Author Topic: THD of output tubes in different configurations  (Read 610 times)

Offline Mars

  • Commercial Member
  • Trade Count: (+1)
  • *****
  • Posts: 1,186
  • Total likes: 52
  • “The path is the goal.” Mahatma Ghandi
    • Mars Tube Audio
THD of output tubes in different configurations
« on: November 08, 2019, 11:28:44 AM »
I am working my way through all the discussions around fixed bias versus cathode bias. One aspect that seems clear (or is it?), is that fixed bias gives lower distortion in power amplifiers mainly due to the fact that bias do not change as the voltages in the output tube climb and descend. Am I correct? Ampdog? Handsome? fredeb? Help me if I need to be corrected :comfort:

But when I look at published data of regular output tubes, the lowest THD figures are for cathode bias pairs, usually Ultra Linear connection. I took some screenshots from the TDSL page to illustrate. See below:

6550


KT88


6L6GC


EL34


So the question is: does cathode bias really give lower distortion figures than fixed bias in amplifiers?
OEM manufacture of Tube Audio
Power and Integrated Amplifiers
Phono Pre-amplifiers
DIY Kits
DIY Workshops

Online fredeb

Re: THD of output tubes in different configurations
« Reply #1 on: November 08, 2019, 03:32:01 PM »
I really don't know , but it seems evident that cathode bias gives lower distortion from the tables you posted . I know that Coen tried both ways with his RCA KT88 PP UL amps , and he preferred the sound of cathode bias .

Then again , one can possibly optimize a circuit for either condition differently . For example a higher value grid resistor with fixed bias could possibly reduce distortion . Adjusting NFB could also make a substantial  difference .   
...evolution is the gradual development and stratification of progressive series of wholes, stretching from the inorganic beginnings to the highest level of spiritual creation.
Jan Smuts

Offline handsome

Re: THD of output tubes in different configurations
« Reply #2 on: November 08, 2019, 06:33:29 PM »
You need to compare THD at the same operating point and into the same load. Converting a cathode bias amp to fixed bias will change the operating point unless you modify the power supply as well - otherwise the fixed bias amp gains the bias voltage as HT. Converting a fixed bias amp to cathode biased, then the amp must have individual cathode resistors that are fully bypassed. There should be no difference in THD because how the bias is derived is irrelevant, its the operating point that counts and that’s what the bias defines. Clipping behavior on the other hand, will be different between fixed and the various cathode bias options. Of course fixed bias amps can be adjusted so their performance can be “tuned” in the face of tube aging, differences between transformer halves and differences between tubes in a pair while cathode biased amps rely on their ironically, fixed resistor values.

Online fredeb

Re: THD of output tubes in different configurations
« Reply #3 on: November 08, 2019, 08:19:23 PM »
There should be no difference in THD because how the bias is derived is irrelevant, its the operating point that counts and that’s what the bias defines. Clipping behavior on the other hand, will be different between fixed and the various cathode bias options.

Thanks handsome . Do you mind extrapolating regarding clipping behaviour with various methods ? Eg Which one will clip earlier ?

It appears lower anode voltage and lower power also offer lower THD , perhaps parallel push-pull with low anode voltage is the way to go for power and low distortion .
...evolution is the gradual development and stratification of progressive series of wholes, stretching from the inorganic beginnings to the highest level of spiritual creation.
Jan Smuts

Offline handsome

Re: THD of output tubes in different configurations
« Reply #4 on: November 08, 2019, 09:21:59 PM »
Clipping is sharper with fixed bias, softer with cathode bias. Meaning on a graph of output power vs distortion the fixed bias amp the distortion rises suddenly at the clipping point (usually 1% on transistor amps) while a cathode biased amp goes gently into gross distortion and the actual point of clipping is hard to define. The last two paragraphs and the picture on this page http://www.gmarts.org/index.php?go=216 describe and show it it better than i can.

Distortion for a (particular, they all have different characteristics) valve is a function of load and operating point. There is no hard n fast rule like low voltage. It depends on how much power you want and the load you are using then you choose your operating point. Distortion increases with power so when you talk about distortion you need to specify at what power level. An amplifier with 10% distortion at 10watts may have less than 1% thd at 1 watt....

Read this from Valve wizards website on designing a triode gain stage - section 1.11 discusses distortion and its the same for power amps and pentodes http://www.valvewizard.co.uk/Common_Gain_Stage.pdf

Online fredeb

Re: THD of output tubes in different configurations
« Reply #5 on: November 08, 2019, 09:26:30 PM »
Cool handsome - now I've got some reading to do . :)

Sent from my SM-G935F using Tapatalk

...evolution is the gradual development and stratification of progressive series of wholes, stretching from the inorganic beginnings to the highest level of spiritual creation.
Jan Smuts

Offline Ampdog

Re: THD of output tubes in different configurations
« Reply #6 on: November 10, 2019, 12:24:27 AM »
O Jene - - I wondered when this aspect will rear its multi-faceted head  :EGrin:  Oh well then . . . .

