Bias question

[VMS]

Hi guys!

I was watching clips of the new marshall 2525 mini jubilee amp which is 20 watt amp with EL34 tubes and it got me thinking about biasing.

So if you make a fixed-bias class AB amp that is lower wattage than the tubes used could produce do you still idle bias it to the 70% of the max dissipation of the tube?


ps. I have no idea how marshall did their amp and if it's fixed- or cathode-biased, class A or AB, that amp just made me wonder this thing.

[HotBluePlates]

I was watching clips of the new marshall 2525 mini jubilee amp which is 20 watt amp with EL34 tubes and it got me thinking about biasing.

So if you make a fixed-bias class AB amp that is lower wattage than the tubes used could produce do you still idle bias it to the 70% of the max dissipation of the tube? ...

Not necessarily. The notion that "70% dissipation" is somehow a "rule" for class AB biasing is a myth. It might be a handy guideline, but that's about it. The myth is built upon the "amp guru" phenomenon and layers of (presumably well-meaning) logical extension. The real truth is you should bias the amp to whatever bias voltage the amp designer chose, as long as the tubes used also don't have an average dissipation over the entire signal cycle which exceeds their dissipation rating. The problem is we often don't know what that correct bias voltage is (though I suspect the bias voltage indicated on 60's Fender schematics to likely be this same number, assuming non-defective tubes).

-  Sometime in the late-80's Ken Fischer put in print an observation that a method used for biasing at the time (the o'scope method) did not necessarily give the best results and was subjective (therefore not easily-repeatable). So Fischer advocated measuring tube plate current to set bias, as it was an objective, repeatable number. But Fischer never said "what number" should be used.

-  Gerald Weber popularized Fischer's information/guidance. In his books/articles, Weber also tried to avoid giving a specific idle current number. But after spilling so much ink telling everyone they needed to use the OT Shunt method to bias their amps by measuring plate current, he acknowledged readers demanded a number to use. In at least one article (which was then incorporated into Desktop Reference of Hip Vintage Amplifiers), he said there isn't one number but provided "35mA" as a safe guess. At that point, everyone assumed they needed to bias their amps to 35mA to "have tone" (but note that 6V6's in blackface Fender amps are operating at a high enough plate voltage that 20-22mA is closer to an upper limit in that case).

-  Internet discussion forums brought up the issue that "35mA" was too hot for some amps/output tubes (like our 6V6 Fenders above) and would bias a designed-class AB amp over 100% dissipation at idle. The "70% idle dissipation rule" was then promulgated, but I never saw a satisfactory explanation for "why 70%"? That said, I've biased amps that way at times just to get it done.

[HotBluePlates]

The definition of Class A is that the tubes conduct during the entire signal cycle, meaning they never shut off. It is also "poor engineering practice" (meaning "wasteful of resources) to use bigger tubes than you need, and not try to get all the power output you can for the operating conditions (i.e., don't choose a KT88 to deliver 2w of output power). As a result, where Class A is feasible and the designer tries to get every watt of output possible, the output tube is biased to 100% dissipation.

The definition of Class B is the tubes conduct for exactly 1/2 the signal cycle. Normally, this means the tube would be biased exactly to the point of idle current cutoff, but tubes are hard to turn off; it takes more bias voltage to go from nearly-off to "cut-off" than to go from nearly-off to full-on. So in practice, Class B tube amps are biased on but at very low plate current (a number up to 5-10mA, but maybe less).

"Class AB" means literally "anything between Class A and Class B." It is usually stated as, "output tubes conduct for more than 180-degrees (half) of the signal cycle but less than 360-degrees" (meaning the tubes cut off some amount of the time). If you look at it in terms of idle current and assume tubes are perfect devices that easily bias just to the point of cutoff, Class AB is then an idle current between 0%-100% dissipation.

In light of that, how rigid could that 70% idle dissipation possibly be?

A designer would design an output stage based on assumptions about the amount of output power needed, the supply voltage that's likely to be available, and a minimum number of just-big-enough tubes to get the desired output power. There are simplified ways to get a first-estimate of the ideal load impedance. From that, another simplification takes impedance estimated and the supply voltage available to estimate plate signal current for full output power. The estimate are compared against "data sheet reality" and the value of voltage, impedance and signal current are tweaked to get a workable output stage with the "average tube" described by the data sheet.

One way to arrive at the bias voltage needed for this output stage is to note the transconductance (Gm) of the tube in question, in the area of the likely operating point. Americans typically specify Gm in terms of micromhos, but the European unit is more-directly informative: mA/V (milliamps-per-volt). If you know the max-signal plate current you'll need, then (max signal current)/(tube Gm in mA/V) = required grid signal for max output power.

