All-Purpose Stereo 5-20 front-end PCB

thorpej

Super Member
In my day job, computer software is my thing, and one thing that really irks me in that discipline is having to do one-offs. Even worse, a one-off that's really similar, but ever so slightly different, to the previous one-off. When that happens, I like to take the time to refactor the code and in so doing build myself an easily-adaptable, re-usable building block.

So, I was taking inventory of the projects I have on the drawing board... and all of the power amps in that queue all use the Mullard 5-20 front-end topology (favored by @kward, myself, and others) -- a gain stage direct-coupled to a long-tail pair phase-inverter/driver stage.

So, I thought to myself -- hey, perhaps I should attempt to make building block that will be flexible enough for all of these projects (no two of those front-ends were exactly the same), but will still save me a bunch of time and hassle.

And so here it is: Jason's All-Purpose Stereo 5-20 Front-End board. Features:
  • Fairly small ... as currently laid out, it measures approx. 7.25" long by 3.25".
  • No tube sockets ... it just has provisions for using TE-style screw terminals or directly-soldered flying leads to make the tube connections, so tube choice and chassis layout is flexible.
  • Resistor footprints are big enough for 2W resistors, if needed.
  • Provision for radial electrolytic cathode bypass capacitors on the gain stage, if desired (or you can just leave them out).
  • Provision for split cathode bias resistors in the gain stage, if needed by your NFB loop. If you don't need a split resistor, just solder in a jumper for the bottom leg.
  • To power supply inputs -- one for the gain stage, one for the inverter/driver stage, in the event that different supply voltages are required.
  • No explicit terminal connections, but soldering on flying leads for CCS modules (for the gain stage plate load or for the LTP tail) should be fairly straight-forward (in fact, I'm counting on the latter).
  • Similarly, if you wanted to add a gain adjustment pot for the non-inverting triode of the LTP, that shouldn't be too difficult with flying leads.
  • Large axial footprints for the the LTP output coupling caps and the LTP grid bypass caps. Polarized footprints are used to indicate correct orientation of the outside foil end of the cap for lowest noise (negative end).
This isn't quite complete -- there is one tweak I want to make (to provide the ability to use a negative rail for the tail of the LTP, if desired); this will get drawn as a small resistor footprint that connects the tail resistor to ground -- solder in a jumper to use the ground rail, or solder in a flying lead to use a negative rail. And I have yet to write up the documentation of how to use the board.

But I do plan to make a small run of these boards, and would be happy to offer them up to other interested AK'ers for cost + mailing (and would increase the production run, if necessary -- minimum order for the fabricator I use is 10).

Attached are a couple of CAD teaser photos.
 

Attachments

  • Screen Shot 2017-12-23 at 7.24.41 AM.png
    Screen Shot 2017-12-23 at 7.24.41 AM.png
    258.9 KB · Views: 102
  • Screen Shot 2017-12-23 at 7.28.10 AM.png
    Screen Shot 2017-12-23 at 7.28.10 AM.png
    138.7 KB · Views: 92
Nice idea. Mount that underneath from standoffs and it saves a lot of manual wiring. I like that it doesn't have the sockets specifically because you can run different tubes. Guessing this only works with triodes though, not the pentode front end that traditional Mullard amps use?
 
Nice idea. Mount that underneath from standoffs and it saves a lot of manual wiring. I like that it doesn't have the sockets specifically because you can run different tubes. Guessing this only works with triodes though, not the pentode front end that traditional Mullard amps use?

Yah, that was exactly my thought. It should also work just fine with a pentode gain stage -- you would just need to provision the screen supply / bypass cap as point-to-point wiring at the socket.
 
Spent a little time refining it a bit more, and I think it's basically ready to go, I think. R17 is now a place to tie in a negative rail for the LTP, if desired (lower pad). If you don't want it, solder a jumper across. And J8 and J9 have been added to provide a place to solder flying leads for the NFB (use either J8 or J9, depending on which part of the split resistor you want to tie in the NFB loop). You can use an appropriate Phoenix connector if you like :)

Any comments on the layout? Over the next couple of days, I'll write up a document that describes how to use the board.
 

Attachments

  • Screen Shot 2017-12-23 at 7.16.21 PM.png
    Screen Shot 2017-12-23 at 7.16.21 PM.png
    246.5 KB · Views: 42
How wide are the traces? I'd consider making them as wide as you can without getting ridiculous. On a sample board I just got back from the fab house, I made the traces .05" (1.27 mm) and I wish I would have made them double that size. Also perhaps use a copper pour for the ground. Also make the capacitor pads significantly larger. Perhaps make room for someone to use radial caps or axial caps in one outline (multiple pads on the one outline).Things I wish I had done on my board.

I like the idea of not having the filament traces on the board.
 
How wide are the traces? I'd consider making them as wide as you can without getting ridiculous. On a sample board I just got back from the fab house, I made the traces .05" (1.27 mm) and I wish I would have made them double that size. Also perhaps use a copper pour for the ground. Also make the capacitor pads significantly larger. Perhaps make room for someone to use radial caps or axial caps in one outline (multiple pads on the one outline).Things I wish I had done on my board.

I like the idea of not having the filament traces on the board.

Thanks for the feedback, Kevin. Traces were the KiCad default of 0.25mm. I went ahead and bumped them all to 1.0mm (had to re-route a couple slightly). This is all low-current, small-signal stuff, so that seems sufficient, yah? The routing between pads where all the resistors are gets tricky if it gets any wider.

All of the ground paths on this are pretty short, except for one (the R channel LTP tail-to-negative-or-ground trace), and on that one I just managed to find a path without having to use a via. I suppose I could fill a bunch of the board and tie it to input ground... I'll experiment a little.

(This is why code review is good! :) )
 
BTW, those large axial capacitor footprints are 20mm wide with 41mm pad spacing. Easily accommodates yellow tubular polypropylene caps, and is pretty close for 600V orange drops of the same capacity;
 
Guessing this only works with triodes though, not the pentode front end that traditional Mullard amps use?

One other thing on using a pentode gain stage... in that case, if you were building a true reproduction of the original Mullard 5-20, the negative feedback loop would tie in to J9 (the lower leg of the split bias resistor) and the screen bypass capacitor would tie in at J8 (the upper leg of the split bias resistor, right at the cathode). The easiest way to do this would be to mount those capacitors on a terminal strip elsewhere, and run flying leads to the board. Even in this case, the board saves a ton of manual wiring.
 
I've replaced 3 Marshall power boards because the filament traces were so close to the B+ traces that (with the help of the aging board itself being conductive) the two voltage sources would mix and melt the tubes in some amps because the filaments received B+. Keeping the filament traces off the board CAN be a good idea.
 
Another compromise solution for delivering heater power is to put a two pin connector next to each tube. If using vias holes, one can solder them on either side of the board. I’ve used this solution on two boards and liked it more the one with traces in the board.
Thanks. Paul
 
Very cool! I have seen boards with 9 and 8 pin socket holes. The 8 pin socket mounting holes spaced around the 9 pin socket holes connected with traces.
That would make the board larger though.
 
I have an old piece of gear that is all tubes on a circuit board. The heater wiring is all done with twisted pair wiring going tube to tube. No heater current on the board at all.
 
Again, the primary design principle here is that tubes are not on the board, and tube choice is not dictated. My plan for the filament wiring would be "well, that's something you just do by hand".
 
British tag board-ish. Interesting, since another source already offers boards with either nine pin or octal sockets as front end plus drive 5-20 modules.
I could definitely see diy demand.
 
Back
Top Bottom