I think if 10KHz and 1KHz are the same the transformer is probably OK. If your getting 3.5W and 3% THD that's not bad. Anything else will require something other than an EL34 and more current.
I think the 6SN7s cathode resistors are too low - surely cathode needs to be a bit higher than 1.15v for a 1v rms input signal?
Don't forget that the negative feedback will also take the cathode voltage down as you hit the bottom of the waveform. In fact it goes negative maintaining the valve in conduction. Without the NFB you can only apply .25v before clipping.
Don't forget that the negative feedback will also take the cathode voltage down as you hit the bottom of the waveform. In fact it goes negative maintaining the valve in conduction. Without the NFB you can only apply .25v before clipping.
C7 is the cathode bypass capacitor for the input cascode.
If you remove C7 that degenerates the stage, meaning it adds current feedback (for AC signal), which
- reduces the gain from the cascode stage
- increases the plate resistance of the cascode stage
So, long story short, if you remove C7 the 6SN7 has lower gain, so there's less negative feedback around the amp, which will probably result in higher distortion at the output. It would also require higher input voltage to reach full power (will lower the amp's sensitivity).
If you put back C7 across R10, the 6SN7 cascode will then have more gain, which should result in more gNFB applied around the amp, which should lower distortion.
But that's all 'should' stuff.
Also, 1.9mA plate current is awfully low for a 6SN7. One usually sees at least 3mA, and >8mA is where many like to run it for lowest distortion, etc. Perhaps all that doesn't matter when using it in a cascode arrangement.
EDIT TO ADD: I just had a thought... Was this cascode stage designed for use with 6SL7?
Hmmm.... Scratch that. Rk of only 510R wouldn't be used with a 6SL7.
Last edited:
Sorry YES keep the bypass cap on the 6SN7. Try removing the bypass cap on the EL34. For what its worth I don't think its a bad circuit.
Normally removing the cathode bypass cap (Ck) from a power output tube is not desirable, because it increases the plate resistance (rp) of the output tube, which degrades bass response (the higher rp into the primary inductance forms a high-pass filter). But if this is a tweeter-only amp, then that's not a concern. However, for a full range amp that would be a concern, and you'd normally keep Ck in place.
I woudn't be surprised if it was originally designed to use a 6SL7.
I also felt that the cathode resistor should be higher (1.5k) and the current needs increasing to 5ma. I can adjust it easily when using a current regulator.
An Ixys 10M45 may be a better choice than the LR8.
The other problem is that 6SN7 likes higher plate voltages than what we're seeing here. Only 55V at the bottom tube's plate? That means there's only about 55V plate-cathode (Vp-k) for that 6SN7 triode. If you look at a 6SN7 load line, that's way down at the left side, with very little voltage swing available. That is not what you want for a power amp driver stage.
Perhaps the 6SN7 voltage divider needs to be adjusted so that each triode gets 33% of the B+. Since the B+ is about 300V to the first stage, the plate of U2 should be at 200V, the plate of U3 at 100V, and 100V dropped across the plate load resistor R5.
If R5 was 47k, 100V dropped across it would be 2.1mA, which is about where the 6SN7 is biased now.
I'd have to look at the 6SN7 load lines to see what Vg should be for Vp = 100V and Ip = 2.1mA.
If that's -2V, then using a 1k resistor for R10 would yield about -2V bias.
Perhaps the 6SN7 voltage divider needs to be adjusted so that each triode gets 33% of the B+. Since the B+ is about 300V to the first stage, the plate of U2 should be at 200V, the plate of U3 at 100V, and 100V dropped across the plate load resistor R5.
If R5 was 47k, 100V dropped across it would be 2.1mA, which is about where the 6SN7 is biased now.
I'd have to look at the 6SN7 load lines to see what Vg should be for Vp = 100V and Ip = 2.1mA.
If that's -2V, then using a 1k resistor for R10 would yield about -2V bias.
Last edited:
If the 6SL7 is run at maybe 750uA plate current, then that would work. The one problem could be slew limiting with such low plate current sunk into the EL34 grid capacitance -- especially if run triode wired.
Also, 6SL7 likes high plate-cathode voltage (Vp-k). It's usually run with >130Vp-k. This amp has B+ of only about 300V to play with, which makes it tough to get the cascode working with both triodes getting enough Vp-k.
