• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Fixed or self bias?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Good day to everybody,

Im building a PP amplifier using two 6B4G per channel driven by a current sourced differential amplifier using 6N6P double triode.

If the power output is not an issue what would you all suggest to use, self bias or fixed bias for the output stage.

Thanks all in advance

Chris
 
Ex-Moderator
Joined 2003
If you have self bias then you will need to bypass the cathode resistors with capacitors. If you then overload the stage, the charge on those capacitors will change and it will take time for them to recover. Meanwhile, your output valves will not be correctly biased. Further, those capacitors are adding another LF time constant to your amplifier. If you apply global feedback, that's an issue.

Alternatively, you go fixed bias. You don't have any additional time constants and recovery after clipping is immediate. You now start to worry about whether your output valves are passing the correct current and whether there's an imbalance that is causing your output transformer to saturate prematurely.

On balance, I'd go for fixed bias. The bass sounds cleaner to me, whereas with cathode bias it can sound a little woolly. 6B4G has a sufficiently low mutual conductance that I doubt if there will be any significant drift in anode current. You could always install a quartet of nice round black moving coil meters with Gothic pointers to monitor current. Or, if you want to be a little more modern, those postage stamp panel-mounting DVMs are 199.9mV FSD, so put one in parallel with a 1 Ohm sense resistor and it will read out directly in mA. One per valve is a bit expensive, but it certainly gives peace of mind.
 
A difficult question. While I agree with all EC8010 says it's not always so clear cut. People often complain about inferior tone in fixed bias even though bass is better.

I suspect this is largely dependent upon the way the fixed bias voltage is derived. A single diode/capacitor/pot tapped off the main transformer PS is guaranteed to sound dreadful IME. After all this voltage feeds directly the grids. There are a lot of ways to get clean bias supply, none of them too easy.
 
I find fixed bias sounds a bit harder, but has more power. This is probably partly due to the higher voltage (no cathode resistor)

The great benefit of self bias is that it is incredibly easy to set up. A shared cathode resistor, bypassed by a high qulaity cap and a pair of matched tubes will sound great.

Seeing power isn't an issue, I would go for self bias, simply for the more delicate sound.


Or try both :)
 
Ex-Moderator
Joined 2004
NO peace of mind if you use a shared cathode resistor for self-bias: if one of the OP tubes stops conducting for any reason, the other becomes severely under-biased and might melt. Separate bias rsistors are much safer and assist with balancing but they need bypassing, thus requiring those nasty electrolytics that many people prefer to avoid.

Fixed bias is better IMHO. It has none of the disadvantages of self-bias and it allows you to set the operating point of your choice. In addition, it avoids wasting 30v or more of B+.
 
I'll share my experience with the same thing on a 2A3 design...

You should really study some of the older tube manuals when the tube was still active to some degree. For example, my RCA RC-16 manual still had a full characteristics listing and curves as opposed to later manuals that stuffed them in the back under the obsolete section.

Anyway, it shows a max grid resistance for the 2A3 at 500K for cathode bias and 50k for fixed bias. The Sylvania manual recommends 10k max grid resistance. What they worried about on these low mutual conductance tubes is positive charge build up on the grid. Some say it's ions from gas molecules, others say particles stripped from the plate material...whatever.

You can get away with large grid resistance (500k) with cathode bias because this bias method self compensates if the grid floats more positive, i.e. more bias voltage at cathode resistor.

With fixed bias, you don't have the self compensation, so they wanted you to use smaller grid resistance to help bleed the charge off the grid at a faster rate.

My 2A3 setup will use both, fixed for a little adjustablity, cathode for a little self compensation. It's been suggested to use a least half of the normal cathode bias resistance so that you develop sufficient compensation drop across cathode resistor.

My purpose to use a little of both is to be able to use a higher value of grid resitance. Why??? It comes down to how I want to drive the 2A3.

For typical resistance-capacitor coupled stages your anode load (plate resistor) is to be x2 times bigger that ra (anode resistance at operating point) and the following grid resistance in the next stage should be around x4 times the previous plate resistor value.

With my 2A3 amp, I wanted to try to use only one gain stage. Most designs use the 6SN7, I'm trying to use a 6SL7. I'm running Sovtek 2A3's at 20W idle (400V, 50ma) which results in more negative bias, and hence, more potential signal swing and more power out. One stage of 6SN7 doesn't have the gain I'd like but does have low annode resistance. The 6SL7 does have more gain but also has much higher annode resistance. This requires that the plate resistor and the following 2A3 grid resistor be of much higher value so that I don't load the gain stage too much.

They also say that the 6SL7 doesn't have the current abilities (drive) that the 6SN7 does, which is true. The 2A3 has relatively large input capacitance so you need some current to be able to keep up with the slew rate at high frequencies, or else you reduce your bandwidth.

So, I want to try fixed and cathode bias, use around 390k grid resistance as part of the bias circuit string, 6SL7 plate resistor of 100k (x2 ra) and still get more gain than a 6SN7 would produce and hopefully enough drive current to not roll off the high end. I'll experimentally test the bandwidth.

