I bought a Knight PA amplifier at an estate sale awhile back and found two Bugle Boy EL34s in it. They test around 80%. I also bought a stash of type 36 tubes recently from a guy. This week I designed a amplifier for them. I've never had an amp with a tetrode or pentode gain stage. I have two questions. What do you think about the design before I build it? Does anyone have experience with a 36 tube? Is it worth building, or a waste of time? If it turns out crap, I'll replace the gain stage with a 6Q7G or something. Have a look and let me know.
Last edited:
My two cents:
A link to the Tung-Sol datasheet for the 36 (which is a tetrode):
36 Tung-Sol
The maximum screen grid voltage is 90 V and the maximum screen grid current is 1.7 mA. I think that at least one of these maximums will be violated in your schematic.
The screen resistor in your schematic is 62K. In order to stay below the maximum screen grid voltage of 90 V, the screen grid current would have to be (290 - 90) / 62K = 3,2 mA, which is way over the maximum screen grid current of 1.7 mA.
But with the maximum screen grid current of 1.7 mA, the voltage drop over the screen grid resistor would be 0.0017 x 62K = 105.4 V. This would result in a screen grid voltage of 290 - 105.4 = 184.6 V, which is way over the maximum screen grid voltage of 90 V.
Besides the above: Even with the lowish plate resistor of 22K, the gain will be rather high since there is no form of negative feedback being applied to the 36. So the sensitivity of the amplifier will be high, and much higher than it has to be when you plan to use it with line level sources.
A link to the Tung-Sol datasheet for the 36 (which is a tetrode):
36 Tung-Sol
The maximum screen grid voltage is 90 V and the maximum screen grid current is 1.7 mA. I think that at least one of these maximums will be violated in your schematic.
The screen resistor in your schematic is 62K. In order to stay below the maximum screen grid voltage of 90 V, the screen grid current would have to be (290 - 90) / 62K = 3,2 mA, which is way over the maximum screen grid current of 1.7 mA.
But with the maximum screen grid current of 1.7 mA, the voltage drop over the screen grid resistor would be 0.0017 x 62K = 105.4 V. This would result in a screen grid voltage of 290 - 105.4 = 184.6 V, which is way over the maximum screen grid voltage of 90 V.
Besides the above: Even with the lowish plate resistor of 22K, the gain will be rather high since there is no form of negative feedback being applied to the 36. So the sensitivity of the amplifier will be high, and much higher than it has to be when you plan to use it with line level sources.
You are correct about the screen grid resistor. I'm not sure how I made that mistake. I was calculating to target 1.6mA and somehow ended up at twice that number. I just looked at my notes and sure enough had 125K marked down earlier. Thanks for pointing that out, I'll correct it in the schematic. BTW, Do you have any thoughts on a 36 tetrode gain stage? Also, I calculated the gain to be 24 with a 22K load. That shouldn't be too bad. I actually looked at larger loads but the the gain was too high with those. The lower load resistor value also makes the impedance look better. Depending on how it works I could always try a 20K or 18K load resistor. I corrected the drawing and attached the updated version using 127K dropping resistor.
I had some time today and did some looking. It turns out that a builder that I respect greatly, Bartola Valves, has used the 24a tetrode to use as a driver stage. He put up three or four posts regarding it. https://www.bartola.co.uk/valves/tag/24a-tube/
The 24a is very similar to the 36, although the 36 looks a little better to me and uses 6.3V heaters. Ale says that his driver sounded quite good. I guess that the 36 might just be a good idea. Check out his website and have a look at his schematic.
The 24a is very similar to the 36, although the 36 looks a little better to me and uses 6.3V heaters. Ale says that his driver sounded quite good. I guess that the 36 might just be a good idea. Check out his website and have a look at his schematic.
Actually, I noticed that the Bartola design doesn't use a grid stopper at all. So I probably can. I used the equation Rg=1÷[2×PI×fc×Cgk].
I used 1Mhz for the cutoff frequency and 3.7pF for Cgk. That is 43K. Is that wrong? It could be but I thought it was that equation??
In fact, it seems that the 24a and 36 tubes are nearly identical save the heaters. You can basically say the 36 is a 24a with 6.3 volt heaters. I bet you could drop it into Bartola's circuit, as is.
I used 1Mhz for the cutoff frequency and 3.7pF for Cgk. That is 43K. Is that wrong? It could be but I thought it was that equation??
In fact, it seems that the 24a and 36 tubes are nearly identical save the heaters. You can basically say the 36 is a 24a with 6.3 volt heaters. I bet you could drop it into Bartola's circuit, as is.
To calculate a gridstopper resistor value for a pentode or tetrode tube, you need to consider the tube's grid-to-cathode capacitance (Cgk) and the frequencies you're working with.
