1-stage (killer, literally) amp made of junk

[dovla]

Got restless while reading what other people do but had no parts on hand, other than tubes and iron that cost me a princely sum of 15 Euros for PT, 4xOPTs (2 small and 2 medium-sized, Grundig and Telefunken). Took apart two old (but working) PC power supplies and harvested what could be used - not much luck with resistors of needed values but got barely enough for a stereo amp.

Power is 250V AC unloaded into 4A 1-piece bridge, then 165 uF cap (2x320 200v in series), followed by 2 small OPT in parallel used as choke, then 235uF cap (2x470 200v in series, didn't have any spare resistors to put across caps but all 4 share the voltage evenly). Small OPT/chokes are 12K to 4R originally used with EL95s so they should be good for at least 30mA each. Paralleling them gives higher B+ as DC resistance is 360 Ohm per primary (have no idea what inductance so I put 3H for the two in parallel into PSUD and got 7mV ripple). B+ comes to 283V B+ at 70mA current draw and 265V with 94 mA.

The amp consists of an EL84 with 200K grid leak, biased by LEDs (3 or 4 pairs of red for 6.2V or 8.3V bias giving 42 or 32 mA plate current approx), the plate connected to the bigger OPT, which is 3500:91 turns ratio (made for 5.2K to 3.5R in its original use). Screen voltage is dropped by some 3.9K and comes to just about the same the same as plate voltage after OPT drop (OPTs have rather high 410R DC resistance).
Datasheet shows EL84 can give 2W output with 2V RMS in and it does seem to be right if the load is correct. I get just about enough volume with 96dB 8-Ohm speakers (so loading the tube with 10-11K) in the big room and almost enough with 90dB 8-Ohm bookshelf near-field but with only 0.6V input from the laptop phones jack.
OPTs seem to be decent, I think they are the same size as Edcor XSE 10W. I need to increase the bass below 80 Hz by some 6dB but then it does play it, there's no audible distortion or saturation and cathode LEDs do not blink.

What came as the biggest surprise is the total absence of any noise or hum with the ear in the horn of a 96dB speaker. When you look at the pictures it's hard to believe as the whole thing is solderless (apart from connecting power caps in series), some resistor strings are held together by mini-grabbers and shrink-wrap.

Now I'll transfer the thing into some form of chassis - probably two Dutch waffle metal jars side-by-side. (I WILL get sockets and proper wattage resistors and put grid stoppers also).

I'm sure not all of the power components would fit together so the question is how to package it:

1. PT and caps together, chokes with the OPT under the tubes (not good because of transformer heat close to the caps?, what's the impact of chokes and OPTs in the same can if oriented perpendicular?)
2. PT and chokes together (PT on top of chokes but fastened to the top cover, not sitting on them), caps with OPTs under the tubes (heat from tubes goes up and OPTs do not heat up?)
3. Eat more Dutch waffles and put everything in 3 cans/jars?

When I left the amp working for some 2 hours in the open boxes nothing was even warm (tubes excepted), can't say to what degree it'll remain so when enclosed. PT came which some schematic that shows 160mA fuse on the secondary so it's probably what it's rated for and I'm using half that.

I'd also like to leave some space in the 'amp can' to add a double triode for the front end if I change my mind and go two-stage (can't try it now with no resistors though I have some coupling caps)

Some other questions not directly related:

4. one of the PC power supplies I disassembled had a 0.33uF cap paralleled with 470k resistor across the IEC socket - what does that do and is it worth retaining for the tube amp?
5. that MKP cap is marked 275V AC but no DC rating - would that automatically mean some 400V DC (so use it as a bypass on electrolytic) or should it not be used for DC (the size is decent - some 2 x 1.5 cm)?

[tomchr]

Don't try this at home, kids... We're what you call experts.

I love death traps like that. But one has to know what one is doing, know one's hand position at all times, and keep kids and pets at bay.

Do you have a schematic for the contraption?