6BQ7 and 6DJ8 are the usual suspects for a cascode input stage. 6N24P too. 6N1P likes higher plate voltages than what we see here. But those are all noval tubes, so that's a problem.
Also, 6SL7 likes high plate-cathode voltage (Vp-k). It's usually run with >130Vp-k. This amp has B+ of only about 300V to play with, which makes it tough to get the cascode working with both triodes getting enough Vp-k.
6BQ7 and 6DJ8 are the usual suspects for a cascode input stage. 6N24P too. 6N1P likes higher plate voltages than what we see here. But those are all noval tubes, so that's a problem.
Last edited:
If you want more power, another thing you could do is change the output stage to pure pentode. Connect the EL34 screen grid (g2) to a 300V supply. The screen grid voltage (Vg2) doesn't need to be very low in an EL34. It's designed to be run with 400V on both its plate and its screen grid. 315V on the plate and Vg2 = 300V would be fine, and you'd get more power that way. You'd want to use more NFB to reduce the output impedance to acceptable levels.
The 6SL7 you end up with lower currents and slew rate issues. I think Rongon is suggesting staying with the 6SN7. Changing the plate 68K for a current source does increase loop gain and drop distortion. Going pentode made my simulation oscillate with NFB - so check on dummy load. Other than that its a 6550. Oddly I found the cathode resistor 6SN7 had little effect on distortion.
What model are you using for the OPT? That would be important for closed loop stability.
The large increase in gain from going with pentode mode could be the cause of the instability, which would then require compensation. That's pretty normal for a pentode amp with its higher levels of NFB.
I'm still concerned that the problems you're seeing are caused by poor choice of operating points for the 6SN7 cascode. It just doesn't look right to have 55V at the plate of the lower 6SN7.
I'm wondering if a 6AC7 or 6SJ7 (octal pentode) might not be a better choice for the driver stage. I mean, a cascode is basically similar to a pentode anyway, but with a pentode you'd get more voltage swing, which is what you want from a driver stage in a power amp.
The large increase in gain from going with pentode mode could be the cause of the instability, which would then require compensation. That's pretty normal for a pentode amp with its higher levels of NFB.
I'm still concerned that the problems you're seeing are caused by poor choice of operating points for the 6SN7 cascode. It just doesn't look right to have 55V at the plate of the lower 6SN7.
I'm wondering if a 6AC7 or 6SJ7 (octal pentode) might not be a better choice for the driver stage. I mean, a cascode is basically similar to a pentode anyway, but with a pentode you'd get more voltage swing, which is what you want from a driver stage in a power amp.
Thanks to you both. So I'll keep the SN7s and in the meantime, have removed the 470uf cathode bypass capacitors.
I'll use REW and Picoscope again tomorrow and then fit the LR8 or 10M90 set for 4ma. This will avoid the need to change the SN7 cathode resistor.
I'll use REW and Picoscope again tomorrow and then fit the LR8 or 10M90 set for 4ma. This will avoid the need to change the SN7 cathode resistor.
As I mentioned earlier, I have a pair of D3As which I will try in due course.
The OPTs are some unbranded Chinese ones and won't be the best.
Its nearer 2.1ma at the moment.
The OPT is a best guess on coupling factor and primary inductance. It seems to be tamed by placing a cap across the FB resistor of 470R (~2n2);
Great discussion.
The OPT is a best guess on coupling factor and primary inductance. It seems to be tamed by placing a cap across the FB resistor of 470R (~2n2);
Great discussion.
That is closer with the schematic I posted from Tube CAD Journal , then you should have also higher voltage on the lower triode , maybe close to 100V
If not , change the bias network for the upper tube grid , connect it from B+ not plate
Like it is , there is local negative feedback from plate to grid , could be good or not
Last edited:
The local fb plate R7 to grid is DC only. It will help a lot in keeping the plate voltage on U2 more constant as you change the current. If you want higher plate voltage on U3 too maybe up R10 to say 820R and R7 820k too. Higher plate voltage on U3 will give you slightly less swing, but the circuit seems very tolerant of component variations. The plate voltage on U3 is also static. The main thing the CCS is doing is increasing the open loop gain which should help with distortion. The UL is also helping here too. Anyway you seem to have a good grasp of the theory.
Last edited: