Some notes, observations and ramblings about my build of Douglas Self's Trimodal Amplifier. The opinions expressed are my own; use at your own risk.
Please do not copy and repost these photographs or link directly to them from my server. This page is not intended to be a step-by-step guide to building the Trimodal nor a replacement for the documentation that comes with the boards, but simply a resource for information about my own build. Strangely, there seems to be a dearth of construction information about these amplifiers on the web so hopefully this will be the start of a trend. Introduction
A few years ago while passing through Singapore, I picked up a copy of "Self on Audio" at the Kinokuniya book store on Orchard Road. Much to my wife's chagrin, I spent the rest of the holiday with my nose buried in that book. I've wanted to build one of the amplifiers described in the book ever since. But choosing which one seemed impossible; the "Blameless" class B looked like a solid performer but would I want to enjoy the measurable and obvious (on paper, anyway) benefits of the class A version?
But because the Trimodal amplifier works in class A/B and B it seemed like a logical choice. Switching modes is as easy as flicking a switch.
There is a catch, however. To take advantage of the potential for increased output power in the class B mode would also require switchable rail voltages - something a bit too complex and costly that would probably require a two separate power transformers or a custom unit that could switch the rail voltages. Since I ordinarily listen at fairly low levels so the modest 20 Watt output using +/-24V rails isn't much of a problem.
It's a real hoot to read comments on the various forums about Douglas Self's amplifiers but I feel confident that actually building one will be much more satisfying than merely reading the rampant speculation about his designs.
There is one "review" of a fully-built but otherwise apparently vaporware version of the amplifier that exists here.
Sourcing Parts
RESISTORS
I purchased all the boards bare since I believed I had many of the parts lying around. Starting with the common resistor values I began by using Dale CMF series resistors having a surplus of them around from another project.
Thinking that it might look odd to start adding the normal blue 1% resistors to the board, I decided to order the missing values in the CMF series. Some of them, however, were unavailable in the needed values so I substituted Military RN series resistors, forgetting that the 1/4W body size is much larger than the standard 1/4W resistor. No matter, at least the color matched. Be aware, however, that many of the resistors are clumped together in tight groups on the board so larger resistors simply won't nestle tightly to the PCB if surrounded by other normal resistors, let alone larger ones. A few odd 1/4W values were not available in either style so I just used some normal resistors I had in stock.
The 3/4 Watt 100 Ohm base stopper resistors are p/n CMF65100R00FKEB. For the 2.5W 10 Ohm on the zobel, I used the black-bodied Dale RS02C10R00FB12. Also, before receiving the boards and documentation I ordered the 0.1 Ohm emitter resistors, guessing 5W would be sufficient (the documentation that came with the PCBs specifies the odd value of 6 Watts). The 5W wirewound resistors are Vishay/Draloric p/n AC05000001007JAC00. If you're pedantic or paranoid, I've found the Vishay RW67VR10B12 in 6.5 Watt, but have not managed to find a 6W 0.1 Ohm resistor from my usual sources.
CAPACITORS
The numerous 0.1 uF decoupling caps have a 0.394" pitch (10mm in new money). I used Mallory 160104J100C-F mainly because the brand has a nostalgic appeal to me. In reality they ended up being rather boring white Chinese made film caps.
The small RF blocking cap is specified as polyester in the official BOM. I only had some multilayer ceramic C0G caps but in the correct value so I used this instead, re-bending the leads to match the 5mm pitch of the PCB footprint.
I chose the Panasonic NHG series ECA-1EHG470 47uF 25V for the numerous decoupling caps and input DC blocking cap. The pitch seemed slightly narrow for the board footprints but wasn't a big issue.
The two 220uf 63V should have a 5mm lead pitch meaning the large supply of squat short Nichicons I had was useless. Instead I ordered the Nichicon UPW1J221MPD6 which fits the footprint perfectly. Because I tend to be a bit careless with leaving the boards lying around when building, I prefer to mount components flush with the PCB. This keeps them from getting pushed over like limp scarecrows. For electrolytics, this might not be the best practice as they tend to leak electrolyte from the bottom when they fail and this can damage PCB tracks. Elevating them off the board can prevent or minimize the damage caused by leaks.