The designer would probably try to shoot for the max output power being as clean as possible, and also note that many output tubes have an onset of grid current (and heavy distortion) once the grid voltage rises to about -1v. So the first-pass output tube bias will likely be a couple-volts more than the voltage derived by dividing max signal current by tube Gm.

Then the designer has to evaluate whether the chosen bias voltage and other operating conditions will keep the tube from overheating in use. There's a fairly tedious graphical method of picking five-to-eleven grid voltage points during the signal cycle at max drive & output power, and determining the plate voltage & plate current for each of those points, then applying a mathematical formula to find average plate dissipation over the signal cycle. If the average dissipation is too high, the supply voltage, load impedance (and indirectly, the plate current) and/or the bias point get changed and the whole design evaluated again.

At the end of all this, the designer arrives at a design bias voltage, with implied idle current. Nowhere in this did "70%" ever come up; however, there are undoubtedly amps which coincidentally have a design bias point which idles at 70% of the plate dissipation rating.

[alerich]

It's a good question. I have been wondering about it, too. There are numerous "lunchbox" style amps on the market but most use a duet of EL84 output tubes so they are biased in the conventional manner. However,  that new Marshall uses a pair of EL34 power tubes and the recent Soldano Hot Rod 25 uses a pair of 6L6 tubes. Not much tech info yet on the Marshall Silver Jubilee mini head. The Soldano reportedly utilizes a lower B+ voltage to reduce the power output borrowing from the EVH variac trick. Don't know if the preamp voltages are similarly lowered. Haven't seen any schematics of the HR25 on the net.

[jjasilli]

I can't find a schematic but apparently this amp is SE (see: http://www.thegearpage.net/board/index.php?threads/marshall-mini-jubilee.1663577/page-8)


That's 2 EL34's in SE.  To get Class A a low plate voltage is needed.  Per the tube charts, for 250 - 350V, there's 11W output (for one tube); http://www.mif.pg.gda.pl/homepages/frank/sheets/129/e/EL34.pdf  So 2 tubes in parallel could yield about 2X that. 


Per the chart, at 250V I think plate dissipation is about 22W per tube (someone please verify).  Max is 25W.  25W / 22W = 88%.  For Class A 90% bias is OK.  (70% is for PP).  But in concept, I think the answer to the original question is yes.  That is, regardless of choosing a high or low plate voltage, figure bias so as to stick with the applicable percentage of max plate diss.

[HotBluePlates]

I can't find a schematic but apparently this amp is SE (see: http://www.thegearpage.net/board/index.php?threads/marshall-mini-jubilee.1663577/page-8)


That's 2 EL34's in SE.  ...  Per the tube charts, for 250 - 350V, there's 11W output (for one tube); http://www.mif.pg.gda.pl/homepages/frank/sheets/129/e/EL34.pdf  So 2 tubes in parallel could yield about 2X that. ...

Sounds reasonable.

Per the chart, at 250V I think plate dissipation is about 22W per tube (someone please verify).  Max is 25W.  25W / 22W = 88%.  For Class A 90% bias is OK.  (70% is for PP). 

The data sheet you linked idles at 25w for both the 250v and 300v conditions (I call the 24.9w plate dissipation of the 300v condition "25w").

My point in my earlier post is there are unstated assumptions which are not hard & fast rules. One of these assumptions is, "I want all the power output possible for my given class of operation."

If you're willing to get less than "all the power" you can idle to a lower dissipation and still stay in Class A (this itself assumes the output tube can't/won't be driven hard enough to ever cut off). So it's possible to run an EL34 in class A with 550v on the plate idling at 20% plate dissipation to get 1/2w of output power using a 10kΩ plate load. You just won't see any production amp doing that because the heater consumes 9.5w of power by itself (we're not considering screen power sunk), so it seems a waste to only get 1/2w of output power from such an output stage.


I may have lost track on whether we're trying to reverse-engineer the Mini Jubilee or if we're talking about biasing in general terms.

[jjasilli]

Agreed.  The stated plate diss ratio's are the general rule of thumb for max safe operation of the tube.  It is not necessary to max out the hot operation of the tube (though at some point cold bias will shut-of the tube).  Tonal bliss is somewhere in that operational range, subject to personal preference.


Also, there are the general, down and dirty, rules of thumb.  Hotblue is showing deeper analysis.