You need to compare THD at the same operating point and into the same load. Converting a cathode bias amp to fixed bias will change the operating point unless you modify the power supply as well - otherwise the fixed bias amp gains the bias voltage as HT. Converting a fixed bias amp to cathode biased, then the amp must have individual cathode resistors that are fully bypassed.

Quite - where power valves may require bias voltages of some 40V - 60V, a meaningful difference could be made. This makes comparisons from valve datasheets difficult; popular datasheets/graphs are seldom set out to cater for just such otherwise 'same-condition' circumstances.

Quote
There should be no difference in THD because how the bias is derived is irrelevant, its the operating point that counts and that’s what the bias defines.

Er - yes and no - this is exactly where the problem lies!!

Point does not lie in the magnitude of the bias voltage, but the fact that with cathode bias said voltage will not remain constant. In practice the textbook definition of class A always drawing a constant cathode current irrespective of bias voltage, just is not so; as the output increases the mean cathode current always increases, [e.g.: For p.p. 6L6's in class A under certain conditions, zero signal cathode current may be 145mA, while rising to 172 mA at maximum output. (RCA Tube Manual)] With cathode bias this magnitude will certainly shift the operating point, thus generating less than optimal conditions. In fact, it follows intuitively that distortion will always be higher with cathode bias than with fixed bias.

This is borne out in valve data sheets and particularly graphs of distortion vs. output power/cathode current.  Similarly will the maximum available output power always be lower for cathode bias than for fixed bias, other matters remaining the same.

[Trying to google this I found an unexpected variety of opinions not always supported by reality.  Despite what some claim, RDH-IV (Radio Designer's Handbook) is silent on this. But a good read on this can be found on the 'Audiokarma' site, under the thread heading "Is there any real advantage to Cathode Bias?" by Dr. David Gillespie (round about April 2012; fear of violating copyright rules does not permit me to copy directly.)]

As said directly comparable data is difficult to find from data lists and graphs; I will try to do so for conditions as close as possible to reality. But to repeat: The increase in total cathode current from zero to maximum output basically makes it impossible for cathode bias to yield the same results than for fixed bias.

One can try all sorts of twists and turns to try side-step this. E.g. make the maximum cathode current work-point the same as for the optimum fixed bias work-point. But then one will find that invariably there will be too little bias at low output, causing  anode and screen dissipation to go beyond specs.   Etc. etc.

Side Note: I might as well while I am at it: An interesting argument resulted some decades ago re what was called 'music power' operating conditions.  AND NO! I AM NOT TALKING ABOUT THE SO-CALLED P.M.P.O.!! (Grant me at least some wherewithal  :) )

The logic is that with sufficient bypass capacity - cathode as well as h.t. - a semblance of fixed bias operation may be created for most music because music's mark-space ratio is commonly small (low average energy; peaks not too frequent) and large enough capacitors will hold the operating points comparatively constant. This is a procedure that I also support, particularly what with comparatively physically small modern electrolytic capacitors. Folks will often find that I use up to 6,8K.µF cathode bypass caps and into many 100s of µF power supply filter caps.  (Again realists may argue whether the cost will not be the same as for 'extra' fixed bias; even if so, one still has the advantage of slow stabilising of working points.)

[Perhaps someone can recall: Was this 'music-power' matter raised in the original Mullard booklet by W. Ferguson - the one introducing the famous '5-10' and '5-20' amplifiers?]

The above was the easy part. How the tables carefully sought out by Karel came to be, plus a load of other conflicting arguments which the diligent researcher will encounter . . .  Perhaps Karel can enlighten us to the source of his information (as always, he would have had good reason for quoting the same) . . .

As promised I can at least try and compile a similar list from valve data sheets I have (may take some time; somewhat busy at present). Also, sadly, I have not yet found relevant explanations by Rod Elliott of ESP whom I hold in high esteem. Second to that, may I again suggest reading the given post by Dr Gillespie - an astute analyser if ever there was one. (Also I wonder if Morgan Jones had anything to say about all this. I fear I sadly am not a proud possessor of any of his books.)

Apology for length. As members might gather, long ago this was a 'sore point' with protagonists of one school or the other. Still seems to be . . . .

and finally: Listening tests again: The discussed effects will be audibly mild; one needs a certain type of music to make matters immediately audible when encountered. But as the question was asked . . .
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)

Offline handsome

Re: THD of output tubes in different configurations
« Reply #7 on: November 10, 2019, 07:56:04 AM »
You are of course right as ever though and  I do stand corrected: ultimately fixed bias will give better distortion performance. However in absolute terms it shouldn't: as I said considering just valve characteristics, load and operating point - in other words the datasheet - type of bias is irrelevant. And no doubt at lower volumes where intrinsic distortion is low and we are free and far from clipping (which is where 90% of our listening should occur) distortion should be the same.

Regarding the shifting of bias point but the opposite to what you mention can also occur with a bypassed cathode resistor: as the signal get larger and intrinsic valve distortion begins to rise,  the cathode capacitor will begin to integrate this "offset" voltage and thus begin biasing the stage colder lessening the mean current. This of course, is the onset of blocking distortion which is both heinous and specific to cathode bias.

