Lumiere

The Warning

This amp operates at very high voltages that can kill you. Do not attempt to build this amplifier unless you have already built several other amplifiers, preferably kits so that you can learn good techniques. I would highly recommend the Bottlehead kits.

The Competition

See this link for the rules, updates and eventually the results.

The Schematic

The Parts List

	Description	Manufacturer	Vendor	Vendor part #	cost	qty	total
C1	6.8 uf 600v	Solen	Tube Store	CA-SO-560UF-630V	$3.40	1	$3.40
C2	0.47 uf 400 film & foil	Dayton	Parts Express	027-458	$1.47	2	$2.94
C3	7.5uf 370 vac motor run	ASC	Allied	225-2238	$6.87	2	$13.74
C4, 5, 9	47uf 350v	Nichicon	Mouser	UVR2V470MHD	$1.51	3	$4.53
C6, 7	180uf 400v	Nichicon	Mouser	UPZ2G181MHD	$3.48	2	$6.96
C8	22uf 450v	Nichicon	Mouser	UVR2W220MHD	$1.04	1	$1.04
D1 - D6	UF4007	Fairchild	Mouser	512-UF4007	$0.18	6	$1.08
L1	5h choke 600v 200ma	Hammond	Angela	P-T193H	$28.00	1	$28.00
M1	2SK3564	Toshiba	Mouser	757-2SK3564	$1.47	2	$2.94
M2, 3	DN2540	Supertex	Mouser	689-DN2540N5-G	$1.60	4	$6.40
R1	249k metal film 0.25 watt	Xicon	Mouser	271-249k	$0.09	2	$0.18
R2	50 carbon film 0.5 watt	Koa	Mouser	660-CF1/2C	$0.19	2	$0.38
R3, 4, 5	1k carbon film 0.5 watt	Koa	Mouser	660-CF1/2C	$0.19	6	$1.14
R6	300 wire wound 5 watt	Vishay	Mouser	594-AC05W300R0J	$0.51	2	$1.02
R7	100 thick film 3 watt	IRC	Mouser	66-GS31001000JLF	$0.51	2	$1.02
R8	121K carbon film 0.5 watt	Koa	Mouser	660-CF1/2C	$0.19	2	$0.38
R9	100k pot 0.5 watt	Bournes	Mouser	3362W-1-104LF	$0.75	2	$1.50
R10	22k carbon film 0.5 watt	Koa	Mouser	660-CF1/2C	$0.19	2	$0.38
R11	130k carbon film 0.5 watt	Koa	Mouser	660-CF1/2C	$0.19	2	$0.38
R12	20k metal film 3 watt	Vishay	Mouser	594-5093	$0.46	2	$0.92
R13	5k wire wound 25 watt	Vishay	Mouser	71-HL25-06Z	$2.28	2	$4.56
R14	180carbon film 2 watt	IRC	Mouser	660-SPR2CT521R181J	$0.16	2	$0.32
R15	2k wire wound 20 watt	Vishay	Mouser	71-CP20-J-2K	$0.75	1	$0.75
R16	200 carbon film 0.5 watt	Koa	Mouser	660-CF1/2C	$0.19	2	$0.38
R17, 18	200k carbon film 0.5 watt	Koa	Mouser	660-CF1/2C	$0.19	2	$0.38
R19	1k carbon film 2 watt	Koa	Mouser	660-SPR2CT521R102J	$0.16	1	$0.16
T1	Toroid 115:115 - 9:9 25VA	Antek	Antek	AN0209	$11.00	2	$22.00
T2, 3	VPS230-350  80 va	Triad	Allied	967-8035	$23.65	1	$23.65
T3, 4	12.6 volt 3 amp	Radio Shack	Radio Shack	273-1511	$10.49	2	$20.98
U1	12HG7	AES	AES	12HG7	$4.50	2	$9.00
U2	6V6	AES	AES	6V6	$8.95	2	$17.90
	9 pin socket	AES	AES	P-ST9-300	$1.50	2	$3.00
	8 pin socket	AES	AES	P-ST8-807	$1.95	2	$3.90
	RCA socket	Kobiconn	Mouser	161-1002	$0.98	2	$1.96
	Binding posts	Pomona	Mouser	565-6883	$2.13	2	$4.26
	IEC with fuse	Kobiconn	Mouser	161-R3013-E	$2.00	1	$2.00
	DPDT switches	Radio Shack	Radio Shack	275-666	$2.99	2	$5.98
	Total						$199.51

