Pentodes; plate resistance; damping...

[Johan Potgieter]

Good post as usual, PRR.

My bit on flat impedance (well, taking a lot of poetic licence on "flat"): One can get a reasonably useful impedance with most driver combinations by using the right compensating (conjugating) networks. One can also get a real horror show using ordinary off-the-shelf networks that appear to have been designed round resistive values (the 2xC - 2xL types for a single cross-over). It must be remembered that at a cross-over frequency of say 2.5 KHz, 8 ohm drivers are hardly 8 ohm, apart from having a healthy phase angle. Get this wrong, and one is in for a 10 dB peak at 2 KHz with a 15 dB dip at 2.6 KHz (actual example). I have never been able to get an acceptable impedance performance with fewer than 5 components for a single cross-over. (Acceptable like amplitude between 4 and 8 ohm; phase angle not exceeding +/- 25 degrees.)

The point I wish to make is that the impedance/frequency characteristic of a number of commercial loudspeakers can be improved by using the right reactive/resistive (additional) components. Purists are likely to go Texas on me, but it is often a case where the corrected loudspeaker giving a reasonable impedance to the amplifier to work with, comes out still better than a grand low-component loudspeaker but presenting such a lousy load to the poor amplifier that the end result is worse. I am no more a fan of cross-overs than any other person, but where the options are limited, I prefer some impedance correction to some of the "difficult-to-drive" models on the market.

[planet10]

Originally posted by mr mojo
I also gotta admit, as a graphic artist by trade I've got to read it over a few times before some of it starts to sink in.

Here is an impedance magnitude graph (RSC 8" FR)

dave

[PRR]

> a speaker whose impedance varies with changing loads.

Wait a minute. What "changing loads"? The air-load is essentially constant. You can shift it a little in the bass with corner placement, or box venting, or horn loading. But corner placement effects are small at the amplifier, and effective horn-loading is a completely different design.

SY> Speaker impedances ... vary with level

Yes, when the coil starts to come out of the magnet gap (huge excursions) or when the copper gets so hot its resistance rises (high average power). When the coil does start to come out of the gap, back-EMF and impedance drops; but not lower than the midband impedance. If the coil actually comes out of the gap and bounces back, huge oscillations are possible, but this is speaker abuse, more than amp-abuse. If a speaker is designed to work to the edge of the gap and beyond, as in competition car-sound woofers, the designer will balance a bunch of factors, including the ready availability of cheap (transistor) amplifiers. A tube amp may not be the wisest pick for throwing a woofer that far; but few tube amps have the power to really get crazy that way.

So I don't know what mojo's quote means.

The commonest impedance variation is with frequency. I took dave's plot of a very typical full-range speaker, assumed it was flat with a modern zero-Z amplifier, and plotted for several non-zero-Z sources for DF of 8 to 0.25.



You see that a speaker with perfectly flat voltage response but a bumpy impedance response will have "errors" with a non-zero source. (OTOH, you can design a speaker to be flat with a hi-Z source, and it will be bumpy with a zero-Z source. Most "good" speakers before the mid 1950s were tuned for significant source impedance, and sound blah on a zero-Z source. As an extreme, Nelson Pass is working/playing with high-Z drive and selecting and tuning the speaker to suit that condition.)

Since you have a VOLume knob and will trim any amp to a "reasonable" sound level, I have corrected for average output.

DF=Infinite is the modern standard, and I've assumed it is "flat".

DF=8 would be a transistor amp with long thin wires (typical BestBuys stuff) or a light-feedback tube amp. It gives +/-0.5Db errors.

DF=2 approximates a naked triode. Errors are OTOO +/-1.5dB. The bass-bump and top-rise may be pleasing.

DF=0.5 seems to be about right for UltraLinear without other NFB. DF=0.25 is getting into naked pentode territory. Both of these give an obviously boomy one-note bass for a speaker designed for large DF. However this and the top-rise may compliment the limits of a single-driver full-range system, by boosting where it starts to fall.

