PIC to pull down the PWR_FAIL line - this will drop out the un-mute relays, disconnecting the power amps from the outputs and reducing power dissipation slightly. This would be used during Standby, after a routine dump-to-EEPROM has occurred (as obviously the PIC won't be able to respond to the PWR_FAIL signal in this state). This neatly saves tying up another port on the PIC and having to run an extra signal between the PSU and Control PCB's

Mechanics:

I chose an all-aluminium case that features extruded front and rear panels, resulting in no screw heads on the front panel. These are much more solid than the cheaper cases, and being made from 1.6mm aluminium instead of thin steel meant that all the screws on the bottom panel can be neatly countersunk (unlike on the NICAM Tuner, which was built in a cheap Maplin case.

Despite buying some really nice panel-mount gold-plated phono sockets a while ago (as seen on original box and some very expensive hi-fi), I decided to use PCB-mount sockets. These aren't as nice as the separate sockets, but have lots of advantages for the Production dept. There is much less assembly - the free sockets have a hex pattern, which makes it even more difficult to line them up neatly. You don't need to attach lengths of wire and terminate them in Molex connectors (which are another connection to potentially upset your music!). The PCB-mount sockets can support the rear of the PCBs which makes assembly easier...

The front panel has a sub-panel to hold the switch PCBs and the VFD. This was an important detail to get right, as a thin panel will flex and resonate when switches are pressed. We've all seen products that are built like this - press a switch and notice the others move in sympathy!

Sub-panel made from 1/8 inch "L" stock - neatly notched to clear the bolt heads. That's because I'm no good at tapping aluminium!

The front view, showing how the various PCB's are mounted.

This image shows the sub-panel mounted on the chassis, along with the rotary encoder and the headphone socket PCB. From the position of the spacers, you can see how the PCB's overlap the bottom of the bracket as mentioned above. You can also see that more material had to be removed from the bottom of the L-bracket to accommodate these spacers. Obviously, I could have just made quick cut-outs to clear them, but I think it looks nicer to clean the opening out properly. There isn't any significant loss of rigidity as the panel is 3mm stock...

This image shows the detail of the Store switch - this is much smaller than the rest of the switches and, like the Reset button on your computer, is intended to be less easy to press. The tactile switch and switch button came from an old car stereo - these can be good sources of switches if they need to be illuminated.

The clear perspex is required because the switch is offset to allow maximum light from the LED to get to the button. It's held by a simple M3 spacer and bolt - this works rather well in practice.

This shows the final assembly - you can see the headphone PCB, the rotary encoder, switches PCB, VFD display, control PCB, and bits of the phono PCB and main analogue PCB. Not bad for a prototype!

Talking of which, here's how the opposite corner worked out. The wiring to the D-type sockets had been worrying me a bit, but as you can see, it all made sense in the end. When planning ribbon-cable runs, it's easy to get things confused and you end up having to make extra folds and twists in the run. I last made that mistake with my NICAM tuner - when I was planning the layout of the PSU/analogue board, I was effectively viewing the PCB from the rear of the unit. If you have a look at the ribbon cable that joins them, you'll see the extra twists required to correct the error.

You might have spotted that I've mounted the mains socket on the inside of the rear panel. This is a temporary measure that will be corrected once I've finished the rear panel. Otherwise, I'd have to unsolder the mains connections each time I need to remove it.

You'll also note that I've maintained the length of the primary connections so that they reach the mains inlet. This is good practice, because you are then able to re-wire the input easily for 120V operation if you emigrate to the US!

Note that the earth connection has its own bolt that is not used for anything else. This is also good practice and possibly a mandatory safety requirement in some parts of the world.

You can just about see the corner of a piece of aluminium that I added between the bottom panel and the transformers. This is needed because the bolts are M4 - it would otherwise be impossible to countersink them into a 1.6mm panel. Rather than a long bolt, I like to use a short bolt to secure a 20mm spacer to the bottom panel. These securely hold the extra plate, and allow you to use another short bolt to fasten the dished washer that holds the transformer. This saves you having to accurately cut short the supplied bolt, minimising the risk of the top panel coming into contact with the bolt.

Although the transformer is close to the surround PCB, experiments show that this is not a problem. I did a lot of testing with the workshop hifi, and I could only hear mains-induced hum if I set the amplifier to maximum gain. In this condition, it has a sensitivity of some 100mV - a typical power amplifier would be nearer to a volt. I found that I could minimise the hum by carefully rotating the transformer - this is normal, and it's wise to allow for this during construction. I also found that placing a small piece of magic metal cancelled the hum completely, so this is an option that I can use if the slight hum is audible when used with real power amplifiers. This so-called 'magic' material came from an old reel-to-reel tape deck, and obviously has special magnetic properties, as 'normal' steel was not as effective here...

Conclusion:

At the time of writing, there isn't much outstanding work. Apart from the software, I just need to finish the front and rear panels. But, these things always take much longer than you think!