~Tom

[dovla]

1-stage EL84.jpgIt takes more time to draw it (cut & paste from other people's work) then to actually build the thing. I suppose there is something to the whole 'wire with gain' thing.
Not recommended to have it running on your desk for hours but a teepee above it made of a sizable hard-covered book is ok until it gets disassembled and put in an enclosure. Everybody in the household is aware it can kill (and inquired how much I'm insured for, with hopeful expression on their faces) so nobody is coming close.

Edit to schematic - first two caps are 330u, not 220 as drawn

[leadbelly]

Whoa who there. It's more killer than you realize! Don't turn it on again until you put some resistors across each filter cap. Right now there's nothing to prevent each filter cap from getting more than 200V across it and blowing up!

[tomchr]

Love the comment about the ground not really being attached to the cardboard box....

[dovla]

Originally Posted by leadbelly

Whoa who there. It's more killer than you realize! Don't turn it on again until you put some resistors across each filter cap. Right now there's nothing to prevent each filter cap from getting more than 200V across it and blowing up!

As I noted: "didn't have any spare resistors to put across caps but all 4 share the voltage evenly". That was the first thing to measure with the foot ready to step on the extension cord switch if needed - unloaded DC is 350V so each cap gets 175 for 10-15 seconds before the tubes warmup, which does not leave any margin for voltage drop differences. Anyway, all caps behave and voltage division is 50/50 to within 1/2 of a volt.
They were also series-connected in PC power supplies and I see no resistors across (and one of those supplies was even made in Taiwan, not China mainland) so maybe they are actually of decent quality.
The biggest pain is waiting for caps to discharge before changing anything - with only the meter connected (1 Mohm) it takes 20 minutes to drain them from some 130V that they have at switch-off. I've now located several 1/8W 1M resistors in the leftovers so I should have enough for a safe bleeder of some 250-300K.

[leadbelly]

Originally Posted by dovla

As I noted: "didn't have any spare resistors to put across caps but all 4 share the voltage evenly". That was the first thing to measure with the foot ready to step on the extension cord switch if needed - unloaded DC is 350V so each cap gets 175 for 10-15 seconds before the tubes warmup, which does not leave any margin for voltage drop differences. Anyway, all caps behave and voltage division is 50/50 to within 1/2 of a volt.
They were also series-connected in PC power supplies and I see no resistors across (and one of those supplies was even made in Taiwan, not China mainland) so maybe they are actually of decent quality.
The biggest pain is waiting for caps to discharge before changing anything - with only the meter connected (1 Mohm) it takes 20 minutes to drain them from some 130V that they have at switch-off. I've now located several 1/8W 1M resistors in the leftovers so I should have enough for a safe bleeder of some 250-300K.

Completely irrelevant. That a circuit operates safely when everything works perfectly is not enough to make it a safe design. It's more about safety when something fails. Ever read the electrical code? What portion of it is about protection from failure? Know what happens when a big cap blows?

Did you trace out the entire PC PS circuit around those caps? Sure there weren't any other traces to anything else that was keeping voltage on the caps balanced?

[dovla]

Leadbelly, calm down

I'm not defending the 'design' - of course there should be resistors across those caps but I just didn't have any that could be used. No point in deliberately making uneven voltage split with different value resistor if caps themselves agreed fine. It's not going to stay this way, small parts are on the way to make it compliant to common sense (and electrical code)

BTW, still nobody answered my original questions about component placement within the chassis (and plural would be ?)

[Nikolas Ojala]

I like your amplifier because it is so simple, but would it be worse or better if there was also a capacitor parallel to LEDs at the cathode?

[dovla]

Originally Posted by Nikolas Ojala

I like your amplifier because it is so simple, but would it be worse or better if there was also a capacitor parallel to LEDs at the cathode?

The point of the LED is to not to have a capacitor - SY's RLD article is a good scientific explanation on what goes on with LED biasing. Paralleling LEDs reduces impendance and makes sure the current passing though EL84 cathode is handled without burned plastic. On the other hand, too many parallel string will not have enough current per LED so 3 is ok to share 30-40 mA but 4 is probably too many.
One important thing to keep in mind (and SY's article points it out) is that LED bias is fixed voltage so grid resistance limit applicable to fixed bias should be adhered to, not the value specified for cathode bias (so 300K max instead of 1M)