Hopefully you can see how important your choice is. And now as you review low transconductance designs you'll see most are cathode biased and have a clue why it's that way.
 

PRR

Member
Joined 2003
Paid Member
> use, self bias or fixed bias for the output stage

If you have to ask: "Use automatic {self} bias whenever possible". That's what Radtiotron says and I agree.

If you have to maximize Test-Tone Watts per Dollar, above a few watts fixed-bias is your best path.

----------------

A specific reason to self-bias 2A3/6B4: if the driver stage runs on the same voltage as the output stage, you will have a hard time making enough clean drive to put the 2A3s to the limit. If you self-bias, in effect the outputs work at 60V less than the driver supply voltage, and you can now (just-barely) make enough clean drive to slap the 2A3 grids to saturation and cutoff.

----------------

> you will need to bypass the cathode resistors with capacitors. If you then overload the stage

Yes, but as long as it stays in Class A, the common bypass cap isn't doing much at all. You can often remove it.

As you shift up into Class AB, now the cap is working, and working hard as you approach overload. But the question states "power output is not an issue". With hot speakers, a couple P-P 6B4 can loaf far below the AB line most of the day.

> NO peace of mind if you use a shared cathode resistor for self-bias: if one of the OP tubes stops conducting

Hmmmm.... you are of course quite right. But in decades of tube work and lore, I never ran into that argument before.

Tube stuff fails, and is made to be repaired.

Tube failures are not the most common failures (caps die a lot).

Working push-pull, you don't have to set your idle plate dissipation up to the maximum to get full power; if dissipation rises a bit due to a one-side failure, you are still probably not far over the rating.

I'm also cheap and lazy. So I would not do the separate bias resistor thing. More to buy, more to wire, more to go wrong. Opinions can differ on this point; that's mine.
 
Since the subject of a clean negative bias supply has been raised, could the brains on the forum suggest a compact way of getting such a clean supply? I'm in the middle of building a bias supply so any schematics and ideas would be very welcome right now. I need up to -50v for the driver stage and -15v for the input diff stage CCS. Up to now I've been using diodes>cap>50k pot(to output tubes)>-15v voltage reg>input CCS. Pretty basic, and plenty room for improvement! Andy
 
I would make the selection on a case-by-case basis. I'm doing a design using 807s, and went with fixed bias. As you can see from this snippet from the STC application report, fixed bias has a significantly lower THD, output impedance, and somewhat more power out for identical voltage and current. In this case, fixed wins for multiple reasons. As for deriving bias voltage, that's NBD these days since a solid state regulator isn't that difficult to build, and these are much quieter than the method from the "good ol' days": half wave diode/ripple capacitor/pot from one half of the main plate xfmr.

It isn't quite as easy, but then again, best to design for good operation before applying the global feedback. That way, you just might need less of it. Since this is a one of a kind project, and I'm not anticipating building these by the thousands, it doesn't matter if it requires a few more parts.
 

Attachments

  • 807-chars.gif
    807-chars.gif
    14.4 KB · Views: 604
Miles Prower said:
I would make the selection on a case-by-case basis. I'm doing a design using 807s, and went with fixed bias. As you can see from this snippet from the STC application report, fixed bias has a significantly lower THD, output impedance, and somewhat more power out for identical voltage and current.


But the operating voltage for the tube (plate to grid) is not identical. In the first case, the tube is idling at 32 watts plate dissipation and in the second it's at 29 watts (similar ratio for the screen). Would the two results be identical if you raised the B+ to compensate for the 22v rise in the cathode voltage?

sheldon

This raises a novice question. Is plate voltage considered from plate to ground or plate to cathode? Plate to ground doesn't seem logical to me, as ground can be whatever you want it to be. the tube doesn't know where ground is. On the other hand, the tube does know where the cathode is relative to the plate. In other words, plate voltage seem like a different thing than B+.
 
But the operating voltage for the tube (plate to grid) is not identical. In the first case, the tube is idling at 32 watts plate dissipation and in the second it's at 29 watts (similar ratio for the screen). Would the two results be identical if you raised the B+ to compensate for the 22v rise in the cathode voltage?

Don't know; don't care. Take another LQQK at those distortion figures: 1.8% with fixed bias vs. 4.0% with cathode bias. That, alone, is a good enough reason to prefer fixed bias.

This raises a novice question. Is plate voltage considered from plate to ground or plate to cathode? Plate to ground doesn't seem logical to me, as ground can be whatever you want it to be. the tube doesn't know where ground is. On the other hand, the tube does know where the cathode is relative to the plate. In other words, plate voltage seem like a different thing than B+.

Always measure from plate to cathode. Now, there are a great many characteristic charts that'll tell you that they are measuring B+. Of course, this would apply with the cathodes connected to ground. If they are not, then it's plate-to-cathode anyway. Many tube designers will talk about "B+" since, unlike solid state, the B+ to ground will be significantly greater than any positive cathode bias voltage. After all, 20V against 400V is less than a 10% difference. Nevertheless, the cathode is your reference point, just as the emitter/source is when dealing with solid state.
 
