On the top shelf of my workshop for probably the last 20 years has been a little black box with a couple of valves on the top.
You might think it is the original valve based version of The Internet, but it isn't.
You might think this is one of many old projects gathering dust. And indeed it is, but there is a difference. You might just be able to see there is a warm glow coming from the tubes.
That is because this is in use, and has been in use pretty much daily all the time it has been there.
I built this 20-25 years ago. I can't remember when specifically, before the blog, before Tynemouth Software even. I didn't have a digital camera at that point, and there is no email trail for the parts as it was well before I got my gmail account (2007?).
I did think I might get a result from some of the valves I bought from Rapid Electronics (yes, they used to sell valves!) but my order history does not go back that far either.
I can't find my notebooks from back then. I am pretty certain I know where they are, just there are rather a lot of things between me and them.
So I thought it might be interesting to reverse engineer it. See how bad it is, and if there is anything I would do differently today.
Step 1 was to clean it up a bit.
You might notice it is resting on a bit of cardboard, that is because the feet have dissolved. There seem to be a problem with rubberised plastic from that era, things that had that sort of coating break down and go sticky over time (e.g. the old sky remote controls)
With that cleaned up a bit, and some new plastic feet fitted, it looks a lot better.
It is a very minimal design. A power switch on the left and a volume control on the right. With a good old chicken head knob like every valve amp should have.
The back has a fixed mains lead, two phono jacks for the input and four 4mm jacks for the speaker outputs (somewhat faded in a quarter of a century of sunlight).
I managed to round one of the screws that was quite persistently stuck, so I had to drill that out.(well, first I had to find my drill, that I had conveniently ran flat last time I had used it, then find the charger for it, then wait for it to charge......)
That's better, all six screws removed.
One of these things is not like the others.
Time to open it up.
But first, a couple of things.
Main voltages and higher are present in here. They can be lethal. If you ever do anything like this, be sensible, take the usual precautions etc.
Secondly, the design is wrong. I know that. I knew that at the time. The transformers and wrong, the valves are wrong. I could have built one with the right parts (and indeed I did start that), but they are very expensive, and not easily available. So, I thought I would see what I could do with parts that were not ideal, but were cheap and easily available. (I am sure that will not stop people on the internet telling me that it is wrong in even more ways that I realise).
OK, those of a nervous disposition, look away now.
OK, there is a lot going on there. Let me reverse engineer my own 25 year old design.
One moment please.
Let's look at that in more detail.
I'll start with the power. This is the first bodge. A proper valve amp would have a proper power transformer with a high voltage winding for the B+ and a low voltage winding for the tube filaments.
I didn't have that, so I improvised with two normal mains transformers.
The mains comes in via a toggle switch and goes to the primary of the first transformer. This has two 15V secondaries.
B+ High Voltage Supply
The first of these is wired to the secondary of a second transformer. This is two 6V windings in series to give a 12V winding. The primary of that becomes the output, one of the 120V is used.
I have redrawn that in simplified form.
The first transformer is designed for 240V to 15V, so a 16:1 windings ratio. The second is designed as 240V to 12V, so 20:1, but in reverse, so 1:20, and only one of the output windings is used, so 1:10.
When arranged like this, you get a single transformer with a windings ration of 16:10, or 8:5.
So with 240V in, there should be approximately 150V AC output. This is not very efficient, but it does give a high voltage, low current supply, and more importantly, isolation from the mains.
(N.B. I am using a nominal 240V here as these transformers are marked as having dual 120V primaries and it makes the maths neater. The UK mains voltage is more like 230V these days, so the output would be 143.75V not 150V)
The high voltage is rectified with a little bridge rectifier module and smoothed with a 47µF capacitor. It is further filtered with a 100Ω resistor and a 100µF capacitor. In circuit, that gives a measured B+ of around 200V in use. (143.75 * √2 is 203.29, so I am happy that the maths seems to work out, given the tolerances of the transformers and the losses in the inefficient design).
I have redrawn that section using this simplified view of the transformer.
In value amps you often see a series of these R/C filters added, with voltages tapped at each stage to feed different parts of the circuit. I thought I would have taken the output transformer feeds from the first capacitor and the triodes from the second, but it appears I have taken all the supplies from the second one. Not sure why. I am sure I prototyped all of this with crocodile clip leads and would have moved things around until I was happy.
It seems the last thing I was testing out were some EM80 "magic eye" type values for a VU meter project that never happened. But that is the sort of crocodile clip lead spaghetti that would have been involved.
Heater Filament Supply
The second 15V secondary on the mains transformer goes to a second bridge rectifier and a 2200µF smoothing capacitor. This voltage rail is regulated by a 7812 linear regulator, with a 1N4001 diode in it's ground wire, giving an output voltage of 12.6V, perfect for the heater filaments of a pair of ECC82 / 12AU7 valves.(Those valves have two 6.3V filaments that can be wired in parallel or series for 6.3V or 12.6V operation)
I was surprised to see the 7812 in there, I did not remember adding that, but I guess it makes sense to regulate that supply. I think I would have planned to end up with a 6V or 12V AC winding that I could use directly, but this works and gives a clean regulated DC supply for the filaments.
