Jonny Tubetronics: amps and synth stuff now in one thread

I did some calcs in a spreadsheet and both 30k and 15k will be fine, so I ordered it up and here it is:


Duplo block for size reference, which is 32mm square I think. This thing isn't small! Both sides seem to track pretty closely, within a couple of percent, which should be good enough for what I need. Curiously, it's got four connections per track. It's got the usual three connections plus what appears to be a centre tap. This had me confused for a while. Pics here mostly for my reference. Shaft turned fully to the left and these two read zero (or near enough):



Shaft turned fully to the right and these two read zero:



So the tag I've got the red probe attached to above is the wiper.

Then the far left hand tag reads 15k at the ends and zero in the middle, which suggests it's a centre tap. Not seen that before:






Now to answer the question about whether it's going to act linearly or logarithmically, I did a plot in Excel to see. Divided the resistance range by 10 and plotted it against frequency. Turns out it's going to be logarithmic:

I can't remember where I read if you put a fixed resistor in parallel with a linear potentiometer it acts slightly logarithmically, but when I plotted it on a graph it didn't appear to. Not enough to straighten out the frequency response, anyway. Probably because it's doing a log to base 2 instead of base 10 or something, but either way it's something I should probably remember from GCSE maths log tables. So I probably won't bother and I'll just put up with a logarithmic slope on my dial.

I saw those ones on eBay from that seller with the green monsters! It’s the same model pot, so the connections are the same by the looks of it. Mine doesn’t have that sticker on it.

That’s pretty much what I gleaned from the internet after a quick look. I’m guessing my one might be 1960s going on the general construction and the waterslide decal. It’s definitely mega high quality, one of those things that feels carefully put together by hand in a way that doesn’t exist any more. The terminal screws are soldered to the terminals after tightening and the long screws that hold the stack together are lacquered to prevent them coming undone. I know from the things that we buy at work, if you want a modern equivalent you have to pay £££££ for quality like this in electronics. £8.50 delivered for a 60 year old NOS component really is a bit of a bargain in comparison and will outlive anything we use at work (and probably/hopefully me).

Not had much time to spend on this lately. Finally got a few minutes to recap and check my maths. I’m going to breadboard the Wireless World worked example to start with so I can play with it, plus there’s a few things I don’t completely understand and I’ll be able to tinker, then modify it from there by swapping the components in the timing circuit from the fixed 50hz to my variable frequency.

There’s something else I want to try too. The original worked example used an EF86 pentode (usually used as an audio amplifier) and an EF91 sharp cutoff pentode (used in old TVs). They gave ways to select the components to suit any pentode, but they didn’t say why they used an EF91 and not just two EF86s. There’s nothing wrong with the EF91, I’m just trying to get my head around the differences between the different pentodes. I’ve got some pentodes below which should be a direct swap: Russian EF86 on the left, EF91, the EF80 superseded the EF91, and then the EF184 superseded the EF80. I think they basically worked the same, but were improved so they could be driven harder as televisions evolved.

They look like RS wire markers, used them a lot offshore, colours make them easy to identify if the numbers have gone, used to know the number/colour off by heart

Much of this has been ticking away in the background. I've been thinking how I might mount up the V2 of the amplifier and I'd quite like to have it on a PCB this time. I haven't come across a PCB for the Mullard or BBC designs with the EF86 pentode input stages (like I built), but there are a number of boards available with an ECC83 dual triode as the input stage. So a 3-valve design for the stereo pair. I actually have a Tubelab SSE board which is a 3-valve design with the bigger EL34 output pentode for about a 10W output, but when I looked at the transformers I decided it was prohibitively expensive. I'd be looking at minimum £450+ for transformers and I can't justify that.

I looked at modifying the components on the Tubelab SSE board to suit the smaller EL84 output pentode. It would work, but one of the disadvantages is the layout on the board. Because it's designed for the bigger valves, it's a bit spread out. And I'd have to mount the valves off the board to convert from an the EL34's octal base to a B9A base for the EL84. The advantage though is it's probably got a good layout for noise, cooling etc and it's gold plated. I then had a look at the Chinese eBay offerings and decided to have a closer look at one since it was about £10 delivered.

So, as expected, it's not the same as the one described. The board is actually nicer looking in purple instead of yellow. The circuit is very slightly different and when I traced it back over the circuit diagram I found a couple of improvements that I would have made. So that's a good thing and all I have to do is check and tweak a couple of component values to suit how I want it.



Board in the eBay description for comparison:



Nixie clock is on hold for the moment. I've got a backlog of projects and a couple of things ought to come first. They were the right tubes at the right price and take a bit of scouring and watching on eBay, so I snapped them up at the time ready for use later on.

Added a simple negative feedback loop from the output directly to the input with a resistor inline:



I didn’t have any resistors in convenient decade values, so I had to go for what I had closest in multiples of 10. So that was 8.2M, 820k, 82k and 8.2k. They all worked as expected and attenuated the output voltage. I’ve got a nice list of numbers to play with now.

Now I know the voltage gain in a generic circuit is roughly this for the following valves: EF184 RF pentode 285, EF80 RF pentode 211, EF91 sharp cut-off pentode 218, EF86 AF pentode 149. The gain on the AF pentodes are probably lower to make them less microphonic. Audio triodes have a gain of about 70 I think, and they're said to be better in guitar amps because they're less microphonic than pentodes.

Then I did some frequency sweeps to look at the frequency response. Did it on all three valves and plotted them in Excel with the output voltages normalised:





This was where I got a surprise. I was expecting them to have a pretty much flat response, but it’s not so. The bit I don’t fully understand yet is how the response gets flattened in an audio amplifier. I put my negative feedback loop from the output to the input. You can also run the feedback partway into the cathode, which I guess changes the bias with respect to the frequency response. I think it might be this. More experimenting needed…

The other interesting thing was almost no difference in response between the RF, sharp cut-off and AF pentodes. I was expecting them to be different.

Final interesting thing was what happens to the lower response when you change the cathode capacitor:



Loads to take in there. I’ve learned a lot.

Neat and breadboard aren't normally seen together in the same sentence

I got rich to merge the amps and synth threads into one since there was starting to be a lot of overlap and I was losing track of what was in which thread. Thanks for that

When I get 5 I'll edit the first post to include the synth stuff.

----------------------------------------------------------------

I got the circuit working with a thermistor:



So it's not exactly functioning as per the original design said it would. Obviously I'm not using the original thermistor, as that type is long discontinued. That means the rate of change of resistance is probably not exactly as per the original one would have been. However, my circuit does seem to want a higher value resistance than the original circuit said it would want. That is probably a combination of tube age / variation from OEM spec and maybe the fact that I'm using 30-40 year old capacitors which might have drifted out of spec. Either way, it still works with a bit of tweaking so I'm happy.

The way I built it with the potentiometer, it needed around 5.87k to hold 16V output. Increasing the output to clipping (~30V) meant increasing the resistance to 5.92k (so not very much of a change). Reducing the resistance until the oscillations stop meant reducing it to 5.84k (again, not very much of a change). So it was very sensitive to the resistance and very sensitive to changes. With the thermistor inserted, I needed a resistor in series to increase the overall resistance anyway, so I used a 5k pot. Starting at 1k, I increased the resistance until I got the original 22V output. It is nowhere near as sensitive now and much bigger changes to the potentiometer make almost no difference because the thermistor compensates. Now I'm interested to see what happens to the output when the timing circuit is altered.