> The figures given in the 807 app report do reference voltages to ground. They shouldn't've, but this is quite common when dealing with VT specs.
That line should read "Plate Supply Voltage"; someone had tired fingers.
In commercial design, you are "given" a certain voltage (maybe the Buyer got a "deal" on some odd-volt power transformers) and you have to make the most of it. We assume that anyone can read "300v, 250Ω cathode resistor" and know the actual voltage across the tube is less.
> I assume that self bias would be essentially identical to fixed bias.
Only if the individual tubes have ZERO 2nd-harmonic distortion.
Real tubes have significant 2nd-H nonlinearity. 6L6/807 has a lot.
From a typical grid bias point, going 10V positive may increase current 50mA but 10V negative decreases it only 40mA. Tube Gm goes down with current.
The most obvious proof is any tube-manual condition for single-ended operation. The plate current should be the same at idle and at full-signal: in fact it declines significantly at full signal.
Look again at Miles' 807 data. For self-bias we have 88mA+5mA= 93mA flowing in 250Ω, grid bias voltage is -23.25V. For an ideal tube working in "good" conditions, the peak grid voltage will be equal to the grid bias (fixed-bias shows 45V p-p or 45V/2= 22.5V peak for 22.5V bias). But the self-bias condition shows 57V p-p or 28.5V peak... from a -23.25V bias point? They are not driving the grids positive, they are swinging current up and down, and tube nonlinearity means more up than down. Cathode current increases, which shift the bias voltage. I figure 114mA, the chart shows 119mA, good agreement.
And that's ignoring load current. At 6K6 load, the cathode current wants to go up to 154mA. If we were using the 250Ω cathode bias resistor, the grid-cathode voltage would shift to -38.5V. That kind of major shift would put the tube at very low Gm at the zero-crossings (what sand-heads call crossover distortion, but much softer) and a high THD number. To keep that from looking too bad, they spec a higher 9K load for lower signal current.
If the load impedance is very-very high, or signal current is very-very low, then yes: self- and fixed-bias come to about the same thing. But power is very low too. For "interesting" impedance, current, power, self-bias is like a "fixed" bias that actually varies with signal level.
All of that assumes perfect bypassing. And no positive-grid overdrive. Even so, I don't believe the 2% versus 4% sheet-spec has anything to do with electrolytic distortion. Those guys were clever enough to fix or subtract any flaws that were not really in "their" tubes. Cap-distortion will vary with frequency, tube distortion won't, so they could quickly check for cap-crud. While they used the $0.10 caps in the consumer products, a Lab has $1.00 electros and $100 banks of oil-caps they can use to "perfect" their measurements. And it is reasonably valid to use the resistor to find the full-power bias point, then replace it with a tapped truck-battery (0.1Ω dynamic resistance) to take the THD measurement.
But self-bias does not have a bad rap in speech/music work. We don't have sustained full-power tones. Musical dynamic peaks are short enough that a $0.50 capacitor (couple-hundred uFd) will not change much before the transient is passed. Put a voltmeter on a self-biased amp's cathode resistor and play music. If it idles at 23.25V, then most normal playing will stay there or rise to maybe 23.5V occasionally. Driven into the edge of audible clipping, it may rise to 24V, with hard-rock maybe 25V. But speech/music can't force a decently-bypassed self-bias stage all the way to the 28.5V bias that we see in steady-state test-tone testing, not unless it is "beyond gross clipping".
That line should read "Plate Supply Voltage"; someone had tired fingers.
In commercial design, you are "given" a certain voltage (maybe the Buyer got a "deal" on some odd-volt power transformers) and you have to make the most of it. We assume that anyone can read "300v, 250Ω cathode resistor" and know the actual voltage across the tube is less.
> I assume that self bias would be essentially identical to fixed bias.
Only if the individual tubes have ZERO 2nd-harmonic distortion.
Real tubes have significant 2nd-H nonlinearity. 6L6/807 has a lot.
From a typical grid bias point, going 10V positive may increase current 50mA but 10V negative decreases it only 40mA. Tube Gm goes down with current.
