12ax7 loading-tone or consequences

[Chuck H]

A design I'm working on right now is using a fairly low load for a preamp stage. There is a 64k load circuit to the following triode and a 100k plate resistor. I don't "know" but I assume the plate resistor and the load it's driving are in parallel as the plate load.?. So,.. 39k? Substantially lower than typical. And perhaps only half the plate resistance. I know this sort of thing has been done before, but I'm looking for any experienced perspective on the matter. The previous circuit had roughly a 68k total load and I was pretty happy with the tone. I think I may have lost "something" with the new circuit. I'll be doing some experimenting with an increased plate load for myself for ear testing purposes, but I would like any practical insight from others here because the load I'm using now is convenient for a sort of "trick" circuit I'm working on. That is, simply increasing, say, voltage divider resistances would be impossible if I still want to achieve the "trick" Like I said, I'll be testing and listening for myself, but I have a hunch someone here already has some practical experience here.???

TIA

[Zozobra]

Have a looksy about at some of the recent Friedman designs. He uses large plate loads on the first few stages (split as 220k + 100k//500p usually!) but the interstage dividers loading is smaller than usual to temper the gain. This sort of thing is pretty common in a lot of the 80's modded marshall tye designs too (Jose etc...).

[Dave H]

How is it biassed? If it's 100k plate to B+ with a 1k5 cathode resistor (as usual) then it's going to clip at the top first when driving a 64k load. If you want more symmetrical clipping you'd have to change the cathode resistor to something like 820 or 1k.

For the same bass response the output coupling cap has to be larger for a 64k load. The output resistance of the typical gain stage is 38k (100k RL // 62k Ra) and it's driving 64k so the resistance to use to calculate the coupling cap value is 38k + 64k = 102k. e.g. a 22n cap is -3dB at 71Hz

[pdf64]

I know it should be the same but I perceive a difference when a regular 12AX7 CC stage has a heavy ac load; such that a 100k plate resistor feeding a 47k grid leak sounds worse than a 47k plate resistor feeding a 100k grid leak, no matter if the cathode resistor was tweaked to centre up the bias in the latter case or not. My test was via a 0.1uF coupling cap but the corner frequency there should be below significance in either case.
Try it out yourself and see what you find.
It may be a DCCF would be a good option for your application.

[Mike Sulzer]

My experience is what you want is an abundance of low order harmonics in the signal as the signal level increases. This can be done with tube or ss. When you put a heavy ac load on the circuit, you get saturation and cutoff at lower output voltage. One way to say this is that you are missing some of the desirable non-linearity resulting from rp varying with output voltage. Thus you go into clipping without enough low order harmonics. Perhaps you could actually move in the other direction by keeping the ac load high, and replacing the plate resistor with an active current source.

[Chuck H]

Thank you gentlemen

Zozobra and Mike, this is an early stage that never clips. That doesn't mean it's not exhibiting any distortions, just not blatant clipping. I do notice that this amp (when clipping later stages and the power tubes) has never exhibited the "beating" that you often hear in in some designs from sum/difference LF harmonics. Not that I necessarily WANT that, but it does make me wonder if somethings missing since it's a common artifact of many good clipped tones. If ear testing reveals that a higher plate load sounds better I'll try the larger plate resistor and maybe alter the circuit with some compromise toward the total goal.

Dave, good call on the biasing. Yes it's a 1.5k (partial bypass on ck). More to experiment with. I'd neglected to consider the bias with the load. The actual LF change is part of the "trick" I'm doing with this circuit. LF rolls off as gain is increased. And I've used ltspice with a 38k voltage source serving the circuit (and my ears) to determine the coupling cap value.

Pete, I've read you recommending against too low a load before in the interest of linearity (sometimes on my threads ). And it may be that I did lose something by finally going TOO low. Will report back when I know if my experience parallels your own.

Another hitch is that I don't have free gratis for testing anymore due to my household circumstances. Not sure when I can get back in and follow with actual cranked up testing. Great stuff here and thank you all again. I'll be back here when I have something to add.

[Mike Sulzer]

Thank you gentlemen

Zozobra and Mike, this is an early stage that never clips. That doesn't mean it's not exhibiting any distortions, just not blatant clipping. I do notice that this amp (when clipping later stages and the power tubes) has never exhibited the "beating" that you often hear in in some designs from sum/difference LF harmonics. Not that I necessarily WANT that, but it does make me wonder if somethings missing since it's a common artifact of many good clipped tones. If ear testing reveals that a higher plate load sounds better I'll try the larger plate resistor and maybe alter the circuit with some compromise toward the total goal.

Dave, good call on the biasing. Yes it's a 1.5k (partial bypass on ck). More to experiment with. I'd neglected to consider the bias with the load. The actual LF change is part of the "trick" I'm doing with this circuit. LF rolls off as gain is increased. And I've used ltspice with a 38k voltage source serving the circuit (and my ears) to determine the coupling cap value.

Pete, I've read you recommending against too low a load before in the interest of linearity (sometimes on my threads ). And it may be that I did lose something by finally going TOO low. Will report back when I know if my experience parallels your own.

Another hitch is that I don't have free gratis for testing anymore due to my household circumstances. Not sure when I can get back in and follow with actual cranked up testing. Great stuff here and thank you all again. I'll be back here when I have something to add.

But even if you are not clipping in the first stage, the lower resistance load has reduced the voltage gain, and thus you do not generate as a high a level of lower order harmonics for the same level of input signal.

[mozz]

Use 12bz7.

[Chuck H]

Use 12bz7.

I did notice the specs for that tube could be an advantage in this situation, but all the ones I've had were so microphonic that you could record a vocal track singing at them!

[mozz]

Damm, i all the ones i have been hoarding i now have to take them out and test them? 12BZ7 microphone capsules for sale, theremin approved, analog space sounds, must sell.

[Gingertube]

Have a look at this data sheet.
http://tdsl.duncanamps.com/pdf/vm345.pdf
On the 2nd page there are curves of S(gm in mA/V), mu, and Ri or rp as we usually call it vs anode current.
There are 3 graphs for Va = 100V , 200V and 250V
The 200V one is the most typical for most guitar amps.

These curves are the most useful for understanding what happens in a gain stage.

If you bias cold (low Anode Current Ia) you are in the high rp part of the graph and the rp has the highest slope.
So in this part of the curve (low Ia) you get a large voltage division between internal rp and external load (anode load resitor in parallel with the loading circuit) - that is going to give less gain.

If you have a low impedance load that lowers the effective load of the anode load resistor in parallel with the load and so you have to swing more current for the same signal voltage output.
That extra current swing will give you more rp change. Lots more if cold biased in the high rp and high rp slope part than if warm biased (say at 1mA).
That is important because the distortion of the circuit is the change in rp vs the anode load effective impedance ratio. That is to say that the rp variation with signal current swing is the source of the circuits distortion.

You can also then conclude that if you have a lowish effective anode load then that ratio (the distortion) is higher.

You can also conclude that higher Anode Current biasing (Ia) will give more gain and less distortion than cold biasing

Since gain = (mu x Ra/(Ra +rp)) where Ra is the effective anode load then gain will be less when effective Ra is less (think of the 150K grid leak into stage 3 of the Trainwreck which lowers the gain of the 2nd stage and increases its distortion).
This applies only to a fully bypassed cathode bias resistor. Otherwise feedback effects have to be thought about too.

Notice that mu is fairly constant with anode current but rp and gm change considerably. So that gain formula can be applied with confidence.

Note also that the 62K rp value we normally assume is valid only at about 1mA. Most guitar amp stages run colder (less current) than that and the rp will be typically higher.

Not sure I've explained all well BUT examining those graphs tells you most of what you would want to know about a Guitar Amp gain stage. Shame really coz they are not the curves from the datsheet which are normally referenced and are not what is used to draw the load line on. In fact some datsheets don't even give you those graphs.

The gain stage can be thought of as (modelled as) a perfect voltage generator of mu x Vg1-k with a source resitance of rp (which varies with signal current) feeding the effective Ra.

Cheers,
Ian

[soundmasterg]

I was going to suggest the 12BZ7 too, but yeah they are often microphonic. I found they work great as a PI, with much less microphonics there, and I've used it in a project running a LTP and it is essential to the tone of that amp in that spot. Maybe you could try a paralled 12AX7?

Greg

[Chuck H]

Thanks. I haven't run my tests yet. I had a couple of hours of "empty house" time a day ago, but it was impromptu so I just played really loud instead of spending time soldering I have a little compromise room in the design so I'm not so much trying to find a solution as asking about tube behavior under the circumstances. Ian provided an excellent link, but I've never grasped tube charts so it's over my head. I've REALLY tried too. But something about the way they're drawn/aligned and the way I'm wired doesn't jive. Blessedly he did explain a little about what is demonstrated, which was easier to digest.

I may yet use parallel triodes, along with some other suggestions in the future for this sort of thing.

[Roberto]

Since gain = (mu x Ra/(Ra +rp)) where Ra is the effective anode load then gain will be less when effective Ra is less (think of the 150K grid leak into stage 3 of the Trainwreck which lowers the gain of the 2nd stage and increases its distortion).

Years ago I did several trials on changing the load of the 100k 39k cold stage of the Soldano SLO100, by lowering the value of the 2M2 resistor (that is in parallel with a 330k to ground, so 287k real load, no voltage divider): I preferred the sound with something around 150k total load (and a different next stage), because of the way the sound changed with the dynamics of the playing style. In that case the very low current helped alot, even if that big unbypassed rk brings back alot of counteracting local nfb.