Note for Class A with cathode bias, IMHO supply voltage is the determining factor for bias.  Because, as you change the value of the cathode R, the plate voltage-tube current relationship in the circuit keeps self-adjusting around the same bias point.  Class A fixed bias is an alternative (which may be what this amp is using).

[VMS]

I may have lost track on whether we're trying to reverse-engineer the Mini Jubilee or if we're talking about biasing in general terms.

Although I have read somewhere that tubes don't have sound it would be nice to make smaller wattage amp with big bottles with different flavor. This is easy in champ like circuits where you can change tubes and keep the same cathode resistor and it will work.

But in fixed-bias amps you usually set the bias thinking about the tube used and not the amp.

For example what if I wan't to use 6l6 tubes in deluxe reverb, would it be better to bias the tubes to 70% of maximum dissipation or set the bias voltage to -35V as in the schematic?


I think in power scaling amp the bias voltage is tracking the plate or screen voltage but i'm not sure in what proportion.
 

[Willabe]

So if you make a fixed-bias class AB amp that is lower wattage than the tubes used could produce do you still idle bias it to the 70% of the max dissipation of the tube?

[Willabe]

I may have lost track on whether we're trying to reverse-engineer the Mini Jubilee or if we're talking about biasing in general terms.

So if you make a fixed-bias class AB amp that is lower wattage than the tubes used could produce do you still idle bias it to the 70% of the max dissipation of the tube?

[HotBluePlates]

Thanks Willabe!

Then in PRR-fashion: 

So if you make a fixed-bias class AB amp that is lower wattage than the tubes used could produce do you still idle bias it to the 70% of the max dissipation of the tube? ...

Not necessarily. ...

[jojokeo]

I may have lost track on whether we're trying to reverse-engineer the Mini Jubilee or if we're talking about biasing in general terms.

Although I have read somewhere that tubes don't have sound it would be nice to make smaller wattage amp with big bottles with different flavor. This is easy in champ like circuits where you can change tubes and keep the same cathode resistor and it will work.

But in fixed-bias amps you usually set the bias thinking about the tube used and not the amp.

For example what if I wan't to use 6l6 tubes in deluxe reverb, would it be better to bias the tubes to 70% of maximum dissipation or set the bias voltage to -35V as in the schematic?


I think in power scaling amp the bias voltage is tracking the plate or screen voltage but i'm not sure in what proportion.

You have to keep in mind your transformers' specs before you willy nilly wing things around. Then as these parameters change, so to will the primary impedance or your OT. Lastly, the point of all of HBP's input is that you don't have to just go for 70% of pDiss or the -35mA - it's good for a starting point maybe but let your ears and/or arithmetic be the guide. As a side note - if you don't care about replacing tubes often try frying the bejeezus out of them or keep 'em backed off just out of the "red plate" area. But you may get the tone you like and after calculating things the pDiss is down to 50%!?

[jjasilli]

Well, a system tends to work best when pushed toward near, but not over the limit.  With too cold bias the tubes may lapse into inter-modulation distortion, and eventually shut-off as bias is set colder.  Probably will sound best with "warm" to hot bias. 


It seems that the original post inter-mixed general bias questions, with the jubilee amp in particular.  But we don't have much info on that amp, so it's hard to reverse engineer. 

Although I have read somewhere that tubes don't have sound . . .

I dare you to tell that to a tube roller.   

[HotBluePlates]

Although I have read somewhere that tubes don't have sound it would be nice to make smaller wattage amp with big bottles with different flavor. This is easy in champ like circuits where you can change tubes and keep the same cathode resistor and it will work. ...

Joe provided good input on this.

About the "Champ with Big Bottles" you mentioned:
This is the case with the THD UniValve. The circuit was essentially a Champ, and used a cathode resistor. Whatever tube you plug in settles to a bias point determined by the tube & resistor. When you change tubes (even of the same type), there is a good chance the specific voltage across the cathode resistor will change.

The same point I made earlier of a signal peak which drive the output tube grid to ~-1v (relative to the cathode) will be about the limit of clean power. So whatever tube you plug in which yields the biggest voltage across the cathode resistor will require more drive from the preamp to hit the same level of dirt. Some will interpret that as "more headroom" because the volume control had to be turned higher. My opinion is that since the OT load/supply voltage hasn't changed, the power output at the onset of dirt hasn't changed either, so the headroom increase is superficial.

Either way, different broad tube types sound different. Tubes which are true pentodes (EL34, EL84, 6K6) have a different sound that beam power tubes (6550, 6L6, 6V6, and KT66, KT88). The shape of their characteristics means they tend to distort in different ways.

