Tube Biasing Questions

[Melodeath00]

Thanks, moamps. As it turns out, I was measuring current completely incorrectly. And in fact, I'm lucky I didn't damage my mics (as far as I know) by shorting the circuits out from what I understand. Well, at least both mics sound ok still!

 

[Matador]

Rp=dVa/dIa (Vg=const.), not just Va/Ia for one operating point.

Melodeath, Rp isn't plate resistance, it's plate impedance. 

moamps has it: Rp is the first derivative of the reciprocal of the transconductance curve.  If you remember your calculus, it is the slope of the tangent line where the load line intersects each of the curves in my picture.

If you eyeball it, the slope doesn't change much from curve to curve through that particular load line.  Yes, the current does change, but the delta change in resistance doesn't change much.

 

[Melodeath00]

Matador,

Do we use the load line to figure out what plate resistor to use? For example, if I want less anode voltage, how do I calculate what resistor I need?

I thought I figured it out by following the load line, but it seems to not always work. (Maybe it only works sometimes by coincidence. If you follow the load lien of a VF14 with .55mA current at 35V, the load line is about 127k, and in fact we use 130kohm total plate resistance in U47). With an EF800 I'm getting 50V anode with 120V B+. following that load line you get about 1.5mA at 0V, which supposedly gives you a plate resistance of 80kohm, but I'm actually using roughly 150kohm plate resistor when I get 50V on plate. I think I am confused somewhere.

Thanks so much

 

[Matador]

The load line is the plate resistance.  It's slope is actually 1/(plate resistor).  It's aline because the resistance (shouldn't) change as a function of voltage and current...voltage and current are strictly determined by the plate resistance in a linear fashion (via Ohm's Law of course).

However in this analogy the tube is a voltage controlled impedance.  The grid voltage w.r.t the anode voltage determines the "equivalent" resistance of the tube at those exact operating points.  It's no different that if you substituted a pot in for the tube, and dialed it up and down depending on what the grid was saying.  Conceptually we can think of the pot as having zero resistance when it's fully "dialed down", and maximum resistance when it's fully dialed up.  If you measured at points inbetween you would read something between those two extremes.

However in this case, the tube isn't a perfect linear element.  The curves are "bendy", which means that the tube isn't a linear impedance as a strict linear function of the grid voltage.  But it kind of looks "linearish" at certain points, so we try to pick a load line in the "linearish" points on the curves.

Hopefully that makes more sense.

 

[dmnieto]

Matador, would you mind helping me with understanding the biasing of the SRPP stage in the G9? The reason is that I would like to be able to burn in the tubes I have selected for your C12 clone, using a G9 preamp.

If I understand correctly on a SRPP stage with no input each section drops a voltage of Ht/2, in this case 245/2=> 122.5V. Eyeballing on the 12ay7 spec, I believe that it would have a biasing point at iddle of:

Vg ~ -2V
Ia ~ 1.3mA
Vplate ~ 120V

Are my calculations ok? Would be this a good idling point to burn in the tube?

For comparison, my calculations tell me that for the original ECC82 tube the biasing point is

Vg ~ -5V
Ia ~ 3.3mA
Vplate ~ 118V

 

[Melodeath00]

The load line is the plate resistance.  It's slope is actually 1/(plate resistor).  It's aline because the resistance (shouldn't) change as a function of voltage and current...voltage and current are strictly determined by the plate resistance in a linear fashion (via Ohm's Law of course).

However in this analogy the tube is a voltage controlled impedance.  The grid voltage w.r.t the anode voltage determines the "equivalent" resistance of the tube at those exact operating points.  It's no different that if you substituted a pot in for the tube, and dialed it up and down depending on what the grid was saying.  Conceptually we can think of the pot as having zero resistance when it's fully "dialed down", and maximum resistance when it's fully dialed up.  If you measured at points inbetween you would read something between those two extremes.

However in this case, the tube isn't a perfect linear element.  The curves are "bendy", which means that the tube isn't a linear impedance as a strict linear function of the grid voltage.  But it kind of looks "linearish" at certain points, so we try to pick a load line in the "linearish" points on the curves.