[rdf]

Does DF scale linearly with number of output devices? I.E., if single triode out DF=2, two triodes DF = 4, times 3 DF =6, etc.. The previous assumes all triodes share the bias and standing current as the the single triode case, what if current is reduced as tubes are added? I just ordered a metric boatload of Russian EL84 equivs to play with and one consideration is triple or higer PSE.

[SY]

Originally posted by rdf
Does DF scale linearly with number of output devices?

Yes and no. Normally, when you parallel output devices, you reduce the plate to plate load proportionately, so DF becomes a wash.

FWIW, naked pentodes can have a somewhat higher source impedance than even the gloomy estimates from PRR- removing the feedback loop from a pentode-connected ST70 (EL34s, 400V on screens and as B+ feed for the output transformer), I saw source impedances over 100 ohms.

[rdf]

Thx Sy. In this case the prime consideration is to increase the anode load and try for higher linearity, increasing power out is secondary. The transformer's a 5 kohm primary capable of sinking about three times a normal EL84 bias. I want to play around with the load/power out balance as a learning experience. The tubes are cheap like borscht for experimentation, if something good comes out of it the circuit can be adapted to something better.

In this particular case then am I correct in thinking that driving a 5000 ohm primary with multiple tubes results in a DF increase over a single similar tube of about equivalent DC operating condition?

[SY]

Yes, if you keep the load constant but double the number of tubes (along with doubling idle current), you will halve the source impedance.

5K is a nice match for p-p EL84 in triode. Should be good for 10 watts or so.

[Johan Potgieter]

PRR,

Yet again relief from me that someone has pointed out that DFs above 10 don't amount to much. When I suggested under another thread that the popular definition of damping factor is electrically in error I got blasted for daring to do so; I will not resurrect that.

A small comment regarding your guestimates for some DFs (I accept that it was ball-park). But triodes would be nearer the 4 mark - I find that the Ra for the KT66 is only 1.2K as triode under normal conditions. Then, the UL use of a pentode for the few examples given shows an Ra about 80% of that for the equivalent triode connection. (Folks often seem to think that UL operation is close to pentode; it is actually closer to triode except for output.) For the KT66 this is about 2.5K compared to the pentode Ra of 22K, so the "naked" DF there for UL would be nearer 1.5.

But please, not to be pedantic or diminish your contribution. Just thought I would mention. UL data is actually hard to come by.

[PRR]

> Does DF scale linearly with number of output devices?

If you (perversely) kept the same transformer ratio: it would, except that transformer resistance sets a lower limit on DF even for infinite tubes.

But for any commercial design (not a borscht-cheap experimenter's toy) you would scale the load in rough proportion (so increased power justifies the increased price), so (as SY said) it washes out.

> driving a 5000 ohm primary with multiple tubes

Yes, though (again thinking rational commercial design) the increased idle current will demand a bigger hunk of output iron. That can actually reduce DCR, though the reduction with size is slow, and the increase in cost can be large, so the DCR may not fall much across the range of Standard Transformers.

Still, if bottles and heater-power is cheap, parallel away.

Looking for low transformer resistance to improve DF is one place where choke-feed makes sense. Getting the DC current out of the transformer allows a much lower DCR design. The choke's DCR will be high, but is not in series with the load, so does not degrade DF, just wastes DC power. Of course choke-feed generally implies a capacitor, which gives a very complicated output impedance. If you make the cap too small it interacts with speaker impedance; if too big it tends to give a big subsonic rise and resonance.

> if you keep the load constant but double the number of tubes (along with doubling idle current)

That scales linearly. If OTOH you add tubes but keep the total current the same (less current per tube), you still get some improvement, but much less. Roughly square-root of number of tubes. 800Ω Rp into 5K load is DF=6 (ignoring transformer resistance). Four such tubes at the same total current will be near DF= 12, twice as high (not always twice as good). The square-root factor is not exact, but gives a general idea where you are headed. The very-linear tubes like 300B may not show even this much improvement.

[rdf]

Thanks to both for the explanations. The iron is really (at this point) intended for a triode connected 813 so it'll take a lot of abuse from EL84s. It's an opportunity to play with some contentious concepts like PSE and driving like-with-like. Most EL84's I've tried are too microphonic or have a curious glassy-crackling noise to be seriously considered for a front end tube.