If you're looking for audible changes with an output tube change, ideally you'd go for no negative feedback. Speaker bass response will likely be loose as a result, but feedback will tend to squash the differences among output tubes.

... But in fixed-bias amps you usually set the bias thinking about the tube used and not the amp.

For example what if I wan't to use 6l6 tubes in deluxe reverb, would it be better to bias the tubes to 70% of maximum dissipation or set the bias voltage to -35V as in the schematic? ...

According to my original argument, this is only because we're taught to think the wrong things about biasing.

If you want to use 6L6's in your Deluxe Reverb, just plug 'em in (as Jojokeo said). If the PT can support the extra heater current, you might be done.

The -35v was for a design using 6V6's, and at that supply voltage they're idling around 20-22mA. If you compared to a tweed Deluxe, the self-set cathode bias will settle on many tubes around 18-21v. Supply voltage is lower and tube current is higher.

But if you wanted to get more actual output power from the 6L6's in the Deluxe Reverb, you've probably changed power & output transformers to enable that extra power. Now a bias nearer -45v, as is more typical with other 6L6 Fenders, would likely be suitable.

If you bias with a small fixed voltage for hotter output tube operation, what happens? Assuming the tubes don't overheat in use, a smaller driving signal from the phase inverter pushes the output tubes to the limit of clean output power (smaller span from bias voltage to -1v on the tube grid). The smaller voltage swing times the tube's Gm means smaller current swing through the OT primary. So the amp seems to break up faster.

Nothing magic happened in the tube, you just changed it's operating condition away from that which delivers the most clean power.

... I think in power scaling amp the bias voltage is tracking the plate or screen voltage but i'm not sure in what proportion.

What I wrote above could be misconstrued into an approach towards power scaling, as I mentioned "less clean power output". Power scaling makes for a much more dramatic drop in power output.

To answer your question, in power scaling a fixed-bias amp, you set the bias with a bias pot as normal. When the power scale control is turned down to reduce supply voltage, there is a circuit which tracks the reduction of the output tube screen. that reduction is proportionally fed to reduce the voltage input to the overall bias circuit, and without twiddling a pot your bias voltage becomes a smaller value in-step with the reduced screen voltage.

So in those you set the bias one time for a given set of output tubes, and leave it alone when adjusting the output power.

[VMS]

and without twiddling a pot your bias voltage becomes a smaller value in-step with the reduced screen voltage.

So how does one decide how much bias voltage should drop if for example screen voltage drops from 400V to 100V?

[jjasilli]

Champ with Big Bottles:  I put 6L6's in my SF VibroChamp & SF Princeton (since sold).  The VibroChamp also got an OT upgrade and both amps got bigger speakers.  So that's a number of variables, but IMHO the big bottles make the amp sound more robust.

[2deaf]

So how does one decide how much bias voltage should drop if for example screen voltage drops from 400V to 100V?

I've never made a Power Scaling device that had the bias track the screen voltage - - they always have the bias track the plate voltage. 

As the plate voltage decreases, the percentage of a tube's maximum dissipation at idle should also decrease.  So if you biased the tubes at 70% at 400V, you might bias them at 25% at 100V.  If you try to maintain 70% of maximum dissipation at lower voltages, you will reach a bias voltage of zero at some point and if you went further, you would need a positive bias voltage.  Also, if you try to maintain 70% dissipation, the signal will develop some ugly crossover distortion at lower plate voltages.

 

 

[HotBluePlates]

So how does one decide how much bias voltage should drop if for example screen voltage drops from 400V to 100V?

I've never made a Power Scaling device that had the bias track the screen voltage - - they always have the bias track the plate voltage. ...

Fair statement. But if you're power scaling your output tube plate voltage, the screen voltage is dropping too, right? 

and without twiddling a pot your bias voltage becomes a smaller value in-step with the reduced screen voltage.

So how does one decide how much bias voltage should drop if for example screen voltage drops from 400V to 100V?

So here's why I talked about power scaling tracking the bias voltage in relation to screen voltage:
In a beam power tube or pentode, the plate current is not really a function of plate voltage, but rather of screen voltage & control grid (G1) bias. So when you're deciding on the bias of a pentode/beam power tube, it's the screen voltage that matters.

"Mu" (µ) with your basic triode is the ratio of a change in plate voltage divided by a change of grid voltage, with plate current held constant. In other words, how much more effective the change of G1 voltage is at controlling plate current than a change of plate voltage.