Hopefully that makes more sense.

I think it makes sense. At the low mA portion of the tube graphs, the curves are bendy and not linear. We can pick an operating point by moving upwards towards the more linear region. Is this what you're saying? Looks like typically the plate impedance will increase as the Grid gets more negative. So the plate impedance is constantly shifting around as we inject a signal.

So if we draw a load line, we get the mA if we drop all the voltage across the plate resistor (0V). This tells us the plate resistance. Is it just coincidence that it seems to match up relatively close with the total resistance in the U47 and the C12?  Why am I getting strange numbers with other tubes like the EF80? I'm afraid I must still not be understanding.

For example, in the EF80, this uses 150k plate resistor in an alternate U47 circuit. If you divide 120V (B+) by 150k, you get 0.8mA. But the linear load line is actually up at 1.5mA or so at 0V, not 0.8mA.

Thanks again for your help

 

[Melodeath00]

I am confused again. I'm experimenting with underheating my EF802 tube from 5.67V down to 4.8V.

I have a total of 151kohm (51k and 100k in series) on the plate, and B+ is 120V.  So if I'm drawing a load line, this puts the current at about 0.8mA for 0V.  My cathode is at 1.1V.  I am measuring plate voltage at 50V.

When I look at the graphs for an EF80 with 5V heater, in order to pass through a bias of 1.1V, it looks like the plate voltage would be 35V, but I'm measuring 50V. I just don't understand. What am I doing wrong?

 

[Matador]

I am confused again. I'm experimenting with underheating my EF802 tube from 5.67V down to 4.8V.

I have a total of 151kohm (51k and 100k in series) on the plate, and B+ is 120V.  So if I'm drawing a load line, this puts the current at about 0.8mA for 0V.  My cathode is at 1.1V.  I am measuring plate voltage at 50V.

When I look at the graphs for an EF80 with 5V heater, in order to pass through a bias of 1.1V, it looks like the plate voltage would be 35V, but I'm measuring 50V. I just don't understand. What am I doing wrong?

Did you take those curves from the exact tube in the circuit?  Every tube isn't guaranteed to curve out exactly with the datasheet:  in fact, the variance may be quite large.

I don't think you are doing anything wrong:  it's looks like it's in the ballpark.

 

[Melodeath00]

I am confused again. I'm experimenting with underheating my EF802 tube from 5.67V down to 4.8V.

I have a total of 151kohm (51k and 100k in series) on the plate, and B+ is 120V.  So if I'm drawing a load line, this puts the current at about 0.8mA for 0V.  My cathode is at 1.1V.  I am measuring plate voltage at 50V.

When I look at the graphs for an EF80 with 5V heater, in order to pass through a bias of 1.1V, it looks like the plate voltage would be 35V, but I'm measuring 50V. I just don't understand. What am I doing wrong?

Did you take those curves from the exact tube in the circuit?  Every tube isn't guaranteed to curve out exactly with the datasheet:  in fact, the variance may be quite large.

I don't think you are doing anything wrong:  it's looks like it's in the ballpark.

That is a chart someone posted in a VF14 substitute measurement thread. I was just using it as a guideline, as the EF80, EF800, and EF802 are very closely related, from what I understand. 50V seems quite different from the "expected" 35V, but I'm no expert, quite obviously haha.

I was trying to use the load line to estimate the 2nd order harmonic distortion (using voltage swing) and compare to the VF14. It looks like my tube is so different from that EF80 chart that I can't, though. Again, I may be doing something wrong/misunderstanding

 

[Matador]

Think of it this way:  at 150K, it only takes about about a 100 uA change in current to cause a 15V difference.

 

[Melodeath00]

Think of it this way:  at 150K, it only takes about about a 100 uA change in current to cause a 15V difference.