The most obvious proof is any tube-manual condition for single-ended operation. The plate current should be the same at idle and at full-signal: in fact it declines significantly at full signal.
Look again at Miles' 807 data. For self-bias we have 88mA+5mA= 93mA flowing in 250Ω, grid bias voltage is -23.25V. For an ideal tube working in "good" conditions, the peak grid voltage will be equal to the grid bias (fixed-bias shows 45V p-p or 45V/2= 22.5V peak for 22.5V bias). But the self-bias condition shows 57V p-p or 28.5V peak... from a -23.25V bias point? They are not driving the grids positive, they are swinging current up and down, and tube nonlinearity means more up than down. Cathode current increases, which shift the bias voltage. I figure 114mA, the chart shows 119mA, good agreement.
And that's ignoring load current. At 6K6 load, the cathode current wants to go up to 154mA. If we were using the 250Ω cathode bias resistor, the grid-cathode voltage would shift to -38.5V. That kind of major shift would put the tube at very low Gm at the zero-crossings (what sand-heads call crossover distortion, but much softer) and a high THD number. To keep that from looking too bad, they spec a higher 9K load for lower signal current.
If the load impedance is very-very high, or signal current is very-very low, then yes: self- and fixed-bias come to about the same thing. But power is very low too. For "interesting" impedance, current, power, self-bias is like a "fixed" bias that actually varies with signal level.
All of that assumes perfect bypassing. And no positive-grid overdrive. Even so, I don't believe the 2% versus 4% sheet-spec has anything to do with electrolytic distortion. Those guys were clever enough to fix or subtract any flaws that were not really in "their" tubes. Cap-distortion will vary with frequency, tube distortion won't, so they could quickly check for cap-crud. While they used the $0.10 caps in the consumer products, a Lab has $1.00 electros and $100 banks of oil-caps they can use to "perfect" their measurements. And it is reasonably valid to use the resistor to find the full-power bias point, then replace it with a tapped truck-battery (0.1Ω dynamic resistance) to take the THD measurement.
But self-bias does not have a bad rap in speech/music work. We don't have sustained full-power tones. Musical dynamic peaks are short enough that a $0.50 capacitor (couple-hundred uFd) will not change much before the transient is passed. Put a voltmeter on a self-biased amp's cathode resistor and play music. If it idles at 23.25V, then most normal playing will stay there or rise to maybe 23.5V occasionally. Driven into the edge of audible clipping, it may rise to 24V, with hard-rock maybe 25V. But speech/music can't force a decently-bypassed self-bias stage all the way to the 28.5V bias that we see in steady-state test-tone testing, not unless it is "beyond gross clipping".
Its easy, a regulator is not needed at all as the current draw is neglible as long as we are talking about class A1, AB1 or even B1.
Use a rectifier bridge, a cap a series resistor, a cap, a series resistor, pot and resistor in series to ground and then a cap at the viper of the pot.
In my OTL i use 100u, 22k, 220u, 22k, 22k pot and 39k, with 220u at the pot viper. this is compeletly silent and has no effect whatsoever on the hum level, in a push-pull amp the requirements are very low any hum on the bias is cancelled due to balance, but the circuit I describe is more than enough for any SE-amp also.
To use an electronic regulator for bias is serious overkill and a complete waste of time and money.
A problem with using a electronic regulator is that the bias voltage stays fixed even when the mains and anode voltage is dropping which means that the cathode idle current goes down.
What you need is that the bias voltage also is automatically adjusted so that the cathode idle current stays the same, this can be sometimes easily achieved by using a zenerdiode in series of one of the series resistors of the circuit described above.
Without the zener the bias voltage usually will drop to quick which means that the cathode current increases, (at least for high Gm tubes)
Regards Hans
PRR said:
Ok, I'm missing something here. I've assumed that the function of the bypass capacitor is to keep the voltage at the cathode constant for music signals (bypassing AC). If that's true, I don't see how that differs from fixed bias (perfect cap assumption). If the cathode/plate/ grid voltage relationship is the same in both cases, output should be the same. And signal swing should be the same. Charging the cap to a higher level under clipping makes sense, I think I get that. But otherwise, I don't see why the current in the tube should be different from the fixed bias case. Are those chart values taken under static conditions? That would be different. Sorry if I'm slow. I'd like to understand this.
