Oh ok.
I looked at the manual and somehow missed that picture showing the voltages.
So I'll leave the 4k resistor in place.
I looked at the manual and somehow missed that picture showing the voltages.
So I'll leave the 4k resistor in place.
Last edited:
Dynaco had to use a 22k resistor dropping 70V between points B and A because they were limited to a 20uF capacitor (one part of the cap 'can' they used).
You have access to much larger value capacitors, so you should be able to get away with 10k and a 47uF 450V capacitor and achieve an even lower F3. Heck, you could use a larger value cap, within reason.
Since you're starting with only 339VDC available, you can only afford to drop roughly 30V across the decoupling resistor to reach that same 305V B+.
Using a 10k ohm resistor, that would mean a total current draw of 3mA for both the pentode and the triode in the 6GH8A.
I'm sure the 6GH8A would work better biased a little hotter than that.
You could use a 4.7k ohm 1W decoupling resistor with a 100uF 400V decoupling cap, and that would allow the pentode and triode to draw 3mA each, which would be a good spot, I think.
Looking at the 6GH8A load lines from the GE tube data, it looks to me like this pentode-triode is different enough from 7199 that you can't just cut and paste the ST70 circuit and use a 6GH8A instead. You'll need to change resistor values to dial in better operating points.
The 6GH8A pentode section looks like it will be really current-starved with the ST70 resistor values. The situation would get worse with lower B+ voltage.
I'd like to see at least 3mA plate+screen running through the 6GH8A pentode.
The 6GH8A triode section looks to be biased on the cool side too. It would be nice to get 3mA running through that too. Closer to 5mA would be slightly better, but not a night-and-day difference. I'd just try to center bias it. Smaller value plate + cathode load resistors would probably get you there. Experimentation required.
--
You have access to much larger value capacitors, so you should be able to get away with 10k and a 47uF 450V capacitor and achieve an even lower F3. Heck, you could use a larger value cap, within reason.
Since you're starting with only 339VDC available, you can only afford to drop roughly 30V across the decoupling resistor to reach that same 305V B+.
Using a 10k ohm resistor, that would mean a total current draw of 3mA for both the pentode and the triode in the 6GH8A.
I'm sure the 6GH8A would work better biased a little hotter than that.
You could use a 4.7k ohm 1W decoupling resistor with a 100uF 400V decoupling cap, and that would allow the pentode and triode to draw 3mA each, which would be a good spot, I think.
Looking at the 6GH8A load lines from the GE tube data, it looks to me like this pentode-triode is different enough from 7199 that you can't just cut and paste the ST70 circuit and use a 6GH8A instead. You'll need to change resistor values to dial in better operating points.
The 6GH8A pentode section looks like it will be really current-starved with the ST70 resistor values. The situation would get worse with lower B+ voltage.
I'd like to see at least 3mA plate+screen running through the 6GH8A pentode.
The 6GH8A triode section looks to be biased on the cool side too. It would be nice to get 3mA running through that too. Closer to 5mA would be slightly better, but not a night-and-day difference. I'd just try to center bias it. Smaller value plate + cathode load resistors would probably get you there. Experimentation required.
--
I suppose the lower plate resistor would be better as it would knock down some of the extra gain that I don't need.
I did fire it up briefly using the 4k resistor and the output across the 4 ohm load seemed to clip more evenly and looked better right up until the point of clipping.
I did fire it up briefly using the 4k resistor and the output across the 4 ohm load seemed to clip more evenly and looked better right up until the point of clipping.
Last edited:
Please post voltages as follows:
6GH8A pin 1 (triode plate)
6GH8A pin 2 (pentode control grid)
6GH8A pin 3 (pentode screen grid)
6GH8A pin 6 (pentode plate)
6GH8A pin 7 (pentode cathode)
6GH8A pin 8 (triode cathode)
6GH8A pin 9 (triode control grid)
Also please post the B+ voltage going to the 'tops' of the plate load resistors.
Are you using the resistor values from the Dynaco schematic? If not, please post the parts values for:
Triode plate and cathode load resistors
Triode cathode bias resistor
Pentode plate load resistor
Pentode screen grid load resistor
Pentode cathode load resistor
Negative feedback resistor and where that's connected
With the above information we should be able to draw load lines and figure out exactly what's going on with the 6GH8A circuit.