Hmmm... and what would be causing this 100uA change in current? Maybe the hum going into the grid due to the headbasket and body sleeve being off? I am just guessing here haha. I guess that 0.1mA is equivalent to .5V or less difference in the grid, so maybe that makes sense? In other words, is my 50V measurement not reliable at all? If I'm right, why does it always give me 50V, and never 20V (where you'd expect it to be according to the load line for the positive swing)? Or maybe, is 50V the amount of swing from the hum, and not just a static measurement of plate at idle? Is that it?

I feel I am so confused.

If I wanted to increase the plate voltage to reach a more "vertical slope" portion of the curves (and thus, lower output impedance), how would I do it? Lowering the 150kohm would change the load line, and place the bias point "further up" into the more vertical portions?

Could I keep the 150kohm the same, but change the cathode to 2V? Would that raise the plate voltage at idle? It seems like it would, but maybe it would also increase output impedance do to the slope of the -2V grid.

Again, not sure I'm grasping this at all, unfortunately.

 

[Matador]

I feel I am so confused.

If I wanted to increase the plate voltage to reach a more "vertical slope" portion of the curves (and thus, lower output impedance), how would I do it? Lowering the 150kohm would change the load line, and place the bias point "further up" into the more vertical portions?

Could I keep the 150kohm the same, but change the cathode to 2V? Would that raise the plate voltage at idle? It seems like it would, but maybe it would also increase output impedance do to the slope of the -2V grid.

Again, not sure I'm grasping this at all, unfortunately.

You are close:  lowering the plate resistance reduces output impedance.  If you keep the B+ the same, this reduces the slope of the line and makes it "flatter".  Quiescent current will drop, so there is less delta change in current for a given change in grid voltage (e.g., gain is less).

Changing the cathode to grid voltage via a cathode resistance changes where along the curve the tube sits a "idle", or 0V grid input (wrt. ground).  If you want it to idle farther up the load line, you must increase the cathode resistance.  You would plot you load line, then start at 50V at the bottom of the tube curve chart (50V/0mA point), and go "up" until you hit the load line.  You then estimate the grid voltage at that point (VGQ), then follow that point directly left until you intersect the vertical axis and find the current at that point (IQ).  The cathode resistance is then just VGQ/IQ.

Your tube and the tube curve aren't the same:  the actual points might be different.  You could start with a 10K pot at the cathode, then dial it downward until you read 50V on the plate (remembering that 50V is the plate-to-cathode potential, not plate to ground potential).  Then read the cathode voltage and that will give you the grid bias that gives you the 50V plate-to-cathode level (could be -1.5V, maybe -2V?  Hard to say).

This explains it pretty well: http://www.angelfire.com/planet/funwithtransistors/Book_CHAP-4A.html

This pic is from a JFET, but it shows the same principles (load line, Q point, input on gate/grid, output on drain/plate):

Hopefully this helps!

 

[granger.frederic]

the most difficult point in fixed bias is to manage the background noise.
it's difficult to achieve a perfectly stiff and noiseless PSU.
if you bias the grid negatively, remember that you are in the ultra high impedance section....

with the actual analogic to digital convertors and preamp performances, the noise of a mic is essential.
we're not in the 60's !!!
with the actual "loudness war" and the over-compressed actual sounds, we need some very quiet gear...

self bias is the easiest way to achieved a perfectly quiet tube mic
Neumann has modified their M49 and U67 from fixed bias to self bias for that reason (M49C and M367).

i add that one of the most respected mic : the elam 251 is self biased...
the C12 is almost self biased also
the U47 is an exception, but stays quiet because of the large underheating

Remember that for a noiseless heater supply , you need a bigger transformer: 25VAC and 2A min (not Alctron 9,5v...!) ,
the needed voltage drop leads to components heat , which leads to noise , etc...
basic integrated regulators are not suited (LM317,78XX...too much noise), and the tube heater filament introduces a little more noise than a resistor.
Thus you need an extremely well designed schematic, but you will still have more noise ...
diode biasing introduces non-linearities and noise...
Only a 1,3v CadNi battery will be almost silent in fixed bias, but not very easy to use.

since 10 years of mic tests , i often finish my microphones  with a self bias ...
i always custom adjust the cathode resistor in self bias, i often like to bias the tube a bit hot , this reduces impair sub-harmonics
i don't put a random cathode resistor and a general purpose capacitor!
i  measure the cathode voltage after 1/2h warmup...
i use very high quality cathode bypass caps : elna silmic , black gate , panasonic FC or nichicon muse (6,3 v : important to minimize non linearities...)
the difference vs a fixed bias is not that big, and the mic is always quieter.
the only audible difference is with very loud and constant sources : the mic compresses a bit in self bias .
90% of the time, i hear no difference except for the background noise.

