Not so much, no. This is accounted for in the design.
OK, so the 13k tail resistor you mentioned would then require 130V of negative rail. That doesn't strike me as much better.
Where?
In the circuit shown on page 544 there is nothing shown to address this issue. It does show one floating heater winding feeding all heaters. Something that may work fine with brand new tubes but would make a professional engineer cringe. Certainly a recipe for hum as the tubes age. And that power transformer better be well screened between primary and heater - otherwise its going to be noisy. A bit of heater-cathode leakage in one output tube and a bit of heater-cathode leakage in one LTP side and you have a feedback loop - which could be positive or negative but always non-linear. Heater-cathode leakage is in the manner of a semiconductor. That equals distortion directly, and the end of your precious suuposedly precision balance.
In Google Books I cannot determine the HT voltage, or the output triode types, which would give a clue to a suitable HT voltage.
But a typical HT voltage would be upwards of 250 V. 250 less 130 leaves 120 V from LTP cathode to HT. Not much, given the necessary drop in the anode loads but enough. Mostly with power triodes, you would have a fair bit more than 250 V in order to get a reasonable power output. Even a pair of 2A3's require 350V (cathode resistor bias) to deliver just 10 W - barely sufficient for quality audio.
350 V less 130 V equals 220V between LTP cathodes and HT. Even with the high grid drive required with the 2A3's low gm, there's enough.
No, there's no need for a negative supply. You never see them in commercial high quality designs, though I suppose you could find some wierd example somewhere.
Last edited:
Perhaps actually reading the book might be useful so that you don't miss all the discussion of the factors that are used in the design examples.
If you're happy with poor distortion performance, perhaps.
If you're happy with poor distortion performance, perhaps.
Well, I'm not going to buy his book just for a forum discussion. I have the circuit in front of me, and so do you I presume - that's all we need. Its the circuit that determines how it performs, not the author's musings.
If you were comfortable with the engineering (example: how is the hum intermod issue addressed? - you keep saying it is) you can explain it in your own words?
I notice you've never actually attempted that.
If you were comfortable with the engineering (example: how is the hum intermod issue addressed? - you keep saying it is) you can explain it in your own words?
I notice you've never actually attempted that.
Last edited:
If you don't want to read it and figure it out, that's your choice. It all there in the text, along with references to earlier work (including even a few Australians!).
No, that's not correct.
No, that's not correct.
DF96 - you are a sharp guy. I should have expected that you would remind me.
It doesn't really change the picture though does it. Whether 99% (1/gm assumption) or 98% (1/gm + Ra/mu assumption) of the signal current from one cathode goes up the other cathode, and 1% or 2% or 3% goes in the finite tail impedance, you've still got effective distortion cancellation - better than 30 dB on something that is 30 dB or more down from signal level anyway. And more cancellation occurs in the output stage - there's no cathode bypass for the output pair. Due to tube variations, anode resistor tolerance, etc, it's not going to be any better.
Similar comments apply to CMRR. This isn't an instrumentation amp, or an amp with neg feedback to one side, where CMRR really matters
Last edited:
Now I know that YOU haven't read the book. Because it jolly well is correct. Go look at the circuit, Figure 6.56 (it's labelled "Entire Bulwer-Lytton circuit diagram") on page 544 and see.
While looking at it again to make sure I get the page no, etc right, I notice there is a separate double supply for the silicon circuits. So your stament about extra cost of a negative supply was nonsense anyway wasn't it? The cost of extra transformer windings, diodes, electros, etc doesn't change much with voltage.
Though to be fair, a competent engineer could have engineered the silicon to run on the HT supply and not need separate feeds. Certainly not a dual feed.
Last edited:
Look again. And read the text. He does not show the heater circuits for the audio tubes in the schematic. In the text, he explains why and what he did. I will grant you that the caption is inaccurate.
I'm going to guess that the supplies you're looking at are for the MOSFETs- that's a no-extra-cost source of voltage for the CCS or long tail, assuming one hasn't bungled the engineering by requiring a separate 100-200V negative supply or hugely compromising distortion and swing.
I'm going to guess that the supplies you're looking at are for the MOSFETs- that's a no-extra-cost source of voltage for the CCS or long tail, assuming one hasn't bungled the engineering by requiring a separate 100-200V negative supply or hugely compromising distortion and swing.
You look again - or stop your nonsense.
There are two windings feeding diode bridges to supply the silicon. The negative end of the LTP CCS is fed from a three-terminal regulator (talk about unnecessary additions) and the 3-term reg is fed via two series resistors from the negative rail for the MOSFET follwer circuit, which is also excessively complex even if you belived that a precision follower circuit is necessary. And a separate winding clearly for the tube heaters, not connected to diodes or anything.
No way are those two sets of windings, diodes, etc not an extra cost. Of course they are extra cost. More than the cost of a simple negative rail for a resistor tail, not that such a thing is needed anyway.
I posted in good faith. If you are going to be silly the whole thing is pointless.
I notice you still haven't attempted to explain the supposed merits of it all.
There are two windings feeding diode bridges to supply the silicon. The negative end of the LTP CCS is fed from a three-terminal regulator (talk about unnecessary additions) and the 3-term reg is fed via two series resistors from the negative rail for the MOSFET follwer circuit, which is also excessively complex even if you belived that a precision follower circuit is necessary. And a separate winding clearly for the tube heaters, not connected to diodes or anything.
No way are those two sets of windings, diodes, etc not an extra cost. Of course they are extra cost. More than the cost of a simple negative rail for a resistor tail, not that such a thing is needed anyway.
