Question on my biasing technique...

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Mbira

Well-known member
Joined
Jun 4, 2004
Messages
2,422
Location
Austin, TX
I have been doing lots of guitar amp work these days. I want to see what the general consensus is for biasing power tubes. What I was taught was to feed a 1K signal into the 1/4 inch input, and then hook up the scope to the speaker jack (while a speaker is connected) The black goes to chassis ground, and the red goes to the tip of the output jack. I am then looking at the slope of both tubes and seeing if I get an early cutoff (in the middle of the wave). If so, I adjust the bias until there is the most even looking slope.

I have a couple questions about this technique. First, is it a good one? I have seen other techniques with soldering on resistors, etc, but I want something that is quick and cost effective for the customer. The problem I have seen with this method is that sometimes it seems like the trace is real crisp and the cutoff is very pronounced. Other times, it is hard to see the cutoff, but adjusting the bias just makes the whole slope change angles a little (did that make sense?). Also, sometimes, changing the bias doesn't seem to get the the cutoff smooth all the way-in this case, does that mean I need to change the bias resistors? Other times, one side of the wave is clipping while the other isn't-even when the bias is equal on both tubes.

I think that is all the questions for now...

Thanks.
Joel
 
I scope biased an amp once and did not like the sound. Biased it by ear, put it back on the scope, and saw some zero crossing garbage, so if that does not add to the confusion....

When you use a steady sine wave, dynamics go out the window.

I am even finding that transformers react differently to a changing frequency, not just a static frequency. AC current is drawn when you go from one frequency to another. So I guess I have to figure out a way to do a rapid sweep test on these buggers to see what I am missing.

Bias by ear, then check the ma draw, if it's in a somewhat reasonable window, then ship it.
 
alo
i bias all my amps with a biasking,and at 70% plate dissipation,and the amps sound good(at least to me),and the tubes have a long life;and i play everyday,and loud.
best regards
pedro
 
I have been told some stories.

FWIW I am just passing on a story. I would NOT do the following

One is do it by ear and when it sounds good and the plates are not glowing red, good to go. Sounds a bit crazy but the person that told me this had seen this done and said the amps sounded good.

Remember your eyes/brain can't see at the resolution your ears/brain can hear
 
The nastiness of an amp that's biased too cold manifests itself most audibly at low output levels. I don't work on many guitar amps these days, but I used to fix them every day, and here's the method I used. This applies to class-AB, push-pull, fixed bias amps, of course.

After a warmup period, I'd adjust the bias so that the tubes were idling at 50-60% of their rated max plate dissipation. Then, with a dummy load connected, I'd inject a sine and adjust the level to give 1 watt output. Using a distortion analyzer tuned to notch out the fundamental, I'd overlay the distortion product waveform over the output waveform on my dual-trace scope. If the amp was biased too cold, there'd be little spikes in the distortion product at the zero-crossing points; this occurred even when the raw output waveform looked just fine. (And this helps to illustrate why the old "crossover notch" method usually results in biasing too cold). I'd set the bias to smooth out the spikes, but no hotter than that. Then I'd re-check my plate dissipation and if it was still at or below 70% of rating, I'd call it good.

The result was a good-sounding amp that didn't run the tubes hotter than they needed to be--which meant a satisfied customer, long tube life, and no "callbacks."

Oh, I almost forgot to mention: I used 1 Ohm precision resistors in series with the cathodes to measure the current. The current measured is actually the combined plate and screen current, but consider it as simply the plate current in your calculations and you'll have a little extra safety margin.

Another note: years ago, I had an idea for a biasing tool that would sample plate voltage and cathode current and, using an analog multiplier, display directly in watts. I never had time to pursue the idea beyond the general concept. A well-known manufacturer of guitar amp products tried to develop my idea for use in one of their products; but the last I heard, they were having problems with drift in the multiplier and weren't ready to bring it to market. If they ever succeed, I hope they'll have the decency to at least send me some free stuff :wink:
 
[quote author="NewYorkDave"]Another note: years ago, I had an idea for a biasing tool that would sample plate voltage and cathode current and, using an analog multiplier, display directly in watts. I never had time to pursue the idea beyond the general concept. A well-known manufacturer of guitar amp products tried to develop my idea for use in one of their products; but the last I heard, they were having problems with drift in the multiplier and weren't ready to bring it to market. If they ever succeed, I hope they'll have the decency to at least send me some free stuff :wink:[/quote]

The world needs a good and reasonably inexpensive four-quadrant analog multiplier. I'm working on something even as we speak that requires several.

