Time for another projector! This one was brought to me with a fault, described as a shadow in the middle of the image, shortly after the lamp was replaced after exploding. This is an older DLP projector, with a UHP mercury lamp. I’ve already removed the top cover of the projector here, showing the internals. The light engine is along the front of the unit, with the lamp on the right. The main control board on top contains all the image processing logic & control functions.
The other side of the mainboard holds the processing chipset. This is probably one of the biggest flip-chip BGA packages I’ve ever seen, the DDP2000. Along with the DAD1000 on the right, these format & send the image data to the DLP chip, via the large white header.
After the mains PCB is removed from the chassis, the rest of the light engine is visible. The DLP is hidden on the left, behind the large heatsink & interface PCB. The light engine is spread out a lot more on this projector, across the entire front of the unit.
A closeup of the light engine shows the back of the phase sensor for the colourwheel, and the mounting brackets for the optics.
The dichroic colourwheel is tucked into the gap between the lamphouse & the first optic.
Hiding at the back of the projector is the alloy frame holding the power supplies & cooling ducts.
After removing the brackets, the DC power supply & the lamp ballast are visible. Since this projector uses a UHP arc lamp, the DC power supply which has the usual low voltage outputs for the logic board, has an auxiliary output from the +340v rail after the PFC circuit that supplies power to the lamp ballast.
The lamp ballast is a pretty standard design, using an Osram control board.
After removing the top cover with the colourwheel, the main optic chain is visible. The usual mirror tunnel homogenizer at the start, with a convex & aspheric lens on the left.
The lamphouse has a last-resort thermal cutout to shut the ballast down if the cooling fans fail. These lamps output some serious heat, and likely wouldn’t last longer than a couple of minutes without cooling.
The final turning optics before the DLP chip are hidden in the Mg casting of the light engine.
The DLP is the older type, with the large ceramic LGA package.
After the DLP, the light is routed through the objective lens, to the screen. This is the back of the lens inside the light engine.
And here is the main problem with the projector – the last lens in the optical chain before the DLP chip has been roasted by the intense light flux from the lamp. Unfortunately NEC cheaped out on this one – it’s the only optic in the machine that isn’t made of glass. This was likely caused by some contamination on the lens, which starts the process of absorbing light on the surface. The resulting heat then causes discolouring of the lens, which absorbs more heat. A chain reaction ensues, ending in the lens completely destroying itself.
The projection lens has a couple of sensors, for the focus & zoom, along with a focus motor. This is driven by feedback from a distance sensor in the base so no manual focusing is required.
These projectors were very popular when they first appeared on the market with the laser hobbyist community, and for very good reason – they contain a massive array of 445nm Royal Blue laser diodes in their optics engine. Originally very expensive, these units can now be had for under £50 on eBay, usually with damaged DLP chips.
Under the door on the bottom of the projector is the 445nm Laser diode array module, itself secured in place with security screws to the beam combiner. The rack of 3 high speed fans to the left draws air over the substantial heatsink.
After removing the shell securing screws, the top cover comes off with the button panel. This gives a view of the internals, mostly PCBs at this stage.
On the left side of the projector is the main control PCB, with the video handling circuitry.
At the top of the board is the main DLP image processing chipset, these two components are actually custom parts, so no datasheets are available. The main DLP IC has some DRAM & a Spansion serial flash for firmware storage. There’s also a small audio amplifier on the left to drive the onboard 2W speaker.
Further down the board sees an unpopulated BGA footprint, with more space for DRAM. The main system microcontroller is on the right, a Renesas part.
Right at the bottom edge is the connector running off to the phosphor wheel drive motor.
The reverse side of the board is pretty sparse, there’s quite a few passives & power control. Down towards the bottom surrounded by inductors is the system power management IC, the DLPA100. This takes the incoming DC 12v rail from the connector on the right side of the board & produces several supply rails for the internal logic: 1.1v 1.8v, 2.5v, 3.3, 5v & also contains the 3-phase brushless driver for the phosphor wheel motor. The main control board input power connector also has a +5v from the mains supply, for standby power. The main board signals the PSU to switch on the main +12v rail through a pin on this connector.