Pentode/Beam Power tubes have a "Triode Mu" or a "µ_g1-g2" which is just like Mu in a regular triode, but is the measure of how much more effective G1 voltage is at controlling plate current than G2 (the screen) voltage.

Theoretically, G1 could cut off plate current at a bias voltage of (Screen volts)/(µ_g1-g2). So if the tube in question has a µ_g1-g2 = 10 and a screen voltage of 400v, plate current will be near cut-off with a G1 bias of -40v. If the same tube's screen voltage is dropped to 100v, cut-off will happen near a bias of -10v.

Assuming a few things about our theoretical design, if the amp is intended to be operated Class A then we might have an idle bias about halfway between cutoff and 0v of bias. The old books often throw out the formula 0.6 * (Screen volts)/(µ_g1-g2), as this is biased just a little hotter than halfway between cutoff & saturation. For our tube above and a screen voltage of 400v, the estimated bias would be 0.6 * (400v)/(10) = -24v.

Then again, this is just one of those starting-point rules of thumb... From here a designer would have to go back & look at the data sheet, figure how much plate current will pass with the bias provided by the formula, and double-check/adjust as necessary.

For a Class AB amp, you still only know that the bias for our same theoretical output tube should be something between -24v (Class A recommendation) and -40v (near cutoff). The proposed design has to be evaluated over the whole signal cycle to determine whether the output tubes cool enough during their shut-off-time to compensate for the big current peaks when they're conducting.

Or you go the experimental route Jjasilli recommended, and watch for signs of redplating tubes (which may require a dark room for "just starting to redplate").

If you listen to an amp while fiddling the bias, I think you'll find a broad range of different idle currents where the amp sounds essentially unchanged.

[2deaf]

Fair statement. But if you're power scaling your output tube plate voltage, the screen voltage is dropping too, right?

Hey!  I thought I was on your "do not reply" list.

I never wrote down any screen voltages when I was experimenting with power scaling.  I just figured the screens were always within a few volts of the plates.  You must know about a bias-modulating circuit that I am not familiar with, which would be just about all of 'em that are active.

What I did was to take a particular parameter and experimentally determine what bias voltage would get the desired characteristic of that parameter at various plate voltages.  One was the breaking point between a clean sine wave all the way to clipping and a sine wave with crossover notch right before clipping.  When the bias voltage was plotted against plate voltage, it turned out to be a linear function with V_Bias=11.5-(0.123 * V_Plate) for one amp.  Other parameters also yielded linear functions, but the slopes and zero crossings were different.

The upshot here is that you would not use 70% of the maximum dissipation for biasing if you are achieving less than full potential from a tube by using lower B+ voltages.

[PRR]

A pair of EL34 for just 20W output?

That's taking the dump-truck to get beer. A 2-pack of '34 can carry 50, 70, maybe 100W out.

OK, yes, one '34 Single-Ended is 8 to 11W, so could be two tubes (parallel or push-pull) deeply stuck in Class A operation. In which case the idle must be 24.99 Watts (and any '34 you have will do that for years).

> choosing a high or low plate voltage

Philips (many tube makers) picked roughly 250V whenever possible because it is convenient and economical. In Class A, to a point, higher voltage is not more power, it forces lower current and much higher load impedance (many turns of fine wire).

Fender had old radio PTs which with field-coil and filtering gave ~~250 to a full radio, but 350V without field-coil and the lower drain of just-an-amp. That's why all those 6V6 were worked "over rated voltage".

You "can" run much higher plate voltage. '34 can stand 800V (NO more!) on plate. To stay under 25W Pdiss the current must be 31ma, the likely load impedance near 26K. 26K begs for winding problems, also treble problems (self-resonance within the audio band). You still get around 10 Watts output (maybe 11W at high THD). A 250V or 350V design can do the same, better, cheaper.

Also as you raise plate voltage and reduce current, you really want to turn-down G2 voltage. High voltage all around invites MASSIVE current flow. High G2 voltage puts "kinks" in the low-volt zone. Also high G2 voltage forces large G1 voltage which is a drag. I suspect a 800V 31mA amp would like Vg2 nearer 125V. G2 current is substantial and variable. This adds complication to power supply.

If this new Marsh is push-pull, they may have re-discovered a condition I have posted here. 400V supply, 6.6K CT loading, 2K common G2 resistor, 250(?) common cathode resistor, ~~100mA total. Gives right-around 20 Watts clean. Interestingly, a little care in pin-out allows "any" common tube bigger than 6V6 to work-- 6L6, EL34, 6550, 7027..... and no real need for matching (metal 6L6 teamed with 6550 worked fine).