Considering the data that Karel presented this would be gathered from manufacturers' datasheets and therefore would have been measured with single tone sine waves - these are not asymmetric and uncorrelated - unlike a music signal - and would perfectly sum to zero on every cycle thus the bias variations we both mention would perhaps not be as prominent under these (totally unrealistic) conditions?

Offline Mars

  • Commercial Member
  • Trade Count: (+1)
  • *****
  • Posts: 1,186
  • Total likes: 52
  • “The path is the goal.” Mahatma Ghandi
    • Mars Tube Audio
Re: THD of output tubes in different configurations
« Reply #8 on: November 11, 2019, 01:15:45 PM »
Thank you all for you contributions and comments, especially Ampdog's link to the Gillespie answer to the question.

As Charles has mentioned before (somewhere on the forum) if you want to know, measure it.

I agree that the issue is complicated, and there are a few aspects to this discussion that can veer off into much words and debate.

To sum up:

  • Fixed bias beats cathode bias where distortion is concerned
  • Fixed bias can deliver more power for the same cost as cathode bias
  • Cathode bias is easier to implement
  • Cathode bias is more reliable
  • Cathode bias is cheaper to implement than fixed bias. (this could be challenged :whistler:)
  • Fixed bias needs to be adjusted to compensate for tube characteristics changing over time.
  • There are no accounting for taste. Some listeners/constructors prefer the cathode bias, other fixed bias.
  • Cathode bias allows tubes to be run at 100% dissipation, with fixed bias exceeding 70% would be dangerous, unless fuses are placed on the cathodes of each output tube. Guitar amplifiers (basically a heavy duty cost effective tube amplifier) are biased at 50% dissipation (mostly)
  • From my experience: If you are using a power transformer that is designed to just deliver the necessary current/voltage specified by the circuit design, cathode bias sounds best. If you can increase the VA (double/triple) above that what is needed, then fixed bias will make you happier.



Quoting Gillespie from the discussion Ampdog mentioned above:

"Regarding distortion, the only time cathode bias can produce the same low distortion levels as fixed bias can, is with a true class A push-pull amplifier. In such an amplifier, the voltage across a cathode bias resistor remains as stable as if fixed bias were used, because the current draw through the two output tubes is exactly equal and opposite at all times through out the entire power output range of the amplifier. Therefore, in this scenario, distortion will be equally low which ever bias scheme is used."

and...

"Ultimately then, power output is always reduced with cathode bias, but distortion can still be made to be quite reasonable at the further expense of increased quiescent current draw. However, as long as enough usable low distortion power can be safely generated with cathode bias, then the practical sides of cathode biasing come into play; i.e. the auto quiescent current adjustment feature, economy, etc. Since the 6BQ5 family of tubes can safely develop a very respectable 13-14 watts RMS power output with cathode bias, and because so many wonderfully efficient speakers exist that work so well with this power level, it's no wonder then that with all things considered, the cathode biased output stage (particularly using 6BQ5 type tubes) was by far the most popular design of all."


« Last Edit: November 11, 2019, 01:20:43 PM by Mars »
OEM manufacture of Tube Audio
Power and Integrated Amplifiers
Phono Pre-amplifiers
DIY Kits
DIY Workshops

Offline Ampdog

Re: THD of output tubes in different configurations
« Reply #9 on: November 13, 2019, 02:48:13 AM »
A quite good summary, Karel!  (I am not going to comment on minutiae.)

[Taste . . . who knows  :-\  I have had enough of 'taste' for one lifetime. Everyone has a right to taste - that does not automatically grant a right to peddle the same as fact . . .  but sorry; escaped before I could suppress the thought. Not saying you are guilty. Old age brings intolerance  >:( ]

I promised in an earlier post that I will hunt for data with enough equivalency to illustrate lower distortion with fixed bias. I found the task easier than I thought: There is none.  One can but come to the general conclusion as expressed above based on stacks of data. I found the most extensive data for valves EL34 and EL84 (some 20 pages each), but not enough correspondence of other factors to lead to an acceptable conclusion (e.g. all EL84 data based on cathpde bias, while some 6L6 data came close to illustrate the tendency but not exactly so).

Not sure I can add much else. {I have alsready mentioned that my own high power stereo amplifier uses 4 x 6L6GC in fixed bias p.p.p-. for 110W output, but the final stage is different (part cathode loading as in Quad II), so no valid comparison.  Perhaps of mild interest to some: The fixed bias feature is in the form of a 'constant voltage sink' in the cathode circuit resembling cathode bias topology (perhaps to call it an 'amplified' power zener diode). But that was done for control purposes. And yes: Against bias or any other failure one adds some 'protection' circuitry (fast auto turn-off); neither pleasant nor economical to lose eight 6L6s working with an h.t. supply of 600V  :-[  - but I get carried away.]  Relevance perhaps: In a serious fixed bias final stage amplifier the total cost justifies some basic protection circuitry; comparatively simple to design with modern logic i.c.s.]

From me then, pending further questions; Good subject Karel and worthwhile research by you.
« Last Edit: November 13, 2019, 02:51:58 AM by Ampdog »
Audio must be the only branch of engineering where lack of basics' knowledge is considered a superior form of wisdom. (Anon)