Occasionally, you get what you ask for

shoshin

Site Supporter
Site Supporter
In a positive way. 😉

Back in my a year is a long time... thread, I wished for a current production 8B/PX-25 type tube. Such a tube would have almost ideal gain and Rp (at reasonable current) for a two stage amp, either as a driver with interstage step-up to drive a low-medium power output stage or even using the same tube in each stage (with only mild interstage step-up, if any).

It seems my wish, or better, has been granted: EML 12B preliminary datasheet.

This could change my proposed next build, which due to work continues in fits and starts.
 
Last edited:
It is definitely more plausible to use something like that with a 5K interstage, and certainly it's reasonable with a 10K IT. Heck, a 10K IT with a 1:2 voltage step-up would be tempting.

Your custom winder could probably pull off a 1:1 with 70H of inductance. It looks like a win for sure.
 
Your’e right Paul: it does look like a win. Would be a simple, clean build too. Lots of options for using a tube like this.

A lot of discussion went into the current design though... but I have learned from that and the 12B may just be a bit too tempting.
 
Last edited:
Jac has provided curves and basic data in the prelim datasheet: I'd really like firm data, including rp and H2 distortion, at various operating points/loads. Seeing the absence of data as an opportunity, I was encouraged to estimate - er, learn/practice how to estimate(!) - some info from the curves. I do find working from curves a bit imprecise; maybe more like indicative.

rp is about 2K5 at 375V/42mA: higher than I would have liked, but workable. Transconductance calcs to 5.6mA/V: exactly per Jac's datasheet.

Paul Joppa's recommended calcs suggest an "optimum" inductive load of 375V/0.042 - (2.4*2,500) = 3K. That can't be right - seems too low. I'm guessing I stuffed something up or the calc is more suited to lower mu, lower rp tubes and does not apply here. It does seem odd that everything else being equal with those calcs, a higher rp tube would require a lower load (2.4*rp is subtracted).

H2 distortion seems low with this tube. Again, using curves to do this seems, well, imprecise. For the operating points above, a 10K load = 0.3% (seems too low), 7K load = 1.67%, and a 5K load = 1.7%.

I didn't do power calcs etc. - maybe later.
 
Last edited:
Yes, this is a great place to use a ~10K interstage.

Is this what you're using:

"Once you’ve chosen your operating voltage and current, you need to find a load impedance. A good approximation is the operating resistance minus 2.4 times the plate resistance. I derived this in a 3-part article in VALVE; I won’t try to cram the math into this post though!"

I would go through PJ's derivations in the valve articles and see where he might make some assumptions regarding typical values of mu for DHTs.
 
It's funny, if you start with the March 1998 Valve, PJ's roughest estimation is that loading impedance could be about operating voltage divided by operating current. That gets you 9K :)
 
Yes, this is a great place to use a ~10K interstage.

Is this what you're using:

"Once you’ve chosen your operating voltage and current, you need to find a load impedance. A good approximation is the operating resistance minus 2.4 times the plate resistance. I derived this in a 3-part article in VALVE; I won’t try to cram the math into this post though!"

I would go through PJ's derivations in the valve articles and see where he might make some assumptions regarding typical values of mu for DHTs.

That's what I'm using Paul, picked up from the Asylum a while back. I will take a look at the Valve articles - I have a link somewhere.

It's funny, if you start with the March 1998 Valve, PJ's roughest estimation is that loading impedance could be about operating voltage divided by operating current. That gets you 9K :)

Interesting re the rough estimation. It nets a similar result to Valve Wizard: load = anode V squared / anode dissipation, that is Z=Va^2/Pa = (375*375)/15.75 = 8930Ohms. Close enough to 9K for me. Surprise! 😉 I like to err on the side of lower distortion, so 10K is about as low as I'd go. Running as low 7K could also be justified, I guess, if my H2 distortion estimation (1.7%) is off by no more than 30%, though I might be concerned about the output impedance.

The calcs I did yesterday were for an output stage. The Input/driver stage will run similar voltage/current ratios and should have similar rp - I expect the same load to apply... I will check it. Later today I hope to play with a different output stage operating point, something near 350V/50mA.
 
