You are right... I forgot the capacitance... The FET source measured better because of it, the bipolar worse...
I will breadboard a bipolar PNP source, this way the probe will be across the resistor, allowing a direct comparison between the two.
Still, does it explain the huge difference?
I will breadboard a bipolar PNP source, this way the probe will be across the resistor, allowing a direct comparison between the two.
Still, does it explain the huge difference?
Last edited:
Dunno... to me the original schematic looks like a Rube Goldberg device hanging off the plate of the first stage pentode. I mean the 10M45S is itself a remarkably competent current regulator; adding another MOSFET to it to get additional current regulation? Oh. I see. The famous over-kill-er-ator circuit.
Oh well. The real purpose of the constant-current source is to act like a perfect inductor having near-infinite inductance. So that at whatever convenient operating point is established, the 'buck voltage' of the synthetic inductor will push the anode to higher voltage ("amplify") or lower, based on the grid bias and the tube mu. All very nice. But the "problem" comes about if the chosen current leaves the plate at too high a voltage. It then has compromised swing - which is exactly what the oscilloscope snapshot seems to show.
So what is the quiescent voltage at the plate of the input pentode? I could presume it is the 210V listed in the circuit diagram, but maybe that's only the DESIGN point. It really should be adjusted to be no higher than about 60% of B+ supply. And yes, the poor regulator would need that heat sink. 175 to 225 would be fine.
Anyway... do also read DF96's posts - he makes some important and good points.
GoatGuy
Oh well. The real purpose of the constant-current source is to act like a perfect inductor having near-infinite inductance. So that at whatever convenient operating point is established, the 'buck voltage' of the synthetic inductor will push the anode to higher voltage ("amplify") or lower, based on the grid bias and the tube mu. All very nice. But the "problem" comes about if the chosen current leaves the plate at too high a voltage. It then has compromised swing - which is exactly what the oscilloscope snapshot seems to show.
So what is the quiescent voltage at the plate of the input pentode? I could presume it is the 210V listed in the circuit diagram, but maybe that's only the DESIGN point. It really should be adjusted to be no higher than about 60% of B+ supply. And yes, the poor regulator would need that heat sink. 175 to 225 would be fine.
Anyway... do also read DF96's posts - he makes some important and good points.
GoatGuy
GoatGuy: thank you for your comments. The initial problem was solved by giving the CCS the headroom it needed. The schematic seems complicated, but I do have my reasons for that. Believe me, I did a lot of comparisons and listening before I came up with this Rube Goldberg machine.
It sounds beautifully anyway and outputs 29 clean watts...
Now we are discussing CCS impedances. SY pointed out (again) my beginner's error. I will measure again with a 10x probe...
It sounds beautifully anyway and outputs 29 clean watts...
Now we are discussing CCS impedances. SY pointed out (again) my beginner's error. I will measure again with a 10x probe...
I spot a different problem with the setup using a bias current source, or any stabilized source for that matter. I assume that your B+ comes from an unregulated power supply. If that is the case, the DC voltage varies along with the mains variation, which may sometimes be as much as +/- 10 %. The plate voltage will also vary by the same amount. But the bias voltage remains unchanged. Looking at the GM70's characteristics and assuming that you've set the bias current when the mains is at nominal, the actual bias current may change from anemic to a rather strong overload during the day due to mains voltage fluctuation. Even a rather benign +/- 5 % mains voltage change will already cause a larger change in current than you may find acceptable. So of you have an unregulated B+, it is best to also supply the grid bias from an unregulated voltage. In this way, both voltages track each other, and the tube's operating point does not shift as much.
Vincent - your experiments are interesting. I use Rod's regs for my DHTs and he's a very clever chap. So I take his designs seriously. I'd be interested in a circuit for a 300b.
This is an interesting thread, and I look forward to some more schematics. Frankly I'm interested in what it sounds like. I use PSE 4P1L in filament bias for my outputs, and though I'm happy with this solution, I have a draw full of 300bs that I might be more interested in if I could use them in fixed bias. I abandoned cathode bias with a bypassed resistor because all bypassed cathode resistors I've ever used anywhere were inferior to filament bias, and by a long way too in the case of electrolytics, which frankly sounded dreadful.
This is an interesting thread, and I look forward to some more schematics. Frankly I'm interested in what it sounds like. I use PSE 4P1L in filament bias for my outputs, and though I'm happy with this solution, I have a draw full of 300bs that I might be more interested in if I could use them in fixed bias. I abandoned cathode bias with a bypassed resistor because all bypassed cathode resistors I've ever used anywhere were inferior to filament bias, and by a long way too in the case of electrolytics, which frankly sounded dreadful.
You are absolutely right. The first version of this amplifier was made with an unregulated bias supply, for the reason you exposed.
With the regulated (in fact it's stabilized, not regulated) bias supply, I have seen the current change by +/-5mA... that's still acceptable to me.
And there's an Arduino microcontroller overseeing the amplifier. In case of overload or abnormal average current, it turns it off!
Hi Andy,
I am glad to see that I'm not the only one who doesn't like the sound of capacitors! I do whatever I can to keep them out of the signal path.
The filament bias is an amazing idea, really. Rod's CCS bias idea also seems to work - I got the same kind of improvement you'd expect after eliminating a bypass capacitor. I need to experiment some more with it, to make sure everything is stable, but it looks promising anyway.
Salas' SSHV shunt regulator also isolates the PSU caps and has a very low impedance, so it should work very well for fixed bias - I'll have to try that too...
Cheers,
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.