The amplifier circuit is constructed directly on the valve bases. These are mounted in the lid. Normally with a case like that I would have used it the other way up with the lid (and the mounting screws) as the base to give a neater appearance. However here is clearly makes sense to have access to the sockets directly and the big transformers bolted to the base.
Valve circuits can be beautifully simple, this one is pretty minimal. Here are more bodges. These are not really output tubes. The transformers I used are not really suitable for audio frequency. You know the story by now.
I am using a pair of ECC82 / 12AU7 valves. One per channel. These are dual triodes, a pair of identical voltage controlled amplifiers in each tube. These act a bit like FETs. The voltage at the grid controls how much current flows between the cathode and the anode. Google it, I am sure there are many better explanations than that.
There is very low input impedance, which make this an ideal input stage. There is a 1M resistor to ground just to stop it floating. The first triode takes the input signal and amplifies it. The ECC82 has a gain of around 17, but there will be loads of headroom from a 1V peak to peak signal with a 200V supply.
This is capacitively coupled to the one half of the volume control. The output of which goes to the grid of the second triode, where the output is generated across the output transformer.
Yes, you guessed it, another mains transformer. This 240V-6V, 40:1 ratio drops down the very high voltage, very low current output to something low voltage, higher current, suitable for driving a speaker.
Here it is redrawn with the transformer simplified.
A mains transformer is completely the wrong thing to use here. They are not designed for audio frequency use. There are also potentially problems with the placement and proximity of all the transformers which could lead to coupling between them.
The right hand channel is, as you might have guessed, identical (but I drew it anyway)
Testing it out
Using the signal generator in my scope, I can feed in a 1KHz sine wave, and the output looks decent enough. You have to be very careful with things like this as the scope probe ground is connected to mains earth, as are parts of the amplifier circuit. In this case, I was measuring the input (which was grounded to mains earth at both ends) and the speaker output (which was isolated).
If I read that right, yellow is showing the input signal (I didn't disable the 10:1, so it is 1V p-p not 10V), and green is the ouput, looks to be about 500mV. Into a 4Ω speaker that would be 62mW? (I am going to pretend I didn't see that, it sounds loud enough to me.)
I don't have the necessary bits to do a frequency response plot. (I think the scope could do it, but it's one of the things knobbled by a license that I don't have) I expect the response would drop off at higher frequencies due to the properties of the wrong transformers I used. The frequency response of my ears is probably no better these days.
It is not very powerful, but it's fine for driving a couple of 1970s hi-fi speakers in a small room. The volume is usually set between 50% and 100%, I have not needed it to be any higher than that.
I built it as an experiment, to see if I could do it and how well it would work. I didn't have a use for it initially, but when the hi-fi amp in my workshop broke, I swapped this in as a temporary replacement. That must have been about 20 years ago.
Valve circuits can be incredibly simple and incredibly reliable. I use it because it just sits there and works. I have a big switch by the door that cuts the power to the whole workshop when I leave, so this amplifier is on, and in use, any time I am in there.
It sounds fine to me, to listen to music or audio books when I am soldering, played from MP3 files by a Raspberry Pi that is wired to various amplifiers around the house. It is not the audiophile valve amplifier experience, that has been nullified by the use of the wrong parts and the signal source.
I did start to make a better one, with a pair of EL84s per channel, one that I would listen to my vinyl records on my 1969 Goldring Lenco GL75 turntable. I got as far as drilling out a chassis, but I never got around to ordering the correct transformers, as they were so expensive. Any interest in my continuing with that? Any generous donors or anyone with some suitable transformers they no longer need?
Would I do anything different today? well, no, not with those parts. There are things I could improve if I wanted to make a better amplifier, but for this application it is fine.
- I could tie the centre tap of the two filaments to ground via a 10K resistor to maybe reduce noise.
- I could probably have done the wiring more neatly, but no one has had to look at it for 25 years, so it has not been a problem.
- I could use a single ECC83 as the input stage for both channels and get 100x gain, and use a single ECC82 with it's lower gain / higher power as the output stages?
- I could create a second B+ using the other 120V tap on the second transformer to give left and right isolated supplies.
- I don't know if it would help to bypass the grid of the second triode with a capacitor. I have that on the input stage.
So, 25 year service is wiping off the dust, replacing the stick on feet on the bottom of the case and replacing a rounded screw. Here's to another 25 years.
Hope you enjoyed something different for a change.
I normally work on old computers, or new things for old computers, such as these:
ZX80, ZX81, Commodore PET and Jupiter Ace compatible computer kits
You can support me via Patreon, and get access to advance previews of blog posts and behind the scenes updates. This also includes access to my Patreon only Discord server for even more regular updates.