My P3A amplifier
Page: [1] [2] [3]

The problem (or good thing) about DIY is that designs always seem to be evolving. I'm currently rebuilding this amp for the third time or so, trying to evaluate the quality and sound of a tube input stage. The current configuration will probably stabilize around this:

Channels 1+2 Elliot P3A
  • rated 60W continuous (10dB headroom)
  • 300mA bias, 42V rails
  • 50,000uF supply caps/channel RCRCRC filter
  • 400VA xform
  • 5687 input tube
Channels 3+4+5 Elliot P3A
  • rated 60W continuous (10dB headroom)
  • 150mA bias, 44V rails
  • 30,000uF supply caps/channel RCRC filter
  • 400VA xform
  • 5687 input tube
Channels 6+7+8 LM3886
  • rated 40W continuous (7dB headroom)
  • 34V rails
  • 22,000uF supply caps/channel RC filter
  • 350VA xform
  • 2nd order MFB opamp input stage
Valve supply
  • 100V plate supply, <1uV ripple
  • 1880uF supply caps RC filter x 4
  • Regulated heaters

Why so many channels?

Channels 7-8 are actually not being used at the moment. The idea is that the two beefed up P3A channels provide the stereo pair, with channels 3-5 for center and rear surround. Channel 7 is for a center (ES) surround. If (when) I move to active crossovers channels 6-8 will be dedicated to surround allowing me to biamp the stereo pair with the P3A amps. I believe that biamping will have the added effect of never having to leave class A on most material. (effective class A power of about 30W for most programs)

Power supply

The power supply is separated from the chassis containing the actual amp boards. This was actually a lot more work than I had anticipated. I used 10 gauge copper wire for the power leads and installed on-board 100uF bypass capacitors to help minimize instantaneous power losses. All ground connections come back to a bus constructed of 1mm copper plate. The power supply is housed in a standard BUD chassis with a 3RU rack plate bolted to the front.
Image: Supply (32K)

The 'R' in the RCRC supplies is a standard 2ohm cement resistor. For the RCRC supply this reduces the ripple by about 20dB over a capacitor-only filter. In normal listening the stereo amp stays in constant current (class A) operation, so the RCRC filter serves to push supply hum below the noise floor. Each channel has its own filter caps and 35A bridge.

Component selection

Outputs: 2SA1302/2SC3281
Drivers: BD140/BD150
Resistors: Vishay/Dale 1% bulk metal film
Coupling cap: 2uF 450V oil/paper 1%
Input shunt: polystyrene 5%
Miller caps: polystyrene 2%
Bypass/zobel: metallized polyester
Electrolytics: Xicon
Input: Tung-Sol 5687 "black plate"
I'm very skeptical of the popularity of "boutique" components. I must admit that I've never had the opportunity to try a Hovland etc. premium cap in a design, because the $100+ price is somewhat off-putting. At least the premium resistors have some definite basis for quality difference, but at $10+ they too are a somewhat disappointing value regardless of their quality. Therefore, most of the components in my amp are standard issue.

Small signal wiring is teflon coated 22ga silver plated copper (shielded). Outputs are 12ga teflon/silver plated copper. Inputs are RCA and 1/4" TRS (balanced). (I would have preferred XLR, but for various reasons went with the TRS)  The input plate is 1mm copper stock, and I tried to provide as much separation as possible between the TRS inputs and the high level outputs.

The heatsinks are massive Alcoa extrusions, acquired used over Ebay. They are oriented incorrectly, which is not a problem because they are overkill for the task. The chassis is homemade; steel plate makes up the top and bottom, and a standard 3RU steel rack plate serves for the front.
Image: Heatsink (28K)
Image: Chassis (48K)
 
©2002 Tyler Thornblade  Contact Info  Disclaimer