[continued from Part 1]
10) DIY AudioNote M7 clone twin-triode preamp. Didn't work out as simple as I originally envisaged though.
Firstly the components were quite expensive, and required a separate transformer for the heaters. All resistors are Holco, electrolytics Black Gate, PSU caps Solen and coupling caps a parallelled combination of AuriCap and Solen. The pot is a Noble; the triodes are GE 6702's and the rectifier an EZ81. Wiring is silver-coated OFC with Teflon, and coax is vdH.
Secondly I had a PCB made when the point-to-point wiring became a mess. On the PCB is also a 20s timer to soft-start the heater supplies for prolonged life. The heater regulator didn't want to power up under load when turned on, so it had to be replaced by a different one that didn't fit on the PCB, and got much hotter than anticipated, hence the tacky heatsink. The triodes are in a separate section and shielded somewhat by an aluminium sheet. The actual audio stage resistors are all soldered onto the tube sockets directly. Unfortunately the design has a lot of gain, and the volume pot is at the input - that means any internal noise is unattennuated.
Thirdly, the massive CLCLC filters are fine for attennuating ripple, but due to the very low current flowing through them, there were some sporadic low frequency (actually more fluctuating DC) on it that that coupled happily through the preamp due to its low PSRR. This went straight through the highpass filter at the output, and caused unnerving cone excursion even at idle. A simple 300V regulator between the tubes and LC filters fixed this, but raised the SNR to an unacceptable and audible level. Placing it before the LC filters is useless since they were the culprits. So I settled on a more complex dual-stage regulator that fortunately fitted snugly on a small open space just below the rectifier tube, and all was well. The only remaining problem is that I should have included relays to mute the outputs until everything has settled upon poweron and poweroff - it makes some strange noises so I just have to remember to switch it on first. The next version of the system controller will have delayed sequential turnon and turnoff control that will address this.
11,12,13) PC system. Once again a project that grew from a small attempt to improve crappy PC speakers to this monstrosity. The satellite speakers are truly remarkable Boston Micro90's. They sound fabulous with excellent build quality - the cabinets are cast aluminium and each weigh about the same as a brick.
The wooden box contains a 60W stereo amp (discrete, not chip amps); a 240W subwoofer amp, an 8" Vifa woofer in a sealed subenclosure that's equalized flat to 30Hz, a preamp, lowpass filter and phase switch. The ugly fan was added later when it turned out that the 2mm steel plate wasn't enough to dissipate the heat of the internally mounted heatsinks to the outside. The Vifa is scheduled for replacement with a Peerless SLS woofer since it got damaged during experiments.. On the inside can be seen the two amps, at the right is the preamp and subwoofer controller and at the left the toroid for all the low-level stuff. The 625VA power transformer is way too heavy to mount on the plate and is hence inside the box itself.
The box on top is the optional master controller. When it is plugged into the main box its preamp section is bypassed, and the box becomes the master volume, power, source and output selection controller. It contains a USB soundcard based on the PCM2702 chip, normal analog inputs, a preamp based on the PGA2310 volume control chip and a microcontroller, analog and headphone outputs and a dicrete class-A headphone amp. To prevent nasty picked up noises on the speakers, the analog outputs leading to the amps are grounded when the headphone output is selected. This is the fourth version and yes, ver5 already being redesigned for smaller space and better performance
14) Just for interest's sake, a picture of the heart of the Mk2 version of my research (I'm working with Mk4 right now and already designing Mk5). I wanted to add a picture of the actual full Mk1 prototype in its chassis but it's travelling around somewhere. The board contains an S/PDIF receiver, sample rate converter and a massively capable Altera Cyclone2 FPGA. It's a 484-pin Ball Grid Array package, and routing the PCB was a nightmare that took several weeks on an 8-layer board. Was quite a relief when everything worked on powerup! ?
The FPGA does all the functions as explained in my Class-D amplifier basics thread to convert the 192kHz 24-bit PCM input to PWM output. The PWM is then fed to an off-board full bridge MOSFET switching stage (not in the picture) that delivers the power and consists of two gate drivers, the FET's, power supply bus capacitors and output lowpass filter. The MOSFET's used are special SMD ones from International Rectifier that are no bigger than a pinky nail; yet can deliver 100W without any heatsinking. Pressing the board with them on a small aluminum strip is enough to raise the power handling significantly, and with the current implementation the power supply voltage and current rating is the only limiting factor up until 500W, and then it's easy to just add another 4 of those small FET's to increase it even further. Idle dissipation is less than 1W.