Last edited:
If you pull the trigger and buy a pair of these driver tubes, you could temporarily set them up with a CCS and cap coupling on a piece of plywood, then use a big coupling cap (10uF) and a 50K pot setup as a variable grid leak. You could play with different grid leak values to measure distortion cancellation and get some listening impressions before going to your winder to have the interstage transformer made.
 
If you pull the trigger and buy a pair of these driver tubes, you could temporarily set them up with a CCS and cap coupling on a piece of plywood, then use a big coupling cap (10uF) and a 50K pot setup as a variable grid leak. You could play with different grid leak values to measure distortion cancellation and get some listening impressions before going to your winder to have the interstage transformer made.

Might be something I can do as the COVID-19 control measures tighten and I work from home. I'd need to get some equipment in place to be able to do it...
 
A 350V/50mA operating point could be okay for an output stage. rp would be a bit lower and rough(!) estimation methods put the optimum load at 7K. Mu of the tube would be lower (but with less step down in the transformer). Dissipation at 17.5W is still well below the conservative 20W max.

I found it difficult to calculate distortion because the operating point had moved off the drawn curves, exacerbating imprecision - estimated distortion increased despite expecting a decrease (more current relative to voltage). Triodes are pretty flexible and estimations are often sufficient... but I'd like to be able to run sims and measure.

And regarding the estimation of the load, the rough methods appear to give a load about 30 - 50% greater than PJ's more sophisticated/accepted methods. Using the rough approaches, a 2A3 at classic points would need 4K2, my 2A3 operating points would need 6K (4K was the calculated optimum), and the Type 50 at classic points >7K (the recommendation is 1/2 this, though I suspect 4K5 would be better)...

I think that is enough for now... my head hurts. 🤯
 
Last edited:
I read through Paul's workings in VALVE issues 3 and 4, 1998. I could not follow it all, especially the derivations in issue 4. One thing I noticed was the formula RL/RO=0.7. This is the primary load impedance/operating resistance (Plate voltage Eb/Plate current Ib)=0.7. To solve for the load then is (Eb/Ib)*0.7... at least in issue 4. I assume this has changed since, though this calculation seems widely applicable. An important thing to understand is that Paul's calcs are (handy) rules of thumb, based on models of models, to be used as a sound starting point for a design.

Without SPICE skills or other modelling tools, I'm left with graphical estimations using curves. That's okay, if still a time-consuming estimation. And some knowledge of what is likely to be good performance is still needed. I've only done this with the EML12B, with some parts done for other tubes for context. I think I need to do the graphical thing on a nice 300B operating point and then maybe a 45 or 50 (for higher rp tubes) to flesh out my knowledge a touch. Interesting is that the highly regarded Type 50 recommended operating points suggest a low load and unsurprisingly has a high estimated output Z - I'm curious to guesstimate H2 distortion.

I've attached some of the graphical stuff I played with today; probably not interesting to anyone but me! It was painful as the line locations were changing relative to the curves as I zoomed in and out! Hope they print to PDF okay.

Gotta run.

Edit: there is a problem with the calcs: I incorrectly used the plate-to-cathode voltage, not the the full B+ voltage, in determining the loadline. For the 12B it is not a big deal and would make only a small difference (slight improvement, generally).
 

Attachments

  • EML 12B operating points.pdf
    857.6 KB · Views: 9
Last edited:
This is more prep work than I usually do. When I have uncertainties, I will lash stuff together on a piece of plywood, then take measurements.
 
I agree Paul - at some point, probably earlier than where I am now, measurement is the way to go. I'd like to get a basic lab set up. Space is tight here but nothing some planning and organisation won't fix. Until then, these graphical estimations are all I have and they have helped verify that the tube is worth pursuing.
 
A bit off topic.

I wanted to check my working/understanding against some known tubes and their operating points, which as a byproduct, deepened my understanding of these interesting devices.