CONNECTORS
I had wrongly assumed that finding the connectors for the boards would be fairly easy. The BOM refers to them as "Molex Mini-Fit" but a search will reveal a baffling array of connectors with frequently unrepresentative pictures. There are various ratings of material, "junior" and "senior" versions, versions with and without PCB mounting pegs etc. The parts needed for these boards are Mini-Fit Jr. Series with a 4.2mm pitch and two rows. The headers should have no pegs.
Fortunately I ended up with the correct connectors but incorrect quantities (note the missing header from the photograph above). To cut to the chase here are the things you'll need: Each connection will require a vertical header which mounts on the board and contains male pins. In addition, you'll need a wire housing/recepticle which does not come with any metal contacts. So then you'll need female sockets which are crimped to the wire and inserted into the housing.
For the four circuit Mini-Fit assembly:
For the six circuit Mini-Fit assembly:
For the eight circuit Mini-Fit assembly:
For the twelve circuit Mini-Fit assembly:
Mini-Fit sockets for the above housings:
For the 2 pin KK series:
Note that these are not the only versions of these parts you can get that will work. Many decisions are required - in a nutshell I chose UL flammability rating 94V-0 and phosphor bronze sockets. There are both cheaper and much more expensive options, gold plating if you're into that sort of thing.
Make sure that, unlike myself, you choose sufficient quantities to fit both the amplifier and protection boards. Also consider whether you will use the same connectors on your power supply side.
A crimping tool is required to secure the wires to the contacts. After trying two locally available tools - the TH1834 from Jaycar and the T1537 from Altronics - I can confidently say that neither one is appropriate for the above mentioned Mini-Fit contacts. The jaws are simply too wide to effectively crimp the wire contact portion alone - they end up crushing the part that secures the wire's insulation. The latter tool may work for the KK contacts though.
The relatively inexpensive Molex 63811-1000 works well for the Mini-Fit contacts with a section on either side of the pivot for the conductor and insulation portion. Obtaining proper joints requires some practice though and is neither particularly easy or intuitive. Using the sockets listed above even with the maximum specified gauge I found that shortening the insulation crimp tang length improved the final result. A good tutorial on connector crimps is located here.
It's worth purchasing the Molex extraction tool as well. The part number is 11-03-0044. Note that this tool works with the Mini-Fit sockets listed above. The KK contacts are easily removed with either a tiny flat screwdriver or similarly shaped tool.
POWER TRANSFORMER
Virtually all the off-the-shelf power transformers in Australia have a single 240V primary winding. But none of the locally available toroidals appear to come with the electrostatic screening suggested in the construction notes. Initially, I wasn't worried about either aspect and initially bought a 300VA 18+18V toroidal transformer from Altronics.
Later I did a search and found AnTek. They had the AS-4218 - a 400VA transformer with two 120V primaries, two 18V secondaries and an electrostatic shield for around half the price of the locally available Chinese product. Shipping brought the final price to a few dollars more than the Altronic product.
Early in the piece I also contacted Harbuch about one of their "stock" transformers but never heard back from them.
TRANSISTORS
Ordinarily I choose Fairchild semiconductors for most of my projects for no particular reason besides habit, I suppose. The nice side benefit for this project is that the medium power transistors have no exposed metal casing so insulators are unnecessary. I was not able to locate the metal clips used on the transistors that are secured to heatsinks so I just used screws.
For the output stage, I had a pair of On Semi MJL1302A/3281A transistors lying about and was just going to use them rather than the suggested 2SA1295 & 2SC3264. It looked as though they were just another part I would have ordered with the boards had I known how seemingly scarce they were. But a search on Ebay revealed they were plentiful. Later, I discovered they were a stock item on goodluckbuy.com. Make sure you order insulators for the unusual MT-200 package or be prepared to cut bulk material to size.
MISCELLANEOUS PARTS
The on-board fuse holders with covers are available in Australia from Altronics as S5985 or elsewhere as Schurter 0031.8001 (holder) 0853.9561 (cover).
Output Inductor
The manual specifies 10 turns of 16 gauge enameled wire 20mm in diameter. A good trick learned while doing metal work is to take the length of wire, put one end in a vice and grip the other end tightly with a set of locking pliers. Pull until the metal stretches slightly. It's an oddly satisfying feeling when you reach the right tension and the metal "gives". Now you have a perfectly straight piece of wire to wind with.