The other end of the board just has the connectors, a bit of glue logic & the HDMI interface chipset.
After unplugging all the connectors, the massive cast frame of the light engine is visible.
Here’s a closeup of the phosphor stripe around the edge of the wheel. This takes the 445nm light from the laser module, and converts it into green. There’s also a frosted glass section of the wheel to pass some blue for the image. The reason for the phosphor being in a large stripe on the wheel is load spreading – there’s several watts of optical power focused down to a very small spot on this phosphor, and would overheat quickly if it wasn’t moving.
At the back of the light engine is the DLP module, with it’s substantial heatsink.
Hiding under the mains PSU, is the light source control PCB. This contains several DC-DC converters, which run the 4 strings of laser diodes, the large Phlatlight Red LED & it’s associated TEC cooler. This board takes the incoming +12v from the mains PSU through the multi-way loom at top centre. There are multiple cores on this connector to spread the load – at normal brightness, in Eco mode, I measured the power consumption at about 8.5A at 12v input for the entire projector.
The left side of the board is dedicated to the high power section of the controller. There’s a power inductor for every channel.
The other side of the board is very heavily populated with components.
The right hand side has the control logic, a Lattice CPLD, and another Renesas Microcontroller. There’s also some glue logic here & a dedicated DA converter.
The other end of the board has the power drive control logic. There’s a MAX16821AA LED buck driver for the Red LED, and 4 drive ICs for the laser diodes, which are marked <009 LDGC N249. I haven’t been able to find anything about these, so they may be custom.
Removing some screws allows the entire optical assembly to be removed from the lower shell. This may be mostly manufactured from a magnesium alloy from the rather low weight.
On the back of the DLP module is the DLPA200 Micromirror Driver IC. This generates the high voltage bias supplies for the DMD chip (+/-28v) from the 12v rail, generates all the timing waveforms required for the DLP chip. There’s a couple of power inductors for the onboard regulators. Video data is sent from the main image processing chipset to the DMD chip via 2 channels of LVDS.
Now the heatsink has been removed, the rear of the DLP chip can be seen, with the remains of the thermal pad. The mount for the heatsink is sprung, to accommodate thermal expansion.
The Red light required to create a colour image is generated by a giant LED, more on this one later.
Here’s the DLP board removed from the projector with the micromirror surface visible. This DLP has many dead pixels, hence the decommission at ~4500 hours of operation.
The 24 laser diodes have their beams combined by this knife-edge mirror assembly, turning the beam through 90° to the lens on the left, which focuses the 24 beams down to the optics engine.
Removing the beam combiner from the array allows the 24 diodes to be seen, mounted under their collimating lenses. This is one beast of a laser unit!
Taking the cover off the optics assembly allows the main optical path to be seen. The blue laser comes in a bottom left, through the lens, the red LED comes in bottom right. The pair of dichroic mirrors manage the light path for the red & green light, while passing the blue straight through.
Here’s another view of the optical path, with both light sources visible.
After the light source, is the homogeniser – this tunnel of 4 mirrors facing each other evens out the light beam & removes all coherence from the laser light. This is important to not have any speckle in the image.
Underneath the objective lens are the pair of stepper motors that drive the focus & zoom mechanisms, along with their position sensors.
Just after the homogeniser, is the final optical path to the DLP. Here the light comes in a bottom left, and hits the turning mirror, after which it is focused onto the DLP chip by the mirror top centre. The objective lens is through the hole in the centre of the optical block, while the DLP is on the right side.
Here’s where the DLP will be mounted in normal operation, with it’s lens in place.
Finally, the created image is passed out through the objective lens to the projection screen.
There’s a sensor mounted on the side of the lens barrel, that I think is a Hall effect device, but I’m unsure what this would be used for, as there is no magnet anywhere near this to sense. It could also be a temperature sensor though, for the DLP & lens assembly.