The Photo

It was April 2009 and Steve and I were relaxing towards the end of yet another Bottleneck meeting and somehow one of us got the idea that we should have a competition and after a few minutes discussion, the idea came forth for a 6V6 amplifier to be built for under $150. We announced the idea on the Bottlehead site and after further discussion we realized that $200 would be a better limit. With the help of Steve, David and the Dowdy Lama, the rules were refined and I started to ponder how I would actually build such an inexpensive amplifier. One of my suggestions to the builders was that this should be a learning experience and that it was an opportunity to try some new techniques.

One of the rules was that the container should come from Ikea but the cost of the container would not be included in the $200. Just for the fun of it. It was interesting to walk through an Ikea looking at various possible containers and try to work out if they would be viable. I can still remember looking at a candelabra thing and wondering if I could stick a socket, RCA socket or Binding post at the end of each arm. Fortunately, sanity prevailed.

We limited entries to those that were carried through the front door at Steve's place and finished up with about six or seven entries. At the time of writing, all of them are supposed to be single ended; the need for 4 output tubes and the $200 limit discouraged the use of push pull topology.

Since I am frugel by nature, my first decision was to use or reuse parts that I already owned which still could be purchased. Of late, I have been dismantling older projects that were flops or that I don't listen to anymore.

For a 6V6 in triode mode, the RCA manual suggests 250 volts, 45 ma and -12.5 volts with a 5K load. Plate dissipation maximum is 14 watts. I decided to go with those suggested values.

The first question for me was single ended or push pull and since I prefer single ended, that is what I chose. Next, which transformers to use. I presumed most would use Edcore or Hammond, so I decided to try something different. I had read that it was possible to use toroids as output transformers and that they could sound pretty decent. I did the calculations and found that a 115+115 : 9+9 toroid would work out to have an impedence of about 5000 when the 115 volt windings were wired in series and the 9 volt windings were in parallel. Antek had them for $11 each in 25 watt versions. They looked promising and certainly were cheap enough.

((115+115) * (115+115) / (9*9)) * 8 = 5224

Since there is no airgap in these toroids, I would have to use a parafeed topology or possibly some alternative such as running DC through one of the secondaries to offset the DC running through the primary. This sounded like an interesting experiment to see if it would actually work, but I had my doubts since you don't hear of anybody who has actually done it.

I calculated that if I was pushing 45 ma through the primary, I would need 1.15 amps through one of the secondaries which could be generated by a CCS.

(230 / 9) * 0.045 = 1.15

Don't take this equation as gospel because I did not try it and I suspect I am incorrect in my thinking about this concept.

Ikea intervened with a container in the shape of a large red candle holder for $15 called a Borrby. Including the chimney, it stands about 18 inches high and 8 inches on each side and now comes in only black and galvanized. Red was available when I bought mine so who knows why that colour is not available now. I figured I could replace one of the glass sides with polycarbonate so that I could install the IEC connector and switches towards the bottom of the container along with the heavy power transformers and the choke. Once I saw this candle holder, I knew I had to use it.