10) DIY AudioNote M7 clone twin-triode preamp. Didn't work out as simple as I originally envisaged though.
Firstly the components were quite expensive, and required a separate transformer for the heaters. All resistors are Holco, electrolytics Black Gate, PSU caps Solen and coupling caps a parallelled combination of AuriCap and Solen. The pot is a Noble; the triodes are GE 6702's and the rectifier an EZ81. Wiring is silver-coated OFC with Teflon, and coax is vdH.
Secondly I had a PCB made when the point-to-point wiring became a mess. On the PCB is also a 20s timer to soft-start the heater supplies for prolonged life. The heater regulator didn't want to power up under load when turned on, so it had to be replaced by a different one that didn't fit on the PCB, and got much hotter than anticipated, hence the tacky heatsink. The triodes are in a separate section and shielded somewhat by an aluminium sheet. The actual audio stage resistors are all soldered onto the tube sockets directly. Unfortunately the design has a lot of gain, and the volume pot is at the input - that means any internal noise is unattennuated.
Thirdly, the massive CLCLC filters are fine for attennuating ripple, but due to the very low current flowing through them, there were some sporadic low frequency (actually more fluctuating DC) on it that that coupled happily through the preamp due to its low PSRR. This went straight through the highpass filter at the output, and caused unnerving cone excursion even at idle. A simple 300V regulator between the tubes and LC filters fixed this, but raised the SNR to an unacceptable and audible level. Placing it before the LC filters is useless since they were the culprits. So I settled on a more complex dual-stage regulator that fortunately fitted snugly on a small open space just below the rectifier tube, and all was well. The only remaining problem is that I should have included relays to mute the outputs until everything has settled upon poweron and poweroff - it makes some strange noises so I just have to remember to switch it on first. The next version of the system controller will have delayed sequential turnon and turnoff control that will address this.
11,12,13) PC system. Once again a project that grew from a small attempt to improve crappy PC speakers to this monstrosity. The satellite speakers are truly remarkable Boston Micro90's. They sound fabulous with excellent build quality - the cabinets are cast aluminium and each weigh about the same as a brick.
The wooden box contains a 60W stereo amp (discrete, not chip amps); a 240W subwoofer amp, an 8" Vifa woofer in a sealed subenclosure that's equalized flat to 30Hz, a preamp, lowpass filter and phase switch. The ugly fan was added later when it turned out that the 2mm steel plate wasn't enough to dissipate the heat of the internally mounted heatsinks to the outside. The Vifa is scheduled for replacement with a Peerless SLS woofer since it got damaged during experiments.. On the inside can be seen the two amps, at the right is the preamp and subwoofer controller and at the left the toroid for all the low-level stuff. The 625VA power transformer is way too heavy to mount on the plate and is hence inside the box itself.
The box on top is the optional master controller. When it is plugged into the main box its preamp section is bypassed, and the box becomes the master volume, power, source and output selection controller. It contains a USB soundcard based on the PCM2702 chip, normal analog inputs, a preamp based on the PGA2310 volume control chip and a microcontroller, analog and headphone outputs and a dicrete class-A headphone amp. To prevent nasty picked up noises on the speakers, the analog outputs leading to the amps are grounded when the headphone output is selected. This is the fourth version and yes, ver5 already being redesigned for smaller space and better performance
14) Just for interest's sake, a picture of the heart of the Mk2 version of my research (I'm working with Mk4 right now and already designing Mk5). I wanted to add a picture of the actual full Mk1 prototype in its chassis but it's travelling around somewhere. The board contains an S/PDIF receiver, sample rate converter and a massively capable Altera Cyclone2 FPGA. It's a 484-pin Ball Grid Array package, and routing the PCB was a nightmare that took several weeks on an 8-layer board. Was quite a relief when everything worked on powerup! ?
The FPGA does all the functions as explained in my Class-D amplifier basics thread to convert the 192kHz 24-bit PCM input to PWM output. The PWM is then fed to an off-board full bridge MOSFET switching stage (not in the picture) that delivers the power and consists of two gate drivers, the FET's, power supply bus capacitors and output lowpass filter. The MOSFET's used are special SMD ones from International Rectifier that are no bigger than a pinky nail; yet can deliver 100W without any heatsinking. Pressing the board with them on a small aluminum strip is enough to raise the power handling significantly, and with the current implementation the power supply voltage and current rating is the only limiting factor up until 500W, and then it's easy to just add another 4 of those small FET's to increase it even further. Idle dissipation is less than 1W.