One thought; even with a perfect cap, you'd see some ripple at the resistor/cathode. Is that what leads to the higher distortion?
Sheldon
BTW, I really appreciate your posts. You go to great lengths to provide a detailed explanation.
There is no difference all things being equal. But all things aren't equal- in order to keep that bias reasonably constant, the tube operating conditions are generally changed so that the cathode current doesn't vary as much from idle to full power. In theory (and maybe more than theory), you could trade off being able to run sustained power levels above the idle and bias up an automatic bias output stage to run the same conditions as fixed bias. Mullard had a nice little treatment of this in their tube book, calling it "low loading."
Low loading is a fine and efficient way of running automatic bias, BUT it is not suitable for loud music with limited dynamic range (i.e., most pop). And it's unsuitable for test bench analysis, but that's a pretty liveable trade-off.
Low loading is a fine and efficient way of running automatic bias, BUT it is not suitable for loud music with limited dynamic range (i.e., most pop). And it's unsuitable for test bench analysis, but that's a pretty liveable trade-off.
Thanks Both,
I think I've got it. It seems the significant issue is charging up that cap so that the cathode voltage drifts higher. At lower outputs, the circuits are mostly equivalent, sonic subtleties allowed for. I wasn't weighing in on one side or the other, just wanted to understand the issues.
Sheldon
I think I've got it. It seems the significant issue is charging up that cap so that the cathode voltage drifts higher. At lower outputs, the circuits are mostly equivalent, sonic subtleties allowed for. I wasn't weighing in on one side or the other, just wanted to understand the issues.
Sheldon
Can I ask a further question about combination bias? This could be done in several ways, for instance:
a) Bypassed cathode resistor and fixed bias on grid (Bill Beard's 'split bias' for example)
b) Unbypassed cathode resistor and fixed bias on grid (e.g. Axiom amp from Chimera labs, where I believe the unbypassed cathode resistor on each 300b was 100 ohms)
c) the so-called "harmonic equaliser, with an unbypassed common cathode resistor and fixed bias.
Referring to a previous post:
"my RCA RC-16 manual shows a max grid resistance for the 2A3 at 500K for cathode bias and 50k for fixed bias"
My question is how do you calculate the grid resistance needed when you use combination bias? In the case above it's clearly between 500k and 50k, but where???? Andy
a) Bypassed cathode resistor and fixed bias on grid (Bill Beard's 'split bias' for example)
b) Unbypassed cathode resistor and fixed bias on grid (e.g. Axiom amp from Chimera labs, where I believe the unbypassed cathode resistor on each 300b was 100 ohms)
c) the so-called "harmonic equaliser, with an unbypassed common cathode resistor and fixed bias.
Referring to a previous post:
"my RCA RC-16 manual shows a max grid resistance for the 2A3 at 500K for cathode bias and 50k for fixed bias"
My question is how do you calculate the grid resistance needed when you use combination bias? In the case above it's clearly between 500k and 50k, but where???? Andy
Why not just use the lower value just to be safe, then design the driver to accommodate that?>
You're interfering with my love affair with Russian teflon .1uF coupling caps. I already have ten of them per channel in my preamp to make up 1uF. I'm thinking of going balanced which is twenty more. Have mercy - hazard a guess!
You're interfering with my love affair with Russian teflon .1uF coupling caps. I already have ten of them per channel in my preamp to make up 1uF. I'm thinking of going balanced which is twenty more. Have mercy - hazard a guess!
Clean bias supply........
Start out with twice the required bias DC output voltage via a full-wave rectifier & capacitor. (If you need -50 vdc bias, start with -100 vdc).
Supply that semi-filtered voltage in a RCRC filter. Drop voltage approximately 15% in the first resistor & 15% in the second resistor. With 100 volts in, 70 volts is the output. Feed that voltage into a potentiometer that supplies a few millampere load to ground. The wiper on the pot is the adjustable bias voltage out. For better resolution on output voltage, the pot can be in series with another resistor to ground. The first three capacitors are 50uF and I prefer a 10uF poly as the last capacitor connected to the pot wiper output. Adjust the wiper on pot for -50 vdc in this example.