Miles Prower said:
But the operating voltage for the tube (plate to grid) is not identical. In the first case, the tube is idling at 32 watts plate dissipation and in the second it's at 29 watts (similar ratio for the screen). Would the two results be identical if you raised the B+ to compensate for the 22v rise in the cathode voltage?

Don't know; don't care. Take another LQQK at those distortion figures: 1.8% with fixed bias vs. 4.0% with cathode bias. That, alone, is a good enough reason to prefer fixed bias.

This raises a novice question. Is plate voltage considered from plate to ground or plate to cathode? Plate to ground doesn't seem logical to me, as ground can be whatever you want it to be. the tube doesn't know where ground is. On the other hand, the tube does know where the cathode is relative to the plate. In other words, plate voltage seem like a different thing than B+.

Always measure from plate to cathode. Now, there are a great many characteristic charts that'll tell you that they are measuring B+. Of course, this would apply with the cathodes connected to ground. If they are not, then it's plate-to-cathode anyway. Many tube designers will talk about "B+" since, unlike solid state, the B+ to ground will be significantly greater than any positive cathode bias voltage. After all, 20V against 400V is less than a 10% difference. Nevertheless, the cathode is your reference point, just as the emitter/source is when dealing with solid state.

That's what I thought. But the chart in the original post seems to indicate plate to ground, or does it? It has a blank for grid voltage. Does that mean 0 or? If it means 0, then the operating conditions for that tube are different than for the fixed bias example. My question was, could those different operating conditions (plate/cathode, screen/cathode voltages) explain the results, at least partly.

If we assume that B+ is raised 20volts, to keep the plate/cathode and screen/cathode the same for both cases, it seems that the bypass cap is then solely responsible for the distortion. If that's the case, I would think that those results would only apply to the the specific cap used and might well be different with a different cap. Or is there something else?

Sheldon
 
That's what I thought. But the chart in the original post seems to indicate plate to ground, or does it?

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. It's a bit of carelessness that couldn't happen when dealing with low voltage, high current solid state devices where a few volts on the emitter/source will make a big difference since those few volts are a significant percentage of the collector/drain-to-ground voltage. Increasing the screen/plate voltage-to-ground to provide the same voltages referenced to the cathode would make the currents more equal in both cases.

It has a blank for grid voltage. Does that mean 0 or?

This is a common convention to indicate that the bias is derived from a cathode resistor, not from an external source. As for the increased distortion, this may very well be topology dependent. A common cathode bias resistor may very well give different results from one resistor per 807. Also, in those days, they didn't have access to the better capacitors we have now.
 
andyjevans said:


Hello there - I regret to say that I need a bit more information to see this circuit (especially at 1 in the morning...). could you give me some more details?

Here's a sketchy outline. The app notes from National Semiconductor go through the design in reasonable detail. For simplicity, I've eliminated bypass caps and protection diodes, but all the stuff you need to understand the circuit is here. Basically, the PNP transistor has its base held at a zener voltage more negative than the output voltage (e.g., if you use a 6.2V zener and the output is -50V, the transistor's base is at -56.2V). The emitter (and thus the regulator input) is 0.7 volts more positive. This guarantees that the regulator doesn't see an excessive input-output voltage differential.

The two resistors attached to the 337 set the output voltage. It's really pretty simple, so much so that it is scorned by the hair-shirt audiophile brigade. Nonetheless, it works and it works damned well.
 

Attachments

  • bminusreg.gif
    bminusreg.gif
    3.6 KB · Views: 608
Miles Prower said:
That's what I thought. But the chart in the original post seems to indicate plate to ground, or does it?

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. It's a bit of carelessness that couldn't happen when dealing with low voltage, high current solid state devices where a few volts on the emitter/source will make a big difference since those few volts are a significant percentage of the collector/drain-to-ground voltage. Increasing the screen/plate voltage-to-ground to provide the same voltages referenced to the cathode would make the currents more equal in both cases.

It has a blank for grid voltage. Does that mean 0 or?

This is a common convention to indicate that the bias is derived from a cathode resistor, not from an external source. As for the increased distortion, this may very well be topology dependent. A common cathode bias resistor may very well give different results from one resistor per 807. Also, in those days, they didn't have access to the better capacitors we have now.

Thanks,

One final check on my logic assumptions. With one resistor/capacitor per tube, and if we had a perfect capacitor, I assume that self bias would be essentially identical to fixed bias.

Sheldon
 
One final check on my logic assumptions. With one resistor/capacitor per tube, and if we had a perfect capacitor, I assume that self bias would be essentially identical to fixed bias.

Perhaps. There is one big difference: if the finals are over driven, then an unusually large voltage can build up across the cathode resistor bypass. This can take quite some time to discharge, and until it does, it throws off the operating point. This can lead to excessive distortion until the bias returns to normal. You avoid this with fixed bias.

As for which to select, that goes on a case-by-case basis.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.