Connecting the screen cap to the cathode increases the gain of the first stage slightly in simulation by adding a small amount of positive feedback. It has a very detrimental effect on the clipping of the amp. So ground is better as is. Shows to check first.
The input of +/- .6V gives full output which is a little sensitive. If you remove the input cap and the two input resistors, you could add a potential divider there making the lower resistor say 47k the upper 150k and the input cap say 100nF - the centre of the resistors feeding the grid. See diagram.
Last edited:
Wish it were a simple case of pull some gain out. I can't put more feedback in as it is very difficult to stabilise the amp. Since the noise is actually in the preamp at a fairly low level and it's volume control does not affect the level of the noise, only the signal, attenuation is an easier solution.
Thank you. The existing filter network was confusing me with respect to adding an effective L-pad. I'll look at re-drawing the schematic later today.
Apologies moving the screen cap does not affect the clipping. I don't think it makes any real difference. Seems to be hang over from the RF days of keeping the leads as short as possible. You can remove the cap all together which drops the open loop gain by 8dB. The NFB would need adjusting to compensate. Mind that more like using a triode in the first stage maybe a 12AX7 would work just fine here!
Last edited:
ejp,
From the schematic of post 361, if you remove the EF86 cathode bypass cap, it will automatically increase the negative feedback.
That is because the R and C at audio frequencies are already driving 1.5k in parallel with 470 Ohms to ground.
But when the bypass cap is taken out of the circuit, the 470 Ohm resistor is out too.
The R and C drives 358 Ohms with the bypass cap and 470 Ohm in place,
and the R and C drives 1500 Ohms with the bypass cap and 470 Ohm taken out of the circuit.
That is a 4.19 increase of negative feedback (12.4 dB)
You do not have to change the R and C from the output transformer to get that.
But of course, the R and C also have to drive the EF cathode impedance which is in parallel with the ~ 300 Ohm or 1500 Ohm.
That is the one factor I left out of the above analysis. So the negative feedback increase will be a little less than 12dB.
From the schematic of post 361, if you remove the EF86 cathode bypass cap, it will automatically increase the negative feedback.
That is because the R and C at audio frequencies are already driving 1.5k in parallel with 470 Ohms to ground.
But when the bypass cap is taken out of the circuit, the 470 Ohm resistor is out too.
The R and C drives 358 Ohms with the bypass cap and 470 Ohm in place,
and the R and C drives 1500 Ohms with the bypass cap and 470 Ohm taken out of the circuit.
That is a 4.19 increase of negative feedback (12.4 dB)
You do not have to change the R and C from the output transformer to get that.
But of course, the R and C also have to drive the EF cathode impedance which is in parallel with the ~ 300 Ohm or 1500 Ohm.
That is the one factor I left out of the above analysis. So the negative feedback increase will be a little less than 12dB.
Last edited:
OK the bypass cap and the 470R were in there to provide lead-lag compensation at LF. Without them the LF pole of the transformer and the LF pole of the output grid coupling caps push the LF quite close to instability and produce a peak in the LF response at 2-5Hz. Unfortunately it has the knock on effect of making the gain higher in the audio band. A better approach would have been to split the FB resistor 18k in half and put the LF compensation (series cap and R down to ground) there. This would have much less affect on the cathode and forward gain. Removing the screen cap also seems to work but again need the NFB components changing.
Last edited:
Adjusted NFB with LF lead-lag compensation moved away from cathode. This drops the forward gain by ~9dB with 18dB NFB. C14 should really be non polarised.
Last edited:
Sounds good. I'd still like to integrate around 6-9dB of attenuation too. Tweaking the turret board to move that LF lead-lag network will prove interesting. I've adjusted my schematic to match. Ideally I'm looking for roughly 0.8-1V input for full output. I'm thinking the 150K/47K resistors may be a bit excessive and perhaps 68k/33k are more appropriate. C14 is already bipolar.
I'm also looking at shifting the cathode biasing section of the turret board altogether to help reduce the footprint in order to shrink the width of the amp down towards 16" and the depth to about 9".
I'm also looking at shifting the cathode biasing section of the turret board altogether to help reduce the footprint in order to shrink the width of the amp down towards 16" and the depth to about 9".
+/- 1.1V input gives just clipping. That's with the old 220k + 22k circuit i.e. no input divider. About +/- 22V into 8R on the output. Why not fuse the cathode of V3 with a single fuse, with the 1n4007 in I don't think anything bad will happen.
Last edited:
Hmm, I think I may need to test the solid state amp and find out what it's gain is and what the pre amp gain is outputting. Also don't forget that my B+ is almost 25V higher than your simulation at 463V rather than 440V.
I followed Merlin's advice on fusing. Probably overkill but prefer to play it safe.
The Valve Wizard
I followed Merlin's advice on fusing. Probably overkill but prefer to play it safe.
The Valve Wizard
Thinking about it, the higher voltage simply provides greater headroom. That said, how does it affect your combination bias with respect to dissipation limits? I can't see any glow with my eyes but a camera can still see some infrared from the anodes. (i'm not sure if that is normal or not)
These are the turret board adjustments, which I hope are correct. I'll leave the attenuation off the input initially as I can then judge if the gain needs further adjustment.
I am surprised that moving R30/C1 drops the gain. When I left them out during testing, it didn't appear to make any difference. (not that I was really paying much attention to the input level vs output) Does this change provide any other positives or negatives to the likes of THD and stability margin?
These are the turret board adjustments, which I hope are correct. I'll leave the attenuation off the input initially as I can then judge if the gain needs further adjustment.
I am surprised that moving R30/C1 drops the gain. When I left them out during testing, it didn't appear to make any difference. (not that I was really paying much attention to the input level vs output) Does this change provide any other positives or negatives to the likes of THD and stability margin?
Stability may well be better. You may need to tweak the FB cap. The THD as .3% at 20v peak to peak O/P which is very good if you believe the figures. This is before the cathode bias droops.
Just to check, the feedback resistor currently is 18K with 150pF in parallel. (which gave around 16dB of feedback as the overall gain was around 49dB OL and 33dB CL) As per the schematic in this post.
Shifting the lead-lag network as per your adjustments here, the feedback resistor gets split in two but are 6.8K & 2.2K. I'm thinking that the amount of feedback I'm going to see is likely to be higher than the 18dB in the simulation. Obviously R28 has changed from 470R to 1K and I'm not sure how that will affect it. I'm currently making the adjustments to the turret board.
Shifting the lead-lag network as per your adjustments here, the feedback resistor gets split in two but are 6.8K & 2.2K. I'm thinking that the amount of feedback I'm going to see is likely to be higher than the 18dB in the simulation. Obviously R28 has changed from 470R to 1K and I'm not sure how that will affect it. I'm currently making the adjustments to the turret board.
I've made the changes necessary. Just need to reinstall the board into the amp and reconnect all of the cables.
I did have one thought regarding the increased voltages vs the simulations. How does it affect the bias on the output valves? I figured that the PC marking on my valves is their plate current. The JJ KT77's I have are 49.5mA which is 10mA less than both the EL34B's and 6CA7's. It would be nice if they said how that was derived.
I did have one thought regarding the increased voltages vs the simulations. How does it affect the bias on the output valves? I figured that the PC marking on my valves is their plate current. The JJ KT77's I have are 49.5mA which is 10mA less than both the EL34B's and 6CA7's. It would be nice if they said how that was derived.
I finished the reassembly late this afternoon. I then spent nearly an hour trying to find my dummy load resistor as I'd "put it away". I've given up for tonight, better not start testing it this late as I risk making silly mistakes.
Tested this morning. Stability seems similar, it still gets a bit upset below 12Hz with large input signals. The expected reduction in sensitivity is pretty decent. The only not so good thing is the HF response falls off worse with feedback now than it did with no feedback at all previously which I thought was a bit odd. I've not tested open loop in this configuration yet but I'm guessing it will be worse too. Closed loop gain appears to sit around 25dB.
Close to maximum clean output.
Clips around 0.9V input.
10KHz Square and 1KHz square look pretty clean. I was expecting 10KHz to be more rolled in considering the drop off in HF response. Some wrinkles start to form as the input level increases but nothing that looks worrying.
This is the only thing I noticed that I managed to capture. (having had a moment of clarity using the storage function of the scope) I assume this is power supply noise of some sort. It appears on the output at a low level and grid of EF86 at a very low level.
Close to maximum clean output.
Clips around 0.9V input.
10KHz Square and 1KHz square look pretty clean. I was expecting 10KHz to be more rolled in considering the drop off in HF response. Some wrinkles start to form as the input level increases but nothing that looks worrying.
This is the only thing I noticed that I managed to capture. (having had a moment of clarity using the storage function of the scope) I assume this is power supply noise of some sort. It appears on the output at a low level and grid of EF86 at a very low level.
That's R22 in your spice schematic, part of the step network around the anode load of EF86?
Actually, something I wondered about yesterday regarding the step network is the order as mullard used CR whereas we are using RC. I'm assuming in this use case, it doesn't make a difference.