I like and use the 2A3. The 2A3, at classic operating points, generates about 3.5W of power at 5% distortion and roughly 5W before its grid is driven positive, albeit approaching 6% distortion. A decent, if not outstandingly linear small DHT. Its rp and load requirements mean it runs about 2.5Ohm output impedance, about as high as I'd like for full-range duty, even for speakers designed for high source impedance drive. The 2A3 can get better with different operating points and loads, but not near 300B territory, I think. So why do I like it? Maybe it just sounds good despite (because of) the distortion... or the distortion is low enough in use (say <1W). The possibility of using AC filaments is a bonus. I do enjoy my amps that run it - at least with JJ or EML - at admittedly a more linear part of the curves, so I just dunno. Maybe time for something different.

Although I have not done anything extensive, I'm rather impressed with what has been achieved with the venerable 300B. At similar operating points to the 2A3, the 300B produces more power at about 60% of the distortion, with similar gain and lower practical output impedance. Linear. And it scales to comfortably achieve more than twice the output power of a 2A3 while still maintaining linearity. It also is easy to implement, though almost mandates from DC filaments. Sound? Well, that is a personal choice, but I reckon it can be had in the desired flavour - just need to find the right supplier and operating points. I have a newfound respect for this tube: I think it is cool in its competence. Which probably makes me decidedly uncool.

The PX4 was also interesting. Some folks love the sound of this tube. It appears to trounce the 2A3 in linearity... and could be better even than the 300B looking at some of the figures, but seems to me similarly linear as the 300B when working the curves. It has more gain than both, though if you want really low distortion you lose the benefits of the gain in use due to the greater step-down when using the higher (but not prohibitively-high) load. The PX4 also so has an output impedance in the ballpark of the 300B. The EML curves and figures are similar but different to the Osram curves; maybe just the natural variance that can be seen in manufacture. I raise the EML option because I don't entertain using unobtanium NOS tubes. If you have the UX4 socket and filament current covered, and you like their take on sonics, the EMLs could be nice if looking for 2A3-ish power... at around half the distortion and a lower practical output impedance.

I'm curious now about the AD1 - I will take a look at some point. I have not looked at the 50 - a quick eyeballing of the data suggests the output impedance when loaded as recommended (to get decent power) to be too high for me - maybe it would ideal if relieved of bass duties? Nor did I look at the 45 - power is too low for me, I think.

This is pretty obvious stuff to those who have been around, and yeah, I guess I've "known" it through osmosis... but it was real nice to work through the curves and charts to "see" it (I hesitate to say understand it - I really don't!).

And finally, back to the subject of this thread, the EML 12B, which I thought could be used as an output tube. It can, if you are looking for 45-like output power at about 70% greater plate dissipation. The high rp of the tube means that you simply can't have low (enough) output impedance and sufficient power at the same time, assuming you can find a good enough output transformer to provide the load and bandwidth. At about 17W plate dissipation, a little over 2W can be managed at an output impedance somewhere around 2.5Ohms in use, if my calcs are right. On balance, it is not what I’m looking for in an output, even with distortion far less than the others mentioned here. The 12B would be an awesome driver, and could be used in a 2-stage amp with modest 1:2 step-up somewhere in the amp... just as Jac intended.

Okay, enough on this topic. I need to get my posterior into action and finish a design and start a build.
 
Last edited:
It's good that you're going through this exercise and keeping in mind what you can actually get out of an output transformer. This is a bit of an advantage for the 2A3, as there are plenty of 2.5K-5K transformers with 15-17H of primary inductance, which favors the classic 2A3 operating point. This can be doubly compounded for those who use something like the Tango XE-20S/ISO FC-20S transformer that works really nicely at 2.5K but not so much at the higher impedance possibilities listed on the datasheet.
 
Agree Paul. The load really changes things... even before we consider the set of compromises the transformer imposes. Still, if the required load is reasonable, transformers still seem like a really good solution.

I‘d rather not need an output transformer with over 5K primary impedance. The PX4 with 5K can have what I think is low distortion at full power: 3.8W at 2% on the EML website, but I could not replicate that from the curves - I got more like 3.8W at 2.5%, which is still rather good - better than a 300B at published low distortion operating points, for example.
 
Last edited:
I’ve decided to use this tube (EML 12B) in my next build... as input/driver, not output. Some passive gain will get me what I need and I won’t have to screen it like the higher mu EML tubes, though a grid stopper will be used.

Thanks Paul for you contributions - much appreciated.
 
Last edited:
Back
Top