I posted in good faith. If you are going to be silly the whole thing is pointless.
I notice you still haven't attempted to explain the supposed merits of it all.
Just to clear up this minor point, the floating 12V widing goes to the DAC magic as indicated on the diagram. See page 545 for the heater supply:
"All of the preceding supplies can be derived from a common transformer, but the audio valve heaters need their 6.3 V at 6 A from a separate transformer (ideally a split bobbin EI type) to avoid interference from rectification spikes. Unfortunately, the author didn’t have room on the top of his chassis for a canned EI heater transformer, so he was forced to use a 6 V 50 VA toroid under the chassis"
The only other mention of hum wrt differential pairs I could find was on page 540:
"However, there is capacitance between the cathodes and the heater (Chk = 7 pF per triode for 7N7), and the heater transformer could easily have 1 nF capacitance to the mains winding, picking up common-mode interference. Thus, by unbalancing the inputs of the differential pair, we force it to amplify signals on its cathodes, rendering it sensitive to mains interference coupled through heater transformer inter-winding capacitance, and that is why the balanced input sounds better, not because of better rejection of interference on audio cables."
The reason for using beefy followers is explained on page 533:
"The amplifier was required from the first to be scalable, with driver circuitry capable of driving many pairs of output valves, and that immediately implies cathode followers..."
Last edited:
While you've been having fun I've been scoping the depths of my ignorance. In applying WD40 to the memory banks I've found a whole load of stuff I didn't know so currently off starting from first principles again on a few things. It's remarkable how much 'then a miracle occurs' has been written about UL operation!
I have finally put the LED bias to bed for EL34s though. The operating point (around -35v on the grid) eats more than I care for from the HT budget, even if I haven't calculated how much of a gnat fart it loses in output power.
I have finally put the LED bias to bed for EL34s though. The operating point (around -35v on the grid) eats more than I care for from the HT budget, even if I haven't calculated how much of a gnat fart it loses in output power.
Well,
If you use fixed bias and you have enough negative voltage you can bias 6L6,EL34, KT66, KT88, 5881 etc and switch Triode pentode U/L + Electron stream and pentode for a laugh. If you don't have enough bias voltage double.
Or you can use mixed bias.
Make the bias so if the pot fails it shuts down the tubes..and if your still not sure fuse the cathodes.
Make sure the bias voltage is up before the output tubes conduct.
Put a 10 ohm resistor in the cathode of each power tube and set the voltage drop to equal the tube idle current.
People will say its wasting power you can reduce it. Put test points for each tube with a common ground.
Put a 1K on the screen grid.
Just a few ideas..feel free to ignore. 🙂
Regards
M. Gregg
If you use fixed bias and you have enough negative voltage you can bias 6L6,EL34, KT66, KT88, 5881 etc and switch Triode pentode U/L + Electron stream and pentode for a laugh. If you don't have enough bias voltage double.
Or you can use mixed bias.
Make the bias so if the pot fails it shuts down the tubes..and if your still not sure fuse the cathodes.
Make sure the bias voltage is up before the output tubes conduct.
Put a 10 ohm resistor in the cathode of each power tube and set the voltage drop to equal the tube idle current.
People will say its wasting power you can reduce it. Put test points for each tube with a common ground.
Put a 1K on the screen grid.
Just a few ideas..feel free to ignore. 🙂
Regards
M. Gregg
Last edited:
I would,
probably set the bias at about 40mA,
then Give it a try go from there.
You can use the 10 ohms or 12 ohms for a standard value of 7/10W wire wound and increased fixed bias for tube rolling.
With the selector switch triode /UL etc.
Regards
M. Gregg
probably set the bias at about 40mA,
then Give it a try go from there.
You can use the 10 ohms or 12 ohms for a standard value of 7/10W wire wound and increased fixed bias for tube rolling.
With the selector switch triode /UL etc.
Regards
M. Gregg
These are all valid options. front panel space is a bit limited, but should be able to put a couple of extra switches in.
Well the circuit is labelled "Entire Bulwer-Lytton circuit diagram") on page 544.
Google Books wouldn't let me look at p 545, but with that word "Entire" in there, why on earth would I or anyone else suspect he left bits out.
This just illustrates what I've been saying - MJ is not a guru.
The interwinding capacitance can be dealt with quite simply with the usual inter-winding screen - pice of sheet copper forming one open circuit turn around the primary and earthed to the chassis.
The heater-cathode capacitance is the least of the concerns. 7 Pf is a reactance of 380,000,000 ohms at 60 Hz. Still 38 megohms at 600 Hz.
The ral issue that MJ missed, apparently, is heater cathode leakage - a non-linear resistance effective down to DC. As low as a few hundred kohm in an old but still good tube. So, you get hum and alteration of operating conditions - a slight imbalance.
The reason for using beefy followers is explained on page 533:
Oh, right. So cost effectiveness and performance were sacrificed to for another amplifier entirely. Remind me to put a 3-phase power supply in my next guitar amp - that will make it scalable.
Last edited:
It is indeed remarkable how much has been written about UL. And a lot of it without real understanding. For a long time some authors insisted in calling it "partial triode" due to their mistaken assumption that it is some sort of compromise between pentode operation and triode operation.
It's really quite simple. Better than a compromise and certainly no miracle.
By applying negative feedback to the screens, we get to use the curvature (non-linearity) of the screen control over current to offset the curvature of the grid control over current. A bend one way is pulled straight by bending it the other way.
- Status
- Not open for further replies.