The basic formula for the gridstopper resistor is:
R = 1 / (2π × f × Cgk)
Where:
For your tube with Cgk = 3.7pF (3.7 × 10⁻¹² F):
Typical calculations:
For audio applications (rolloff around 100 kHz):
R = 1 / (2π × 100,000 × 3.7 × 10⁻¹²) ≈ 430kΩ
For RF applications (rolloff around 10 MHz):
R = 1 / (2π × 10,000,000 × 3.7 × 10⁻¹²) ≈ 4.3kΩ
Practical considerations:
For general-purpose audio tube circuits, values between 10kΩ and 47kΩ are common starting points. You can then adjust based on stability testing and frequency response measurements.
per Claude AI. That's what I was working from. Keep in mind that the top cap is the grid. What I did with my 85 gain stage was build the top cap cable with shielded wire and placed the grid stopper right at the top cap and covered it with three layers of heat shrink. This was because when I would walk up to the amp with my phone, you could hear something from the amp. With the grid stopper at the cap, it is dead silent.
The basic formula for the gridstopper resistor is:
R = 1 / (2π × f × Cgk)
Where:
- R = gridstopper resistance (ohms)
- f = frequency where you want the RC rolloff (Hz)
- Cgk = grid-to-cathode capacitance (farads)
For your tube with Cgk = 3.7pF (3.7 × 10⁻¹² F):
Typical calculations:
For audio applications (rolloff around 100 kHz):
R = 1 / (2π × 100,000 × 3.7 × 10⁻¹²) ≈ 430kΩ
For RF applications (rolloff around 10 MHz):
R = 1 / (2π × 10,000,000 × 3.7 × 10⁻¹²) ≈ 4.3kΩ
Practical considerations:
- Most gridstopper resistors range from 1kΩ to 100kΩ
- Higher values provide better RF suppression but may affect high-frequency response
- The resistor should be placed as close as possible to the tube socket
- Use a non-inductive resistor type (carbon composition or metal film)
- Consider the circuit's input impedance requirements
For general-purpose audio tube circuits, values between 10kΩ and 47kΩ are common starting points. You can then adjust based on stability testing and frequency response measurements.
per Claude AI. That's what I was working from. Keep in mind that the top cap is the grid. What I did with my 85 gain stage was build the top cap cable with shielded wire and placed the grid stopper right at the top cap and covered it with three layers of heat shrink. This was because when I would walk up to the amp with my phone, you could hear something from the amp. With the grid stopper at the cap, it is dead silent.
Last edited:
What I wrote about the gain being high was wrong. The gain will be roughly mutual conductance times plate resistance so your number of 24 seems right to me.
I do not have experience with a tetrode as a driver. The pentodes I used as drivers (EF6 and EF86) were connected as triodes.
Good luck with your project!
Here is a link describing the calculation for grid stoppers ... Valve Wizard Grid Stoppers and Miller Capacitance
The calculaton uses Cin, which is calculation as Cin = Cgk + Cga(1 + A), where A is the gain of the tube. Cga is small at 0.01pF, but the gain is about 300, so it doubles Cin.
I just feel it is dangerous quoting those AI sources because there is nothing to back up the reasoning. I suppose in many cases Cga is negligible, but that calculation above should use Cin, and state that Cga can normally be ignore for a pentode.
The calculaton uses Cin, which is calculation as Cin = Cgk + Cga(1 + A), where A is the gain of the tube. Cga is small at 0.01pF, but the gain is about 300, so it doubles Cin.
I just feel it is dangerous quoting those AI sources because there is nothing to back up the reasoning. I suppose in many cases Cga is negligible, but that calculation above should use Cin, and state that Cga can normally be ignore for a pentode.
OldHector, thanks for that. I read the web page. I notice that he uses the actual gain of 60 for the ECC83 example. In my case that is 24. Valve Wizard also uses 20Khz for the cutoff frequency. If I use his 20Khz, I would need a 2M grid stopper. That won't work . If I use 100Khz, I need 410K ohm and fc = 1M, 41k. So, what do I do? I think I'll use the 43K. If 34K is fine for a guitar signal, then 43k should be fine for a 2.5V pre-amp input. I can always drop a 1K5 in later to see how it works. I also get your point about the AI answer. Thanks.
Here is an updated schematic with the screen grid dropping resistor value changed on the load line slide. I also included some information regarding the 24-A amplifier that Bartola Valves posted.
What you need is data on the 36's g2 current at the operating point to determine the dropping R from B+. Without that, perhaps assume it will be half the max as a starting point and adjust it while tuning.
Douglas
Douglas
Instead of the double RC sections in the power supply you might do just as well or even better with a single LC section. Even better is a double LC section. You can use a much smaller L for the second inductor and you will get essentially no residual ripple. Takes a little tuning to get a step response without ringing but PSUD will help with that. A SE amp really like a quiet power supply since there isn't any PP cancellation in the output.
Douglas, thanks for looking. The datasheet does give a screen current for a 250v operating point. I'm at 220v if I remember correctly. I'm guessing it will be close to that. With the 127K resistor, I'll be at 1.6mA. That'll be close enough, I think. I'm still thinking about using a 91V zener to set the voltage and drop a bit more with the grid stopper.