I found a disused curtain rod that was slightly smaller than the intended final size - perfect since the wire is bound to unravel slightly after winding. I cut a small notch in the end to hold the wire.
The photo shows the unused part of the coil; wind plenty of material so you're assured of 10 clean winds for each inductor.
Strangely, the finished inductor measured nowhere near the correct value. The text of Audio Power Amplifier Design Handbook specifies 2.3uH for this inductor. If the finished amplifier breaks into spontaneous oscillation under typical loads I will revisit the inductor.
Power Supply
I have a surplus of nearly new 10,000uF 71V capacitors mined from old amplifier boards that are too good to sit idle. They were originally manufactured by Nippon Chemi-Con and the only drawbacks are their 85 degree C rating and the "LI" terminal code that consists of two flat lugs placed 90 degrees to each other.
After creating a new component in Eagle and starting to lay out a simple PCB I happened to have Ebay open and did a quick search. I found a board that was very similar to what I was laying out (how complicated can a bank of caps be?) routed for LI terminals for around $12 for two boards!. Postage was a few bucks from Bogor, Indonesia.
A couple of weeks later I had a pair of boards in hand. I used some heavy gauge solid wire on the bottom between the capacitors to reduce the resistances between the PCB tracks.
I'm not certain what a dioda is but I'm sure I'll eventually work it out. A more intractable problem may be finding a bridge rectifier with a symmetrical flat lug pattern that matches the board. Oh wait, it looks like they are made for a D5SB inline package (which is what is printed right there on the PCB).
Not long after ordering the boards I noticed that Signal Transfer had begun selling a power supply board of their own. The board includes a +/-17 Volt supply for other circuitry such as their balanced input card.
Case
After a lot of indecision I decided to get a Par-Metal Series 20 case even though shipping to Australia was more than the price of the case itself. The stock number was 20-16164N. So far, so good - it looks like everything will fit.
Protection Board
The documentation that comes with the protection board is very specific about the type of relay used, warning about the introduction of distortion by using typical "power relays". Signal Transfer stocks the correct relay so in hindsight it might have been a good idea to purchase them with the board.
However, I was able to find a suitable relay locally at Rockby Electronics but in the GZ-SS-124LM variant, made by the same "GOODSKY" outfit. This relay is SPST rather than SPDT.
This hardly matters as the solder layer of the board shows that the other "throw" of the switch is simply shorted out on the board itself so is unused.
In the middle of the schematic that came with the protection board is a note saying that on issue 4 of the pcb resistor R34 had to be added to the solder side of the board. I did not find reference to this anywhere else in the manual so it could potentially be easy to miss. It seems a surprising oversight for a small board that costs 23 UK Pounds.
Input Boards and +/-17V Power Supply
I wanted to have both unbalanced RCA and balanced XLR inputs and decided to use Self's low noise design for the balanced inputs. Others have pointed out that noise from the previous stage(s) will dominate the overall performance of a balanced input but the allure of having no perceptible increase in noise above the inky black silence of the amplifier when switching between inputs was irresistible. So much so that I designed my own Low Noise Balanced Input SMD PCB.
A small regulated power supply runs off the raw voltage derived from the capacitor bank to power the two input boards which sit on the first layer (the power amplifiers will sit above). The adjustable power supply is my own design that I sell (for the Rev F 1176) with the regulators mounted from the bottom of the PCB and attached to the bottom of the chassis with silcone insulators. Note that the regulated power supply happily accepts a DC input rather than the usual power transformer secondary.
A rear panel toggle switch allows either the unbalanced or balanced connector to be used. The balanced lines get processed into unbalanced signals through the input boards full-time so the switching takes place in unbalanced mode for both inputs.
I changed the mounting and added a sheet metal shield above the input boards in a vain attempt to fix some hum. Later, I traced the hum to a stupid grounding mistake (of course).
After all is said and done I would have preferred that it looked a little nicer than it does internally. It all fits well within the Antek case, is dead silent but for the life of me I cannot hear what is so "diabolical" about the "sound" of this amplifier :^)
The difference between A and B? A little increase in hum. It's tempting to buy a 30V transformer and run it class B full-time. For some reason I seem to want to listen to things a little louder through this amplifier.