The monolithic parts are expensive, if good, and usually single-sourced. The 13700/NE5517 parts can be made into two- and four-quadrant multipliers but they are noisy and slow. If bandwidth is not a great concern one can craft pulse-height/pulse-width contraptions with decent accuracy and drift. With good transistor arrays like THAT Corp.'s you can make some good fast variable transconductance systems or slower log-antilog ones. But all of the above are a royal pain to one extent or another.
 
This application only requires a single quadrant since the voltages never go negative.

I cam across this Usenet post by Winfield Hill that offers a low-cost idea.

One interesting low-cost single-quadrant multiplication
technique is to take advantage of the accurate logarithmic nature of
bipolar transistors. Simply stated, A*B = anti-log (ln A + ln B).

I remember a fellow student at Harvard who became very excited when
he discovered RCA's newly-introduced CA3019 hex diode IC (really six
transistors wired as diodes). He envisioned a form of complex analog
computing made with large numbers of inexpensive matched junctions...

Neither the CA3019, nor the CA3039 which arrived soon after, are
still with us, but five-transistor IC arrays such as the Intersil
CA3046 are easily obtained. One of these plus a quad opamp and a
handful of resistors can complete the task for under $1. :)

From the same author, in another thread:

A multiplier is not a bad idea, but most, like the AD835
from Analog Devices, require split supplies. Instead let me suggest
an approach that's well-suited for single-supplies: log circuits.
Consider, anti-log [log I + log V] = I * V. We'll use Intersil's
CA3096 matched-transistor array IC to implement a single-quadrant
multiplier. The classic Ebers-Moll transistor equations (see AoE
page 80 and) tell us how to execute our design. The equation for
base-emitter voltage is Vbe = kT/q ln Ic/Is, where k is the Boltzman
constant, T is absolute temperature, q is the charge of one electron,
and Is is a very small scaling current called the saturation current.
ln means the natural log. Because the term kT/q will appear often,
we'll simplify it to a voltage Vt = kT/q = 25mV at room temperature.

[ASCII schematic mangled by PHBB, so I eliminated it]

We're using PNP transistors Q1 Q2 to add the log of two currents (one
related to current, the other to voltage), and a third NPN transistor
to offset the result. Using the equation above we can write

Vo = Vt ln I1/Is + Vt ln I2/Is - Vt ln I3/Is

= Vt ln (I1 * I2)/I3 M,

where M is a term wrapping up the three Is values. The Ebers-Moll
equation can also be written to express collector current resulting
from a base bias voltage, this is an antilog or exponential function.

Ic = Is e^ Vbe/Vt

We'll take the voltage Vo we obtained above and use it to drive the
fourth anti-log transistor, so the resulting current will be,

I1 I2 Is3 Is4
Io = ----- -------
I3 Is1 Is2

Ideally the last term can be made to cancel out, giving us I1 * I2
plus a scaling factor from I3. Pretty cool, huh? One low-cost IC
plus a few resistors, and voila, fast real-time power calculations.

An application from National using their LM194 super-matched BJT pair:
PDF
 
RCA introduced the CA3045/3046/3086 years ago---the ads trumpeted "The Economy Five." Sadly the only pretty good version of the CA3046 (high beta, low popcorn noise, the latter something that plagued the other parts) was made for a while by STMicro, and even it didn't have particularly low bulk resistance (which spoils the log conformance). That is THAT's claim to fame.

I used 3046's for a zillion things over the years and it was sad to see one vendor after another discontinue them. I believe they are all discontinued now---for a poetic farewell, see

http://www.atypical.net/archive/2005/02/12/sic-transit-gloria-ca3046

I am sure there is some NOS here and there though.

Thing is, if you really need four quadrant you want at least a matched quad for transconductance multipliers.

There is the quarter-square technique, which extracts the product from the inrterstices of an algebraic identity: xy = 1/4 [( x+y)^2 - (x-y)^2], and which needs two squaring circuits. If you can know that y < x and x > 0 then the squarers can be one-quadrant. If not then absolute values can be used and the sign restored afterwards.

There were big rackmount multiplier/dividers made this way with tubes, including a specially selected tube for realizing the squarers :shock:

The log/antilog stuff is usually temperature sensitive and the usual solution is to use some special resistor somewhere to stabilize against ambient changes.
 
> I adjust the bias until there is the most even looking slope.