The end result is actually a CRCRC to a RC with the pot acting as the last RC filter with adjustable fixed bias. It is a very simple circuit & provides very clean DC voltage.
Start out with twice the required bias DC output voltage via a full-wave rectifier & capacitor. (If you need -50 vdc bias, start with -100 vdc).
Supply that semi-filtered voltage in a RCRC filter. Drop voltage approximately 15% in the first resistor & 15% in the second resistor. With 100 volts in, 70 volts is the output. Feed that voltage into a potentiometer that supplies a few millampere load to ground. The wiper on the pot is the adjustable bias voltage out. For better resolution on output voltage, the pot can be in series with another resistor to ground. The first three capacitors are 50uF and I prefer a 10uF poly as the last capacitor connected to the pot wiper output. Adjust the wiper on pot for -50 vdc in this example.
The end result is actually a CRCRC to a RC with the pot acting as the last RC filter with adjustable fixed bias. It is a very simple circuit & provides very clean DC voltage.
HI
Why don't you make use of the automatic-fixed bias...
It's a design by mattijs from machmat (www.machmat.com/sales/kits/index.htm)
you can manually set the bias current... and whatever changes (even the mains voltage changes within 5-10%) the bias current will be kept on the right level....no more milli-amp meters in your amp!!
i tried it and it was really worth it!! the sound got much better....
i'd really like to hear your comments on this
henk
Why don't you make use of the automatic-fixed bias...
It's a design by mattijs from machmat (www.machmat.com/sales/kits/index.htm)
you can manually set the bias current... and whatever changes (even the mains voltage changes within 5-10%) the bias current will be kept on the right level....no more milli-amp meters in your amp!!
i tried it and it was really worth it!! the sound got much better....
i'd really like to hear your comments on this
henk
> how do you calculate the grid resistance needed when you use combination bias?
By the percent of self-bias.
If the book says 500K self-bias, 50K fixed-bias, and you run 50% fixed bias, use 250K max grid resistance.
No reason you can't use your 0.1uFd caps.
The above derivation ignore the fact that the fixed-bias rating isn't zero ohms. I suspect you could apply the factor to both ratings and add: 250K+25K= 275K max with 50% self bias.
I don't see any point in precision math here. The max grid resistor rating is a worst-case. Look around, you find many working amplifiers using "illegal" grid resistors, and living. Mostly. 90% to 99% of tubes will have an actual grid current much lower than the one used to set the max grid resistor rating. In DIY, you can often "cheat", being prepared to replace any tubes that get unhappy with too-large grid resistors. In mass-production, such tube selection (and tube aging inside the warranty period) is just too expensive to consider, so production gear usually stays with the rating.
There is a special case for triodes with large plate resistors. Even if the grid current and voltage runs-away, they can't melt-down. You can run them without cathode bias and let the grid find its own level. 10Meg and 22Meg were common grid resistors in this use.
By the percent of self-bias.
If the book says 500K self-bias, 50K fixed-bias, and you run 50% fixed bias, use 250K max grid resistance.
No reason you can't use your 0.1uFd caps.
The above derivation ignore the fact that the fixed-bias rating isn't zero ohms. I suspect you could apply the factor to both ratings and add: 250K+25K= 275K max with 50% self bias.
I don't see any point in precision math here. The max grid resistor rating is a worst-case. Look around, you find many working amplifiers using "illegal" grid resistors, and living. Mostly. 90% to 99% of tubes will have an actual grid current much lower than the one used to set the max grid resistor rating. In DIY, you can often "cheat", being prepared to replace any tubes that get unhappy with too-large grid resistors. In mass-production, such tube selection (and tube aging inside the warranty period) is just too expensive to consider, so production gear usually stays with the rating.
There is a special case for triodes with large plate resistors. Even if the grid current and voltage runs-away, they can't melt-down. You can run them without cathode bias and let the grid find its own level. 10Meg and 22Meg were common grid resistors in this use.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.