Here's the "napkin" drawing for the 6GK5 voltage gain block.
SY is correct about the I/P cap. not being essential, if the source signal is DC offset free. However, I prefer a carefully sized cap. at the I/P as protection against O/P trafo core saturation. The I/P cap. combines with the voltage amplifier's grid to ground resistor to form a high pass filter. When the O/P trafo is rated at 100 W., size the cap. to get a F3 of approx. 15 Hz. Move that "corner" freq. upwards, when the magnetic headroom in the O/P "iron" is smaller.
SY is correct about the I/P cap. not being essential, if the source signal is DC offset free. However, I prefer a carefully sized cap. at the I/P as protection against O/P trafo core saturation. The I/P cap. combines with the voltage amplifier's grid to ground resistor to form a high pass filter. When the O/P trafo is rated at 100 W., size the cap. to get a F3 of approx. 15 Hz. Move that "corner" freq. upwards, when the magnetic headroom in the O/P "iron" is smaller.
Attachments
Using 12.6v would require an even number of tubes in order to run their heaters in series pairs, (right?) So 6.3v would be more suitable (right?)
Do we have a choice? Those instructions sound rather crucial to me. (But then, I don't know much here.)
..Todd
Last edited:
Eli, would it make sense to also use a 10M45S and LED reference for the LTP tail on the 6GK5 version? If so, can you do another napkin drawing for that?
..Todd
The 10M45S is FET, not BJT, based. A LED is not needed. A single resistor sets the current. Look at the data sheet.
The LED in the 6GK5 gain block biases the cathode positive with respect to the grid.
The LED in the 6GK5 gain block biases the cathode positive with respect to the grid.
Looking at the datasheet won't help me understand electronics the way you do, unfortunately. I guess I'll leave the transistors there and forget about it.
..Todd
I didn't see this thread until yesterday since it started up while I was out of town and I didn't realize that it was about an amp design. I and another member of this forum started down the long journey to design a complete amplifier from scratch. This process is now 14 months long and neither one of us has completed the build (due to life's little diversions). I do however have a working prototype, and it has been tested and listened to with KT88's in it (regardless of the thread title). Our design is fully differential with LTP's in both the first and second stages. Mosfet drivers are used to allow AB2 operation. This gets me 75 watts in triode mode using KT88's on 450 volts. Triode mode sounds quite nice without any feedback. The very long thread is here :
http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp.html
Highlights:
KT-88 testing in posts #272, #310 and #319.
Chris's schematic is in post #268 and my driver board schematic is in post # 277
Just copy ours. We use it on both stages. The negative supply is not needed for the second stage, and you don't need it either. Just connect R12 and R19 to ground.
I think that your design will work, but I do see one mistake that I believe you are aware of. Move the G3 connection on the EF86 to the cathode. Pin 8 connects to pin 3, not pin 9
A question was asked somewhere in the middle of your thread about using a CCS in the output stage. The output stage can be made fully differential too just like our first two stages. I have tried it and it does work with two BIG restrictions. It only works in class A and grid current must be kept to a minimum (no A2). Either class AB (one tube driven to cutoff) or the presence of significant grid current will unbalance the differential pair leading to distortion and recovery issues. This makes your monster amp into a BIG but very nice sounding 10 watt amp!
http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp.html
Highlights:
KT-88 testing in posts #272, #310 and #319.
Chris's schematic is in post #268 and my driver board schematic is in post # 277
Just copy ours. We use it on both stages. The negative supply is not needed for the second stage, and you don't need it either. Just connect R12 and R19 to ground.
I think that your design will work, but I do see one mistake that I believe you are aware of. Move the G3 connection on the EF86 to the cathode. Pin 8 connects to pin 3, not pin 9
A question was asked somewhere in the middle of your thread about using a CCS in the output stage. The output stage can be made fully differential too just like our first two stages. I have tried it and it does work with two BIG restrictions. It only works in class A and grid current must be kept to a minimum (no A2). Either class AB (one tube driven to cutoff) or the presence of significant grid current will unbalance the differential pair leading to distortion and recovery issues. This makes your monster amp into a BIG but very nice sounding 10 watt amp!
Brilliant! Never thought of it that way, but the R value(s) are about right...
Consider please mine as well: plain linear frequency independent 27K resistor.
If you guys are expecting me to do arithmetic, you'll just have to get used to correcting me. (I have an arts education for a reason! Or maybe I should say I don't have a science/math/electronics education for a reason!) 
Having said that, I calculated the filament bias voltage divider for a stand-alone 6.3 vct transformer. (see attached diagram). It should bias it to 60v at 4 mA if my crappy math and dubious comprehension is by some miracle correct.
Okay, let me have it.
..Todd
Having said that, I calculated the filament bias voltage divider for a stand-alone 6.3 vct transformer. (see attached diagram). It should bias it to 60v at 4 mA if my crappy math and dubious comprehension is by some miracle correct.
Okay, let me have it.
..Todd
Attachments
Last edited:
I calculated the filament bias voltage divider for a stand-alone 6.3 vct transformer. (see attached diagram). It should bias it to 60v at 4 mA if my crappy math and dubious comprehension is by some miracle correct.
Okay, let me have it.
..Todd
Taj: At 4 ma, the voltage divider will run a tad hot
....The 100k R will be dissipating 1.6 W which will require a 3W or 5W R for some margin. If you double the values to 200K & 30K, the divider will pass 2ma and the 200K will dissipate .8W.Keep in mind that the divider is just a reference voltage, as does not have to pass any current.
Also, per Wavebourn's suggestion, you can replace the 220R bleed resistor in the power supply with the voltage divider, and simplify the design, since it will serve double duty as the bleeder and voltage divider.
We could get the dissipated power down and the total resistance down by tapping off of the 240V B+.....and the divider could still be used a the bleeder. Off of the 240V B+ the R values are 75K and 27K for a heater bias voltage of 62V with the 75K dissipating .4W. Don't forget the cap across the lower R.
Last edited:
In theory, that's true, and I have also found that to be the case. However, at least one of my friends has seen leakage currents that could upset the string and recommends using emitter followers to drive that voltage (see "Valve Amplifiers," 3rd edition, the THINGY). Personally, I think that's extreme, and if any tube I used had that much leakage, I'd replace it.
Yeah, but resistors are cheap and your life isn't. Ever since I got zapped because of an open bleeder resistor, all of my amps have two bleeder paths. One will be on the input side of the choke(s) to discharge the input cap if the choke goes open or is disconnected. The other will be the heater bias divider which I place downstream somewhere, usually across the biggest cap. Use 3 watt resistors too, less chance of it going open.
Thanks! Never thought of that either.........I'm going to add another bleeder R upstream of the choke on my present project when I get home tonite.
Resistors don't replace good habits and safety practice. I never touch capacitors I did not discharge by resistors, even when I know that bleeders are there.
However, redundancy is a good thing, but such a way we should think about reserves of other things as well.
How about a smart crowbar?
Honestly well done and connected one bleeder is enough, if we concentrate all the current through the bleeder which is going to work as a voltage divider too we can make it a stiffer point for the filament circuit as we would need lower resistor values here.
But for safety measures during build and tweak we could solder additional bleeder resistor directly on to each cap which can be removed in the finalised design.
EDIT: Now I see Waveborun already addressed the issue and I would tend to agree.
Cheers Michael
Last edited: