I've been experimenting with a constant current source substituting for the load resistor on a 6SN7. I was reading Broskie's Tube Cad Journal last night and noticed that he was using the CCS on the cathode. I wired that up and experimented with it too.
I suppose it's obvious to many, but you seem to need a bypass capacitor if you use the CCS on the cathode. No big deal because that was my intent anyway.
I find the use of CCS extremely effective, but a bit fussy. I've selected an operating point of 235v and 2.5ma. with the CCS on top (load resistor substitute), i need a trim pot on the cathode resistor to tune in the desired plate voltage. It works very well, but pop in a different tube and it will need an adjustment every time. I guess there is just that much variation between individual tubes.
In the cathode position I thought it might be easier to hit the operating point. Select the a load resistor to achieve the desired drop from B+ to plate OP given the design current and everything should just fall into place. No, not really. Now the operating point changes volt for volt with the B+ voltage. That cant be a good thing.
It seems like a coin flip, except that it's easier to trim a cathode resistor than install a voltage regulator. I have a TubeLabSE and note that George designed it with the CCS on top with a cathode trim pot.
I have not listened to these two options and that is of course where the rubber meets the road. So far it has just been scope and meter excercise.
So, I'm asking others with more experience what they have learned on this subject. Are there considerations I've overlooked? Has anyone used CCS on the cathode to good effect?
I suppose it's obvious to many, but you seem to need a bypass capacitor if you use the CCS on the cathode. No big deal because that was my intent anyway.
I find the use of CCS extremely effective, but a bit fussy. I've selected an operating point of 235v and 2.5ma. with the CCS on top (load resistor substitute), i need a trim pot on the cathode resistor to tune in the desired plate voltage. It works very well, but pop in a different tube and it will need an adjustment every time. I guess there is just that much variation between individual tubes.
In the cathode position I thought it might be easier to hit the operating point. Select the a load resistor to achieve the desired drop from B+ to plate OP given the design current and everything should just fall into place. No, not really. Now the operating point changes volt for volt with the B+ voltage. That cant be a good thing.
It seems like a coin flip, except that it's easier to trim a cathode resistor than install a voltage regulator. I have a TubeLabSE and note that George designed it with the CCS on top with a cathode trim pot.
I have not listened to these two options and that is of course where the rubber meets the road. So far it has just been scope and meter excercise.
So, I'm asking others with more experience what they have learned on this subject. Are there considerations I've overlooked? Has anyone used CCS on the cathode to good effect?
Last edited:
LOL. That one made me laugh nicely. However, the way real-world CCS's work, its not quite silence, but a darn good volume reducer. You can still hear a bit of signal leaking thru.
(Sadly…) I know because I learned the Bypass Cap thing the hard way.
Thanks for the chuckles.
GoatGuy
Yep… I continue to use the relatively ancient JFETs with grounded gates and small resistor values from source to ground. If you're eating a bag of Cheetos, and testing a bulk bag while watching The Munsters (as for example), its surprising how fast you can 'bin' the things into about 10 useful self-bias cutoff bins. Mark em as such, and play with them endlessly, at an original price of less than 1¢ apiece.
Just saying … let's not forget the OLD old days. Before everything came from Mouser, cough, cough, ahem, kachoo, honk, snort … Cheetos, a 9 volt battery, a breadboard, and a $9.95 (RIP) RadioShack check-out counter "basket" special DVM digital volt meter.
GoatGuy
> the operating point changes volt for volt with the B+ voltage.
This is why you want tube stages with a resistor on the top and a resistor on the bottom. They self adjust.
Start "fixing" things, and you have to fix EVERYthing.
Ultimately you end up awful close to an Analog Chip, ALL op-points fixed by design. "Sterile". "Lifeless".
If you truly want a fixed current and a fixed voltage, put a CCS in the plate and a fixed Voltage grid-cathode. LED is simple but only comes in a few voltage drops. Grounded cathode and a small regulated grid bias has advantages (no crap in the cathode path). The plate voltage depends only on tube Mu, which for a given tube and op-point is pretty much baked-in by electrode dimensions and spacing. Oh, a good slam might throw Mu off a bit. And there is tube-to-tube variation. But who is to say that a higher-Mu tube might not benefit from a higher plate voltage? Endless pondering possible.
This is why you want tube stages with a resistor on the top and a resistor on the bottom. They self adjust.
Start "fixing" things, and you have to fix EVERYthing.
Ultimately you end up awful close to an Analog Chip, ALL op-points fixed by design. "Sterile". "Lifeless".
If you truly want a fixed current and a fixed voltage, put a CCS in the plate and a fixed Voltage grid-cathode. LED is simple but only comes in a few voltage drops. Grounded cathode and a small regulated grid bias has advantages (no crap in the cathode path). The plate voltage depends only on tube Mu, which for a given tube and op-point is pretty much baked-in by electrode dimensions and spacing. Oh, a good slam might throw Mu off a bit. And there is tube-to-tube variation. But who is to say that a higher-Mu tube might not benefit from a higher plate voltage? Endless pondering possible.
CCS in the anode circuit makes some sense, as it presents a high impedance and so maximises gain and minimises distortion.
Putting a CCS in the cathode is a waste of time, as there is no reason why you need to set the anode current precisely to any particular value. Cathode resistors are best, as they are self-adjusting. If you really want to fix the grid voltage, use an LED.