Since the container was a lamp, I decided just for the fun of it that I should bung some lightbulbs in there to use as the parafeed plate load just as I had done for my 829B amp. Each 240 volt 25 watt bulb works out to be 2000 ohms. I figured I would put two per channel with a 1k ohm resistor in series between them to achieve the desired 5k ohms which at 45 ma would give a 225 volt drop. Of course, such bulbs are unobtainium in the US, so I installed two DPDT switches and a 5k 25 watt wirewound resistor for each channel so that for the competition purposes, the wirewound resistors could be used as the parafeed plate load. The bulbs would just be for show. This decision had an implication, since I would now need about 500 volts as B+ which would complicate the power supply. While not at the 800 to 1000 voltage level, this is still quite a high and potentially dangerous voltage.

The next decision was the driver. I have used the 5687 and 5965 in a few projects and recently the 12HG7 wired as a triode. I built an initial breadboard with a switching arrangment where I could quickly compare these three tubes to see what I liked. Here I had an advantage over the other contestents since I had a pretty good idea of what the music would be, The torture track is to be Tamacun by Rodrigo y Gabriela. and it is recorded louder than the other tracks on our competition CD and starts off with a resounding whack. Only the 12HG7 was convincing so that is what I designed around. There is nothing wrong with the other tubes and in certain ways, I preferred the sound of the 5687 but I figured a low powered amp needs all the help it can get.

Pete Millet used the 12HG7 in pentode mode in his 813 amp and also used them in his single driver test of pentode tubes. In his test, he ran them in triode mode at 35 ma with a Constant Current Source load and resulted with a gain of 39. This is a serious driver.

I've been using Gary Pimm's excellent CCS designs in some other projects and have been pleased with the results. The cost of the boards would have been a problem with the $200 limit, so I decided to use a cascaded pair of Supertex DN2450 as the load for each 12HG7. These devices are a bit cheaper than the IXYS 10M45S which I could have used as an alternative. A 300 ohm resistor resulted in 28ma so I figured that would do. Incidently, I have a 45 volt battery that I use to test the CCS's before I insert them into the circuit. I just use clip leads to connect the CCS to a 100 ohm 10 watt power resistor and to the battery. Measuring the voltage across the resistor is easy and a simple calculation gives the current. In this case:

2.8 volts / 100 ohms = .028 amps.

I've been playing with Tubelab's Powerdrive of late so I decided I would try to include this if possible to get a bit more power out of the amp. If you read the link, class A2 becomes possible with more power output from the amp. I didn't do anything special and just used the resistor values that he used for the SE board. The 2SK2700 seems to have become unobtainium these days so I substituted the 2SK3564. They seem to work ok and only cost $1.47 each. Incidently, in this configuration there is about 5 or 6 ma passing through the 20k resistor, R12. One cost saving advantage about using Powerdrive is that you don't need a cathode resistor or cap for the 6V6.

To construct the CCS's and the powerdrive circuitry I just use the little boards from Radio Shack. They are fine for simple projects like this, just don't try to reuse them or the pads come off.

So now I had six heat producing semiconductors. For a heat sink I just attach them to a 7 inch long 1/8" thick piece of aluminium angle bar. The bar is attached to the Ikea Borrby container by the bolt that I eventually used to hold the output transformers and so far this appears to be adequate heatsinking.

There is a philosophy that an amplifier is no better than it's power supply so I tried to make sure I didn't skimp too much there. There were two complications, first the need to get  a bit under 500 volts for B+ and the need for a negative 150 volt supply for the Power drive. So I started off playing around with the excellent PSUD2 tool from Duncan Amps. I knew I would need approximately 45 ma for each of the 6V6's, 30 ma for each 12HG7 and 5 ma for each Powerdrive, a total of 160ma or a total of (500 volts * 160 ma) = 80 va. More VA capacity than 80 would be better.

Since low cost was an objective, I immediately thought of the Triad VPS230-350 which is a useful 80 VA transformer whose inputs and outputs can be configured as 230 or 115 volt. I figured I could configure the input for 115 and the output for 230 and then have a voltage doubler circuit. Triad makes a 130 VA unit that would be a better choice, but it is over $7 more expensive than the 80 VA unit and I already had the 80 va unit so I decided to try it and see how it went, even though it was at it's limit of 80 va. If necessary, I could reduce the current through the 12HG7's. One of our competition rules was that the items could not be purchased from Ebay so that somebody else could build the same amplifier and be able to get the parts for the same cost. If this rule were relaxed, industrial transformers are pretty common on Ebay and would be a good cheap alternative. Some can be configured with 480 volt outputs.

My spreadsheet cost calculations showed that I could probably use a choke so initially I used a Hammond 5h 300 ma 57 ohm 800 volt max choke (193L) since I already had one laying around even though it was overkill for this size amp. The 200 ma 65 600 volt max version (193H) would have been more than sufficient for this project and cost less. Eventually I bought and substituted with the 193H because I made an error in my spreadsheet calculation. I had exceeded the $200 limit so needed to cut something and this was after I had finished building. Curses!!!! The photos actually show the 193L choke. There are less expensive 5h chokes from Hammond such as the 158Q, but the 400 volt voltage rating was too low, and the resistance at 105 ohms was too high.

For the filaments I decided to use a Radio Shack 12.6 VCT 3 amp transformer. They are relatively inexpensive and I had a few sitting around recovered from past projects. For the negative voltage, I simply used another in reverse and connected the 12.6 side to the 12.6 wires on the filament transformer. It's a simple way to get 115 VAC. This second transformer could have been smaller than 3 amps.

Rectification is done with the usual UF4007 diodes and I used Nichicon capacitors that were recycled from other projects. Good alternatives would be the Panasonic ED or EE series from Digikey. R17 & 18 are 200k resistors across the 180uf Nichicons to balance the voltages between the two capacitors wired in series. These resistors also server as bleeders to reduce the voltage to zero after the amp is turned off and as far as I am concerned, all amps should have bleeders. When you are first testing your amps, it is helpful to watch how quickly the voltage returns to zero so that you can get an idea of how long it takes before it is safe to touch the circuitry again.

I use a dual switch arrangement so that I can turn on the filaments and the negative voltage first and the other switch turns on the high voltage. You can read about this here.

Resistors are generally 0.5 watt Koa carbon films from Mouser. In quantities of one they cost 19 cents but if you buy 100 of them they are just 3 cents. Whenever I need a new value, I simply get 100 and I have enough for years. They also have a 400 volt rating which is higher than the typical metal film resistor. The rumour is that these Koa resistors are the same as Kiwame except with a much lower cost. 

The coupling caps are the Dayton Film & Foil caps from Parts Express and I preferred them to the Orange Drop 716P I also tried. You could probably get away with 0.22uf instead of the 0.47 I used. You guessed it, I had them sitting around. If the Ebay constriction was not in force, I would have used the Russian K40Y-9 paper in oils. I tried bypassing the Orange Drops with Russian silver mica SSG-2 caps and for me, they made a huge difference. Of course, the selection of capacitors is a matter of taste.

For the parafeed capacitors I used ASC motor run caps. Incidently, I have taken to mounting these monsters with GOOP which is a very effective glue. I get it at the local ACE hardware shop and it certainly beats messing around with the motor run clamps. Just make sure to allow 24 hours for curing. From a couple of years experience with this glue, caps mounted with GOOP appear to be quite secure but with a reasonable amount of force, they can be removed.

Wire is mostly solid core 18 gauge except for 14 gauge for the ground bus. For signal wire I use some twisted pair stuff from old IBM Type 1 cable. This is 22 gauge solid copper with some sort of foam sheafing that is not affected by soldering iron heat. I use a standard colouring scheme for all of my projects, black is for AC, red for B+ high voltage, orange for driver wiring, blue for negative and yellow for heaters. Green is for ground. I am not sure if there are wire colouring standards, but that is what I use so that I can look at one of my amps and work out pretty quickly what is what.