The small PCB on the side of the lens unit holds the stepper motor drive IC, an LB1937 from Sanyo. There is another IC here, which looks to be a microcontroller.
Removing the top cover allows the moving lens assemblies to be seen. These move independently of each other to implement focus & zoom, via lead screw drives on the stepper motors.
Here I’m shining a separate 445nm diode laser into the optical assembly, through the blue optical path. The phosphor wheel is turned to the clear section, which allows the 445nm light to pass straight through, being turned 180° by the mirrors & directed out towards where the DLP assembly would be.
Turning the phosphor into the light path causes a very bright green light to be generated, and passed back towards the 445nm laser entry point. The dichroic mirror in the way reflects this light to the left, through a lens & then to the other dichroic mirror to be turned another 90° to the DLP assembly. I’m not sure where the magenta light is coming from – the phosphor probably generates light on more wavelengths than just pur green, giving some red to mix with the blue.
Here’s the Red LED removed from it’s cooling & collimation assembly – this has an enormous silicon emitter area, and apparently these LEDs are designed to be uniform in light emission, specially made for projection use. There’s a thermistor onboard for temperature sensing – sensible when the datasheet gives CW currents of 8A, and pulsed currents of 13.5A!
Not surprisingly, cooling this beast of an LED requires more than just a heatsink, so it’s mounted on a TEC module, possibly around 40W thermal capacity.
Fan control is handled by this little PCB, squeezed in between the optics engine & 445nm Laser array. There’s a SMSC EMC2305 I²C 5-channel PWM fan controller on here, communicating back to the main system microcontroller. Besides some passives, and 4 transistors to make sure the fans don’t start at full power when the projector is powered on, there’s not much else.
I figured it was about time I built another valve amplifier, and since I already had most of the required parts in stock, here it is! Above is the lid of a cake tin sourced from a local shop as a case, marked out & drilled for the valve sockets, output transformers & speaker terminals.
The ECL82 valve is a Triode & Audio Output Pentode in a single envelope, requiring only a single valve per audio channel. There are a pair of extra holes drilled here for a couple of EM80 magic-eye valves wired as VU meters to give a bit of a lightshow.
Here’s the base schematic for the Class-A ECL82 amplifier sections, obtained from the interweb. It’s pretty basic, and doesn’t mention a value for the volume potentiometer, so I used a 100K audio taper for that. Power will be supplied from low-voltage DC, running through a high voltage DC-DC converter for the anode supply of 200v, and a 5A buck converter for the 6.3v filament supply.
The EM80 side is as the schematic above, the signal input being taken directly from the Pentode anode of the ECL82. I have removed the second 1N4148 diode down to ground, leaving only a single diode.
Most of the parts comprising the ECL82 amplifier stages are mounted directly on the back of the valve sockets, requiring only a 6.3v filament supply, 200v anode supply & audio I/O connections. Axial electrolytics have been used for ease of assembly, even though they’re getting a little expensive nowadays!
After fitting the components to the top lid, point-to-point wiring is used to connect up the valve socket assemblies. Some large electrolytics provide B+ smoothing, and all the filaments are daisy-chained in parallel. Audio is brought in on micro-coax from the I/O, and straight out to the output transformers on twisted pairs, keeping the audio wiring away from the B+ voltage.
The audio transformers, from a 1960’s Philips Radiogram, are mounted behind the valves, with the wiring emerging through holes in the case. I’ve already done the paint job here, in metallic copper.
Audio & power sockets are on the back of the tin, with both 3.5mm Stereo inputs & phono inputs. A DC barrel jack takes care of the power, accepting 12-24v.
Controls on the front provide volume, balance, bass & treble adjustments.
Here’s the amplifier with it’s valves glowing nicely. Total power consumption is roughly 30W, using NOS Svetlana ECL82s & EM80s. In operation there is no hum or noise in the background, with no audio input the connected speakers are entirely silent.