That's just my two cents' worth.

 

[micaddict]

Nice post(s), but could this be a typo?

fixed bias is the easiest way to achieved a perfectly quiet mic

 

[RuudNL]

One of the most important advantages of self bias is that ageing of the tube will not have a big influence on the operatinmg point.
Personally I have never measured a lower background noise in a self bias configuration vs. fixed bias; this may however depend upon the design. In some cases self bias may also introduce a form of feedback, thus lowering the gain.

 

[granger.frederic]

you can and must rebias the tube with its ageing !
Self bias ALWAYS introduce negative feedback...thus, to avoid that, audio voltages must be shunted by a cathode capacitor...
difference gain is negligible when bypassed by a large capacitor...
could you give me your design because i always had slightly noisier mics in fixed bias ...?
(personally, i use a scope and spectrum analyzer to measure...)

 

[Melodeath00]

I feel I am so confused.

If I wanted to increase the plate voltage to reach a more "vertical slope" portion of the curves (and thus, lower output impedance), how would I do it? Lowering the 150kohm would change the load line, and place the bias point "further up" into the more vertical portions?

Could I keep the 150kohm the same, but change the cathode to 2V? Would that raise the plate voltage at idle? It seems like it would, but maybe it would also increase output impedance do to the slope of the -2V grid.

Again, not sure I'm grasping this at all, unfortunately.

You are close:  lowering the plate resistance reduces output impedance.  If you keep the B+ the same, this reduces the slope of the line and makes it "flatter".  Quiescent current will drop, so there is less delta change in current for a given change in grid voltage (e.g., gain is less).

Changing the cathode to grid voltage via a cathode resistance changes where along the curve the tube sits a "idle", or 0V grid input (wrt. ground).  If you want it to idle farther up the load line, you must increase the cathode resistance.  You would plot you load line, then start at 50V at the bottom of the tube curve chart (50V/0mA point), and go "up" until you hit the load line.  You then estimate the grid voltage at that point (VGQ), then follow that point directly left until you intersect the vertical axis and find the current at that point (IQ).  The cathode resistance is then just VGQ/IQ.

Your tube and the tube curve aren't the same:  the actual points might be different.  You could start with a 10K pot at the cathode, then dial it downward until you read 50V on the plate (remembering that 50V is the plate-to-cathode potential, not plate to ground potential).  Then read the cathode voltage and that will give you the grid bias that gives you the 50V plate-to-cathode level (could be -1.5V, maybe -2V?  Hard to say).

This explains it pretty well: http://www.angelfire.com/planet/funwithtransistors/Book_CHAP-4A.html

This pic is from a JFET, but it shows the same principles (load line, Q point, input on gate/grid, output on drain/plate):

Hopefully this helps!

Hi Matador! Thanks! This indeed helps, but I have a few questions. You said "If you keep the B+ the same, this reduces the slope of the line and makes it "flatter".  Quiescent current will drop." Doesn't keeping B+ the same, but reducing the plate resistor increase the negativeness of the slope of the load line, not make it "flatter"? I thought quiescent current will go up, not drop. If I used 50k Rp instead of 150k, my mA at 0V would be 2.4mA instead of 0.8mA. The load line has gotten more negatively vertical, not flatter, and at -1V grid, we're now at a higher current than when we used 150k. Again, I may be misunderstanding how to use/read these graphs.