That works fine for big tubes at low voltage. So does looking at distortion residual, though 1 Watt is not always the best test level (you will get a different "optimum" at 0.2W or at 5W, and in some amps this matters a lot).

Small tubes at high voltage, or even big tubes at Fender's voltages, will usually over-dissipate before you get the kink out.

Red-glow is usually bad, but that's not a good test. Some heavy-duty tubes are rated to work in the red-glow mode. Most receiving tubes (even 6L6) are rated for a point far below glow. They do make them glow in the factory, to drive gas out of the metal. But it takes a long time to get all the gas out. What the factory does not get out, will come out when you cook the tube. A little bit of gas in a vacuum tube makes it a gas-tube, which will cause BIG problems, including smoking transformers. So you need to know the actual plate dissipation, and check the ratings. For some tubes, experience shows that the ratings can be pushed, but others won't really meet their ratings reliably.

BTW: many tube sheets say you can not hit two max ratings at once. Original 6L6 can take 360V, or 19V, but not 360V and 19V at the same time. Since many guitar amps push the B+ to or beyond rating, you really should not push the dissipation to the max.

So a #1 rule should be: respect the plate rating. You don't really need NYDave's real-time calculator, or bcarso's dream of a good 5-cent multiplier: you know the B+, divide by plate rating, don't exceed that current. B+ will vary somewhat, and NYDave's calculator will track that. But over the useful range of idle currents, B+ won't change much. Use the low-current B+ value and compute the max current. B+ sag will add a little safety factor.

A #2 rule is: bias as hot as possible without breaking the #1 rule. In theory there is an ideal bias current (for a specific goal; lowest THD is not the same as least-annoying distortion). In practice, under-current is bad and over-current is good.

In theory, using a good simple approximation of tube linearity: each tube's idle current should be about 1/4 of peak current. This is probably not useful on the testbench (though you can find peak current with a load, a scope, and that 1Ω resistor you should routinely put in the cathode). For a big 6L6GC amp, 590V idle B+, 550V B+ at full blast, 5,600Ω p-p load, assuming to get to full power it goes Class AB (it will): take plate-drop as 50V, one-tube load is 1,400Ω, 500V/1,400Ω= 357mA peak, the theory suggests 357mA/4= 89mA per tube idle. The per tube idle dissipation is 590V*89mA= 52.7 Watts, far in excess of the 35W rating!

You can't get the kink out without frying the tube, and if the tube was not very well de-gassed (few are these days) you can burn a transformer too. So in this case, the wisest path is to go as much current as possible without violating our #1 rule (don't fry the tubes). Since we alread meet or beat the rated plate Voltage, we should not get too close to rated plate Power. 70% is a nice round number, or 66% or 75%, not fussy. 70% of the rated 35W is 24.5W, but round-up to "25W". Round-up the B+ to "600V". Now we can do the math on our toes (tube calculations rarely need even toy slide-rule accuracy). 25W/600V= 1/24 Amps which is about 40mA. That is your maximum per-tube idle current for this amp. You can go lower if it sounds better. A hi-fi amp won't. Some guitar flavors may work with lower idle current, though generally the more the better.

All that is for fix-bias push-pull (and assuming it can't reach full power while staying Class A; I doubt any hand-caried hi-power amp does.)

For cathode resistor bias, check that idle dissipation is safe, and that cathode current (voltage) rises about 10%-40% from idle to full power sinewave with rated load.

Properly designed transformer coupled SE amps are dead-easy. Short-out the bias and read the cathode current before the tube melts. Idle at half that current. Check against plate power rating and reduce idle current if needed. Let it cook and check transformers for excess heat. Economic SE design is all about pushing things to the almost-melting point, even accepting short life in return for low tube-count and cost, so it is gonna run hot. And sound will generally improve with heat, though slowly. However the SE amp does not have the gross low-level kink that an under-biased p-p AB amp has, so you can work over a wide range of idle current with big change of heat, some change of power, and small change of sound.
 
> An application from National using their LM194 super-matched BJT pair: AN-222

That's the base LM194 essay.

AN-265 has a multiplier especially adapted to power computation. It has some clever details. It is "too good" in the sense that it can handle AC (including reactance) and we only need DC. And of course it adds some circuit to get the signal out without electrocution. The voltage scale would have to be adjusted for typical tube B+ voltage, a trivial mod.

http://www.national.com/an/AN/AN-265.pdf
 

Latest posts

Back
Top