A valve has three parameters: grid-cathode voltage, anode-cathode voltage, anode current. You get to fix two of them; the valve fixes the other one. The best option is to use resistors, then you get to fix two linear combinations and the valve fixes whatever you left unfixed. Valves can vary from their datasheet values by up to around 30% without any fault being present.
Putting a CCS in the cathode is a waste of time, as there is no reason why you need to set the anode current precisely to any particular value. Cathode resistors are best, as they are self-adjusting. If you really want to fix the grid voltage, use an LED.
A valve has three parameters: grid-cathode voltage, anode-cathode voltage, anode current. You get to fix two of them; the valve fixes the other one. The best option is to use resistors, then you get to fix two linear combinations and the valve fixes whatever you left unfixed. Valves can vary from their datasheet values by up to around 30% without any fault being present.
All good points. I’m trying to get a small signal triode to swing about 350 volts to improve on an 845 driver circuit (that is really crap). I’m a bit constrained by the existing power supply and chassis (which are actually very good). It’s a lot to ask, but the anode CCS with LED bias gives up the best voltage and wave form. I was initially put off by the LED option because I couldn’t hit the exact OP I wanted, but that seems pedantic now that I’ve really investigated the options. I’ve been dissuaded from the cathode CCS by my trials and your comments. I think the CCS is necessary, at least in the second stage where the big voltage swing happens.
Did you get any idea of the dynamic resistance? There is very little info on the data sheet.
No it's not... it does wonders with cathode coupled phasesplitters and the like.
Putting a CCS in anode line ? The anode will be presented with near infinity impedance unless well damped by a following stage. I have plenty of evidence with high gm tubes, often tamed as triodes that instability can occur with just the oscilloscope probe capacitance. This too hair raising. Bear in mind the traditional anode resistor acts as a damper so I continue to use it.
I tried the CCS anode method with a flighty high gm 12BY7 in triode mode used as a current driver, and it oscillated. With an internal Z reduced to a few kohms it objected. As I see it, there are limitations.
rJ
Very true. And NOS versus new prod, add even more %, close to 50% variation. This is a nightmare for DC design.
There's a nice datasheet out there… https://www.mouser.com/datasheet/2/240/98704-1546069.pdf
But rather vexingly, it does not actually state the what R for desired mA formula. I just invested, oh, 30 minutes squinting at the datasheet (above) and transcribing various R and mA values to a spreadsheet. Determined ('cuz I'm old and irascible) that the best fit parameters are
R = e(8.15 - 1.1 ln(mA))
where the r-squared (computed estimate of linearity of eyeballed data to a straight line intercept and slope) is 0.9998. Which is rather good as data curve fitting goes.
So, download the PDF, and print it. Then write the above formula next to the graph. Biblically useful formula.
If you want it in base–10, the same formula is
R = 10(3.54 - 1.1 log10(mA))
Just Saying,GoatGuy ✓
… highlights added by me … to point out why I really don't like using semiconductor constant-current loads in valve circuits. Seriously: the best solution, now deeply out of favor due to mass and cost, is inductor anode loading.
The “beauty of choke loading” is that the choke 'figures out' an quiescent current that its happy trying to keep constant due to the usual formula:
V(t) = L di/dt
Except perhaps losing people who've not yet taken calculus, there isn't a more straight-forward way of describing a choke's behavior. Subject a choke to current variation, and it'll happily impose a I-don't-want-to-change-dammit voltage in turn. The “problem” of course is that fairly high value chokes need to be deployed for it to work well at the lower frequencies. In really boldly-optimistic-simplicity terms,
Z = 2πLF
so forF = 20 Hz
Z = 10,000 Ω (pulled out of my hat)
thenZ = 10,000 Ω (pulled out of my hat)
10,000 = 6.28 × 20 × L
L = 10,000 / (20 × 6.28)
L = 80 H
Which is a high value choke to be sure. Costy. L = 10,000 / (20 × 6.28)
L = 80 H
Yet, I've built no end of amplifiers with choke loads “in my day”, and they tirelessly work to do exactly the self-adjusting-yet-current-source-acting job they're intended to do. Valves age. Chokes say: “so what!”.
Just Saying,
GoatGuy ✓
Hey, that's great. I'll get working on it right away. I suppose it would be much the same for the 900 Volt version.
Unfortunately… the datasheet of IXCP10M90S (900 volt, 100 mA max) doesn't promise the same kind of log-log domain curve-fit linearity. If you are truely interested, I can figure out the best-fit curve for the IXCP10M90S, but it'll take at least an hour of fiddling. Better in many ways is just to use the PDF for the thing, and a ruler. To find the R for a desired mA.
You don't have to be terribly accurate! (a factoid almost lost, in this day of having 16 digit calculators on ones' smart phones!!!) ±20% is good-enuf.
Just Saying,
GoatGuy ✓
So to say it in words, just to be clear, Rdyn equals 10 to the power of (3.54 minus 1.1 times log to the base 10 of the milliamps). Have I got the brackets in the right place?
--- No. ---
The 'R' in the above formula is the R needed in series with the cathode of the current-regulator chip to 'set-point' its current value.
The datasheet (pdf, which I included a link to) itself states the dynamic impedance of the chip as a current source. a minimum of 160 kΩ…
You're welcome.
In advance.
Just Saying,
GoatGuy ✓
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.