Your link and remaining paragraphs seem to be talking about self-bias tubes. I don't think a cathode resistor or VGQ/IQ is applicable to my situation here. My EF802 is in a circuit with a fixed cathode bias of 1.1V tapped off of the heater. I cannot change that cathode voltage without adjusting the heater, or changing the voltage divider/circuit. If I raise the heater, I can raise the cathode up to 1.4V. In theory, would that raise the plate voltage? My idea is to reach a point in the curves that has a high slope. The more vertical the tangent/slope of the tube curve at the idle point = less output impedance. It seems to me that the way to do that is keep cathode and B+ constant, and lowering Rp would change the load line, and put the idle point "higher" on the graph, at a more vertical portion of the -1V grid line. Am I not getting it?

If you follow the load line and decrease the grid/increase cathode, you wind up getting at less vertical slope points (the "knees") of the tube curves, which means higher output impedance. I think.

the most difficult point in fixed bias is to manage the background noise.
it's difficult to achieve a perfect stiff and noiseless PSU.
if you bias the grid negatively, remember that you are in the ultra high impedance section....

90% of the time, i hear no difference except for the background noise.

Hi granger.frederic! Forgive me for cutting most of your post out in the quote above - I didn't want to take up too much space.

My "Alternate Tube U47" uses a "fixed bias" on the cathode, which is tapped off of the heater voltage through a 100 ohm and 29 ohm voltage divider. With a 5V heater, this puts the cathode at about 1.1V (the same grid operating point as the VF14 U47). I did not design the PSU, so I don't know how noise is being handled on the filament wire, but I can tell you my "U47" is not noisy at all. It is just the slightest bit noisier than my C12, but it is so minimal. It has a TINY bit of extra 120Hz hum compared to the C12, but it sounds perfectly quiet unless you crank to volume in order to TRY to listen to the hum. Hopefully that makes sense.

you can and must rebias the tube with its ageing !
Self bias ALWAYS introduce negative feedback...thus, to avoid that, audio voltages must be shunted by a cathode capacitor...
difference gain is negligible when bypassed by a large capacitor...
could you give me your design because i always had noisier mics in fixed bias ...?
(personally, i use a scope and spectrum analyzer to measure...)

How exactly does one re-bias a tube that is aging? Perhaps I do not understand because so far I have just been trying to figure out fixed bias designs. Can you rebias a fixed bias?

 

[granger.frederic]

C12 design is one of the most noisier design of the famous mics ...
the biasing method introduces 60hz/50hz and subharmonics and the long cable (and connectors) often introduces bad noises in the high impedance section
AKG has changed the C12VR to self bias ....

MK47 is the noisiest mic i ever tested (sorry but it's the truth) even with a bunch of twenty 408A , i've never found a pair of non noisy and microphonic tubes (filament movement noises..."dings" that made me mad...)
and by definition : 2 tubes in parallel are noisier than 1
unusable with large dynamic compression...

the rebias of a fixed bias is needed after a while, in general, just change a resistor or adjust a pot in the PSU....

 

[Melodeath00]

C12 design is one of the most noisier design of the famous mics ...
the biasing method introduces 60hz/50hz and subharmonics and the long cable (and connectors) often introduces bad noises in the high impedance section
AKG has changed the C12VR to self bias ....

MK47 is the noisiest mic i ever tested (sorry but it's the truth) even with a bunch of twenty 408A , i've never found a pair of non noisy and microphonic tubes (filament movement noises..."dings" that made me mad...)
by definition : 2 tubes in parallel are noisier than 1

the rebias of a fixed bias is needed after a while, in general, just change a resistor or adjust a pot in the PSU....

I've never used an MK47, so I wouldn't know. My C12 is also not any noticeably noisier than my Manley Reference Cardioid. In fact, I think all my tube LDCs are quieter than my dynamic mics.

What resistor do you change to adjust a fixed bias? The plate resistor? What exactly is the purpose? Does the plate output impedance increase as the tube ages?

 

[granger.frederic]

the noise of passive dynamic mic comes from your preamp only (or an impedance mismatch) ...

use a quiet preamp and compare again..

the plate resistor is not for biasing
to adjust fixed bias , you must understand the mic design....it depends...