People who are too keen to get rid of Miller capacitance in 5-20-like circuits often find that they have to add compensation networks to get loop stability. For slew rate purposes the input stage is not bothered whether it is loaded by 100pF LTP Miller capacitance or 100pF compensation capacitor added to the anode circuit.
The ECC83/12AX7 has just about enough mu that it can do LTP PS without a CCS tail. Almost all other valves need either a CCS tail or unbalanced anode resistors. The weakness of this valve for this position is not its linearity (which is good) but its driving ability (which is low). However, on real music it can drive an EL34 UL output without any problems. It might not cope quite so well with a triode output, as that will have more Miller capacitance.
It is true that putting feedback to the input cathode creates scope for common-mode distortion, which feedback cannot reduce. This is most likely to be a problem with those designs which have deliberately engineered low sensitivity (i.e. high input voltage required) as that means the common-mode voltage is high. A pentode is better in this respect, provided g2 is decoupled to the cathode and not ground. However, most triodes will be fine too. I did a back-of-envelope estimate when I decided to use an ECC81/12AT7 as my input triode and decided it was OK.
The ECC83/12AX7 has just about enough mu that it can do LTP PS without a CCS tail. Almost all other valves need either a CCS tail or unbalanced anode resistors. The weakness of this valve for this position is not its linearity (which is good) but its driving ability (which is low). However, on real music it can drive an EL34 UL output without any problems. It might not cope quite so well with a triode output, as that will have more Miller capacitance.
It is true that putting feedback to the input cathode creates scope for common-mode distortion, which feedback cannot reduce. This is most likely to be a problem with those designs which have deliberately engineered low sensitivity (i.e. high input voltage required) as that means the common-mode voltage is high. A pentode is better in this respect, provided g2 is decoupled to the cathode and not ground. However, most triodes will be fine too. I did a back-of-envelope estimate when I decided to use an ECC81/12AT7 as my input triode and decided it was OK.
What's your take on the idea of the first ECC83 with GNFB added to the second grid? The GNFB will be switched in or out as required.
I don't want to use a cathode decoupling capacitor on the EF86 hence the idea of two ECC83s. Also, I don't like the idea of wiring up a pentode as a triode, input or output. Doesn't seem right.
Last edited:
You mean have an LTP as the first stage? That doesn't eliminate the common-mode issue, it just modifies it. Common-mode distortion is always a potential problem whenever the feedback goes to a different point from the input. Unfortunately, this normally creates a choice between CM and thermal noise.
A 5-20 with no feedback will not work too well, unless intended never to have feedback. You can't just add or delete feedback like seasoning - it has to be designed in or designed out.
If you suffer from faradaphobia then you may need to seek assistance from a self-help group. Those of us who don't can find it hard to sympathise and suggest avoidance coping mechanisms. All I will say is that the cathode decoupling cap should be inside the loop, not part of the feedback network.
A 5-20 with no feedback will not work too well, unless intended never to have feedback. You can't just add or delete feedback like seasoning - it has to be designed in or designed out.
If you suffer from faradaphobia then you may need to seek assistance from a self-help group. Those of us who don't can find it hard to sympathise and suggest avoidance coping mechanisms. All I will say is that the cathode decoupling cap should be inside the loop, not part of the feedback network.
That made me smile. I don't have faradaphobia. Just thought that removing it would reduce the gain of the EF86 but that could cause more headaches according to postings on 'ere.
Agreed about slew effect. The difference is though that Miller capacitance effect acts with the input resistance, which can vary from very low to 25K e.g. with something feeding through a 100K pot. The change in effect can be easily seen by feeding a square wave generator to the power amplifier through a 100K pot.
The second thing is that one often finds that this capacitor has to be in series with a resistor, bringing the phase shift back to zero once the loop gain is diminished. Other capacitances start coming in at those frequencies, making life difficult if the first pole is also still in.
P.I.F.E,
You pose difficult questions! Yes, it would work - but after observing several factors. The 'internal resistance' (rp) of an ECC83 will be much higher than that of a triode-strapped EF86, and even more so with an unbypassed cathode there. The gain reduction will be in order, but a re-visit to feedback optimisation will need to be done because of different Miller-effect - er - effects, both before and after the triode. Thus not a no-go, only will need attention to compensating R.Cs.
Now that we are going away from the original 5-20, there are other possibilities, some hinted at before. E.g. a triode-pentode as first stage will still give pentode advantages and a free triode for the intended bias protection (e.g. 6U8A). As my opinion was before, I much prefer a lower µ tube for the phase inverter to lower the nett gain, with its added advantages over a high-µ triode. But it comes down to your decision the further we go - hopefully not mixing you up!
[P.S: I know how it feels, but there really needs to be no stress in using a pentode triode-strapped. One has no alternative in output tubes, and there is no objection to an EF86 triode-strapped; only that one could then have had two triodes instead in one socket.]
Last edited:
You do not need to decouple the cathode resistor fully. It can be partly decoupled and the open loop gain adjusted by this way.
Below is an example. The GNFB is supplied at the top of R15.
The gain of below circuit is 37 dB. When the cathode resistor is fully decoupled, the gain is 35 dB.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Sy,
Does this mean that if I have a LTP with 43K plate loads and a CCS in the tail, and the first grid is AC coupled via a 0.22u cap followed by a 1 Meg resistor to ground, while the second grid is only AC coupled via a 0.22u cap to ground, that there will only be balanced currents through the loads during signal (program) conditions? At idle with no signal applied, only the first triode will conduct due to the cathode bias via the 1 Meg resistor? Is this correct? When I built my amp, I measured a voltage drop on one triode load but no drop on the second.
Thanks,
Ron
What's the DC reference of the grid that's cap coupled to ground?
If the input half of the diff amp is DC ground referenced, take the second section grid straight to ground. At AC, your balance will then be as perfect as the matching of the plate resistors. At DC, the plate voltages will vary unless the sections are perfectly matched or you put some sort of adjustment in the cathodes.
If the input half of the diff amp is DC ground referenced, take the second section grid straight to ground. At AC, your balance will then be as perfect as the matching of the plate resistors. At DC, the plate voltages will vary unless the sections are perfectly matched or you put some sort of adjustment in the cathodes.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.