6GH8A pin 1 (triode plate)
6GH8A pin 2 (pentode control grid)
6GH8A pin 3 (pentode screen grid)
6GH8A pin 6 (pentode plate)
6GH8A pin 7 (pentode cathode)
6GH8A pin 8 (triode cathode)
6GH8A pin 9 (triode control grid)
Also please post the B+ voltage going to the 'tops' of the plate load resistors.
Are you using the resistor values from the Dynaco schematic? If not, please post the parts values for:
Triode plate and cathode load resistors
Triode cathode bias resistor
Pentode plate load resistor
Pentode screen grid load resistor
Pentode cathode load resistor
Negative feedback resistor and where that's connected
With the above information we should be able to draw load lines and figure out exactly what's going on with the 6GH8A circuit.
OK, let's see what we have here...
The B+ is 324V. So far, so good.
Starting with the triode phase inverter, using good old Ohm's Law:
Continuing on to the pentode voltage amp stage:
The operating points you've chosen for your 6GH8A are as follows:
PENTODE
TRIODE
Now let's look at what the GE datasheet says are normal operating points for a 6GH8A:
Here are the load lines for the operating points you've chosen:
For the pentode section, reducing its screen grid voltage to only 33.4V reduces transconductance by a lot more than in the graph above (the graph is based on a screen grid voltage of 125V, and the plate is still current starved). This operating point severely limits both plate and screen current. Doing that makes it similar to using a really worn out 6GH8A (very low gm, higher rp).
I'm not sure why the 6GH8A triode plate curves aren't matching up with your 6GH8A. It could be that the data sheet isn't completely accurate, or it could be that the 6GH8A you've chosen is worn out. I don't know what's causing this behavior.
Both the pentode and the triode are biased very cold. Again, you can't cut and paste the component values for a 7199 to use with a 6GH8A. They are different tubes, and it seems the 6GH8A requires smaller value load resistors to get it to light up fully.
I'll see if I can whip up a simulation that provides us with better values for the load resistors for the pentode and triode. That will take a little time....
The B+ is 324V. So far, so good.
Starting with the triode phase inverter, using good old Ohm's Law:
- On the triode's cathode, 103.8V across 50k ohms means there's about 2.1mA plate current (Ip).
- 104.9V across 50k means there's about 2.1mA plate current there too. Remember Kirchoff's Law states that the current drawn by an device will appear the same at all points in its circuit. So this checks out.
- 103.8V at the triode's cathode minus 96.9V at the pentode's plate means the triode has -6.9V grid bias. Wait... That's a lot! I think those 50k load resistors could be reduced in value.
Continuing on to the pentode voltage amp stage:
- 324.3V minus 96.9V at the pentode's plate equals 227.4V dropped across 270k ohm plate load resistor.
- 227.4V divided by 270k ohms = 842uA (0.842mA) plate current. That's cold.
- Now for the screen grid... 324.3V minus 33.4V = 290.9V dropped across the screen grid load resistor (1M ohms).
- 290.9V divided by 1,000,000 ohms = 291uA (0.29mA). Again, that's really cold.
- The total current drawn by the plate and screen grid of the pentode equal 0.842mA + 0.291mA = 1.133mA. Yes, that's quite cold biased.
- The voltage at the pentode's cathode is only 0.754V. Grid current is likely. That would mean higher THD from the pentode.
- The total load resistance from the pentode's cathode to ground is 620 + 47 ohms = 667 ohms.
- 0.754V divided by 667 ohms = 1.13mA. That checks out perfectly.
The operating points you've chosen for your 6GH8A are as follows:
PENTODE
- Plate voltage = 96.9V
- Screen voltage = 33.4V
- Plate current = 842uA
- Screen current = 291uA
- Total plate + screen current = 1.13mA
- Grid bias = -754uV
TRIODE
- Plate-cathode voltage = 115.6V
- Plate/cathode current = 2.1mA
- Grid bias = -6.9V
Now let's look at what the GE datasheet says are normal operating points for a 6GH8A:
Here are the load lines for the operating points you've chosen:
For the pentode section, reducing its screen grid voltage to only 33.4V reduces transconductance by a lot more than in the graph above (the graph is based on a screen grid voltage of 125V, and the plate is still current starved). This operating point severely limits both plate and screen current. Doing that makes it similar to using a really worn out 6GH8A (very low gm, higher rp).
I'm not sure why the 6GH8A triode plate curves aren't matching up with your 6GH8A. It could be that the data sheet isn't completely accurate, or it could be that the 6GH8A you've chosen is worn out. I don't know what's causing this behavior.
Both the pentode and the triode are biased very cold. Again, you can't cut and paste the component values for a 7199 to use with a 6GH8A. They are different tubes, and it seems the 6GH8A requires smaller value load resistors to get it to light up fully.
I'll see if I can whip up a simulation that provides us with better values for the load resistors for the pentode and triode. That will take a little time....
Well, you can as long as the same supply voltages are provided. This is a proven tube to sub into an ST-70 with an adapter and the ST-70 Series II uses it with only mods to the FB line. All the loads are the same value. Just be VERY certain to keep the pin differences in mind between the two tubes when cloning a circuit based on the 7199.
Last edited:
An example of what you mean using the 7199 data here as an example.
I could always wire the tube up how it is in the Scott LK-72A.
Or I can recheck the compensation and get it right then leave well enough alone.
Van Alstine claimed that he had to re-bias the 6GH8A to replace the 7199. This goes way back to the late 1980s.
It looks to me like the 6GH8A triode is very different from the triode in the 7199. The 6GH8A triode has mu = 46, while the 7199 has mu = 17. Also, the 6GH8A has lower rp than the 7199 triode.
The pentode in the 6GH8A looks like it has similar gm to the 7199 pentode, but at a much lower plate voltage (220V for 7199, 125V for 6GH8A, both at 12mA Ip).
Here are the typical characteristics from the tube manuals:
The 7199 sheet shows operating points for low current, with Vp = 100V, Vg2 = 50V, Ip = 1.1mA, Ig2 = 350uA. gm is low at 1.5mA/V.
Perhaps 6GH8A works better under these very low current conditions.
It looks to me like 6GH8A is more similar to 6U8A than it is to 7199.
6U8A
6U8A is also a lot cheaper and easier to find than 7199.
--
It looks to me like the 6GH8A triode is very different from the triode in the 7199. The 6GH8A triode has mu = 46, while the 7199 has mu = 17. Also, the 6GH8A has lower rp than the 7199 triode.
The pentode in the 6GH8A looks like it has similar gm to the 7199 pentode, but at a much lower plate voltage (220V for 7199, 125V for 6GH8A, both at 12mA Ip).
Here are the typical characteristics from the tube manuals:
The 7199 sheet shows operating points for low current, with Vp = 100V, Vg2 = 50V, Ip = 1.1mA, Ig2 = 350uA. gm is low at 1.5mA/V.
Perhaps 6GH8A works better under these very low current conditions.
It looks to me like 6GH8A is more similar to 6U8A than it is to 7199.
6U8A
6U8A is also a lot cheaper and easier to find than 7199.
--
This is true but it isn't a significant factor in a cathodyne splitter since the gain is just close to unity. All the gain is from the pentode.
The pentode in the 6GH8A looks like it has similar gm to the 7199 pentode, but at a much lower plate voltage (220V for 7199, 125V for 6GH8A, both at 12mA Ip). The 6GH8A pentode looks closer to the pentode section in the 6U8A or the 6BL8.
I haven't checked out the 6AN8A yet. That was used in the Dyna MkIII amp (PP 6550A).
That looks like this, and it looks like it should work well enough.
Note that there is another RC decoupling network between the triode phase splitter and the pentode voltage amp that is not in the current circuit.
I got the voltages as close as possible to those on the Scott schematic. They're very close, certainly within +/-10%.
The gain is really high, at 84.5X.
I came up with this earlier, using a 6U8A model and operating points from load lines that looked good to me. This version has much lower distortion, but also much less gain, at around 50X.
THD for the LK72 circuit is 0.3% at 10V peak output, while the circuit below has THD of 0.042% at 10V peak output.
The THD is so much lower I'm afraid it might be a SPICE hallucination. But here it is anyway....
As you can see, the pentode is biased for higher plate and screen voltage and higher current operation (3.16mA instead of 1.3mA).
The pentode is AC coupled to the triode phase splitter. Otherwise the phase splitter is very much like the one in the LK72.
Hopefully there's information here that will prove useful. If not, well, my apologies.
--
I haven't checked out the 6AN8A yet. That was used in the Dyna MkIII amp (PP 6550A).
That looks like this, and it looks like it should work well enough.
Note that there is another RC decoupling network between the triode phase splitter and the pentode voltage amp that is not in the current circuit.
I got the voltages as close as possible to those on the Scott schematic. They're very close, certainly within +/-10%.
The gain is really high, at 84.5X.
I came up with this earlier, using a 6U8A model and operating points from load lines that looked good to me. This version has much lower distortion, but also much less gain, at around 50X.