So I built the breadboard and after a while, got it to work. There were a lot of clip leads involved. I took the amp to a mini meet with Steve and David and for $200, it's not a bad amp. Those who have heard my other amps at various meetings will be surprised to hear that it actually worked on arrival. It's always interesting to hear your amp in a different location with different speakers and it's a great reality check. I thought the midrange could use help, but the high frequencies had that open quality that I get with my 829B amp. My old octal socket punch had died so before I went home, Steve kindly drilled some socket holes for me in the top part of the candle holder. I was now ready to build.

So I took the breadboard home and dismantled it and started to stuff the parts into the Ikea candle holder. There are essentially three sections. The power supply is at the bottom, the light bulbs and plate resistors are mounted to polycarbonate mouted vertically in the middle and everything else is at the top near the chimney. The toroids fit quite well within the chimney itself and I used some Radio Shack terminal strips to mount most of the other components. Gingerly, I finally put the sides on and finished all the wiring connections from top to bottom.

It worked first time and has been fine ever since. The voltages for the 6V6 and 12HG7 are as expected. The top gets just slightly warm which is to be expected since the container is designed to get rid of heat effectively. If I need to get inside, there is a door that can be opened and it is not too difficult to remove one of the sides if necessary. The beast weighs about 23 pounds.

So far, I haven't used the light bulb option except in some of my initial testing. They do not light up as much as the 829B amp which is a bit disappointing, just not enough current passing through them. However, when the high voltage is turned on, there is a flash of light for a moment as the amp stabilizes. There is a bunch of current passing through the 6V6 initially which is something I would never have caught with a multimeter. The light bulb does give you a visual look into the operation of the amplifier.

Initially, I was not sure that this amp would be a keeper, but the more use it gets and the more I listen to it, the more I like it. It will be very interesting to compare it with the other entries in a blind setting. Eventually when the competition is over, I might change the power supply and use an industrial transformer and some motor run caps that I have kicking around.

So overall, it has been a wonderful project and I have learned a lot from the whole process. It's a real challenge having to stick to a cost limit and it makes you truly respect those manufacturers such as Bottlehead who produce such great kits at such low prices. Try it yourself sometime.

The Breadboard. A goodly number of clips in evidence.

RCA plugs. They should be closer to the 12HG7 tubes, but there was no room. Note the hole ready for the bolt to hold the toroid transformers.

Sockets and binding posts installed. The sides of the candle holder fit into the slots.

I replaced one of the glass sides with a piece of polycarbonate and cut holes for the IEC holder and DPDT switches which are partially wired.

Here you can see the DN2540 CCS on the left and the powerdrive on the right with the bar of aluminium.

The transformers for the power supply.

The Antek toroids in the chimney and the ASC motor run caps attached with GOOP.

The middle part with the back of the light bulb holders and the 5K plate resistors. Just above and below the plate resistors you can see the DPDT switches that can switch between the bulbs and the resistors. Naturally, I would only flip those switches when the power is off. The keen eyed will notice that at this stage I do not have the 1k resistors that connect the bulbs.

I also use GOOP to attach the Nichicon caps to a small piece of polycarbonate which I then can attach to the larger piece of poly. There is a module for the main B+ and another for the negative 150.

This is essentially the view through the back door in case I need to get to anything.

The final test before I assembled it all. Note that I use a bunch of multimeters and hook them up with clip leads. I then write down the expected voltages and after turn on, write down the voltages I see. If there is a big discrepency in any of the voltages, then I turn off immediately and find out what went wrong. The red meters on the right come from Harbor Freight, cost $7.99, supposedly can handle 1000 VDC and work fine for these purposes.

I must admit I am not a very tidy wirer but this is what the back of the top area looks like. Thank heavens it is mostly covered up by the slots at the top of each side. In the foreground are the ASC caps.

And here is the front view of the top. You can see the yellow and black Dayton coupling caps.

And finally, the back door. Even with the door shut, heat is not a problem.