THD for the LK72 circuit is 0.3% at 10V peak output, while the circuit below has THD of 0.042% at 10V peak output.
The THD is so much lower I'm afraid it might be a SPICE hallucination. But here it is anyway....
As you can see, the pentode is biased for higher plate and screen voltage and higher current operation (3.16mA instead of 1.3mA).
The pentode is AC coupled to the triode phase splitter. Otherwise the phase splitter is very much like the one in the LK72.
Hopefully there's information here that will prove useful. If not, well, my apologies.
--
That circuit is definitely worth trying.
The only thing I don't have is the 15k 2W resistors and the 68k 2W resistor, but I might can combine 1% resistors to make the value and wattage.
The only thing I don't have is the 15k 2W resistors and the 68k 2W resistor, but I might can combine 1% resistors to make the value and wattage.
Is there a missing stage decoupling resistor/cap for this circuit? The VA/splitter stage usually gets decoupled to avoid LF motorboating type instability. The other point to recocognize is the additional coupling cap creates a Williamson style driver section that tends to be more sensitive to phase shifting across the amp and the tendancy toward HF instability with GNFB applied over moderate amounts and quality of the OPT.
The GEC 912 plus used a similar approach only using a dual triode and needed no decoupling cap and resistor. I've used that circuit as well for a couple other amps with no issues.
Deleted my post because I missed the point, 20to20 was talking about the B+, and I agree, I always add an RC filter for the voltage amplification stage.
Yes, over the years it seems anyone who has posted for help beating a motorboating issue has nearly always had a basic circuit that omitted driver/splitter stage decoupling, along with whatever design factors that combined brought it out. Some work out, some don't.
Good criticisms, thank you.
This does not become a Williamson type driver because there is no differential driver stage after the split load PI. The Williamson is a 4-stage design, this 6V6 amp has 3 stages. The Williamson had to make do with very small capacitor values (8uF) which is not a limitation we have. We can buy 470uF 450V capacitors these days. A time constant is a time constant is a time constant. Motorboating problem solved (but a very long ramp up for the B+). Also, the Williamson applied 20dB of NFB, requiring a specially designed OPT with minimal phase shifts and wide bandwidth. I don't think that much NFB will be applied here, but I don't actually know for sure.
DC coupling the VA to the PI is going to be problematic with a B+ of less than 300V. The plate of the VA pentode will have to be at or below 100V, which leaves only about 100V plate-to-cathode across the PI triode.
Fix one problem and you'll make another. Everything is a balancing act of fixing problems. Which are the biggest problems? Which can be left alone? At what point is 'good enough' good enough?
Everything is a compromise.
If you can stand trying the circuit I came up with that doesn't have interstage decoupling network, then I think that one will have very low THD and enough gain. Otherwise...
The Scott LK72 circuit looks as good as anything I can come up with given the limitations imposed (DC-coupled, interstage decoupling network).
I'd say go with that.
This does not become a Williamson type driver because there is no differential driver stage after the split load PI. The Williamson is a 4-stage design, this 6V6 amp has 3 stages. The Williamson had to make do with very small capacitor values (8uF) which is not a limitation we have. We can buy 470uF 450V capacitors these days. A time constant is a time constant is a time constant. Motorboating problem solved (but a very long ramp up for the B+). Also, the Williamson applied 20dB of NFB, requiring a specially designed OPT with minimal phase shifts and wide bandwidth. I don't think that much NFB will be applied here, but I don't actually know for sure.
DC coupling the VA to the PI is going to be problematic with a B+ of less than 300V. The plate of the VA pentode will have to be at or below 100V, which leaves only about 100V plate-to-cathode across the PI triode.
Fix one problem and you'll make another. Everything is a balancing act of fixing problems. Which are the biggest problems? Which can be left alone? At what point is 'good enough' good enough?
Everything is a compromise.
If you can stand trying the circuit I came up with that doesn't have interstage decoupling network, then I think that one will have very low THD and enough gain. Otherwise...
The Scott LK72 circuit looks as good as anything I can come up with given the limitations imposed (DC-coupled, interstage decoupling network).
I'd say go with that.
Could that be why this circuit initially experienced a very low frequency oscillation that tended to come and go?
I'll definitely give that one a try.
What voltage should the 4.7uF cap be rated at and do I have to use that value or can I go larger without negatively affecting the circuit? I'd assume it needs to be rated at 450V.
Alao for the 270k resistor I don't have one, but do have a 75k and 200k for 275k. Will that work?
Last edited: