That 6E5P looks pretty good over a selected span. Thanks adalin! Those tubes are (were?) still pretty cheap.
The 12HL7 and 6BQ7 I just found appropriate curves for on Frank's site, so no need to bother with tracing them. Not as good as the tubes looked at so far for cascode.
http://frank.pocnet.net/sheets/135/1/12HL7.pdf
http://frank.pocnet.net/sheets/093/6/6BQ7A.pdf
The 6E5P was used in the Bartoli article on Rod's circuit, and with %dist measurements even, so is a very useful reference curve to have.
Piano3: If you run across those TV tubes again, I would make a list.
PS:
I just noticed that
Vacuum Tubes Radio Tubes - 5,000 different tubes in stock - Over 10 million tubes!
has got a new list of cheap TV tubes on sale. Somewhat limited choice though.
Then there is also the existing TV tube sale list at ESRC:
Summer Dollar Days - Vacuum Tube Sale - $1.00 Vacuum Tubes
Maybe some $0.50 tube will work well for cascode.....
The 12HL7 and 6BQ7 I just found appropriate curves for on Frank's site, so no need to bother with tracing them. Not as good as the tubes looked at so far for cascode.
http://frank.pocnet.net/sheets/135/1/12HL7.pdf
http://frank.pocnet.net/sheets/093/6/6BQ7A.pdf
The 6E5P was used in the Bartoli article on Rod's circuit, and with %dist measurements even, so is a very useful reference curve to have.
Piano3: If you run across those TV tubes again, I would make a list.
PS:
I just noticed that
Vacuum Tubes Radio Tubes - 5,000 different tubes in stock - Over 10 million tubes!
has got a new list of cheap TV tubes on sale. Somewhat limited choice though.
Then there is also the existing TV tube sale list at ESRC:
Summer Dollar Days - Vacuum Tube Sale - $1.00 Vacuum Tubes
Maybe some $0.50 tube will work well for cascode.....
Last edited:
Found one!
Take a look at the plate curves (bottom page 6, for the triode section, especially the 150V Vp curve) for this 6KT8 TV tube. Only $1 at ESRC. Looks fairly straight Ip versus Vg1. (now we need to know if this is a real curve, or someones ruler extension...)
Seems this plate voltage range (approx. 175 V) is special thus far, but it may just be that most tubes are straightening out when approaching current saturation of the cathode.
http://frank.pocnet.net/sheets/135/6/6KT8.pdf
Take a look at the plate curves (bottom page 6, for the triode section, especially the 150V Vp curve) for this 6KT8 TV tube. Only $1 at ESRC. Looks fairly straight Ip versus Vg1. (now we need to know if this is a real curve, or someones ruler extension...)
Seems this plate voltage range (approx. 175 V) is special thus far, but it may just be that most tubes are straightening out when approaching current saturation of the cathode.
http://frank.pocnet.net/sheets/135/6/6KT8.pdf
Last edited:
It may be possible to straighten out the Vg1 versus Ip curve for many triodes by putting a little resistance in series with the plate. That way the "fixed" cascode voltage drops a little as the current goes up. Have to try this on the curve tracer to see how well it works.
Of course, putting some un-bypassed resistance in the cathode circuit always linearizes things too. But that is an asymptotic (NFdbk) fix. The small series plate resistor may just fix the problem straight away.
Putting some resistance in the screen circuit of a pentode may do similar things (at least until the plate voltage gets well below the g2 voltage, where the whole thing will collapse). Might be what the original RH-84 circuit was up to, with its non-trivial un-bypassed screen resistor.
Of course, putting some un-bypassed resistance in the cathode circuit always linearizes things too. But that is an asymptotic (NFdbk) fix. The small series plate resistor may just fix the problem straight away.
Putting some resistance in the screen circuit of a pentode may do similar things (at least until the plate voltage gets well below the g2 voltage, where the whole thing will collapse). Might be what the original RH-84 circuit was up to, with its non-trivial un-bypassed screen resistor.
Last edited:
I put the 6KT8 on the curve tracer here and it does look quite linear for Ip versus Vg1. You have a point though about the low gm. Will have to look some more.
That 6sp3-ev does look interesting, although it is difficult to tell just how straight the Ip versus Vg1 curves are with such a vertical scale on the datasheet. Everything looks like a vertical line.
http://frank.pocnet.net/sheets/113/6/6S3PE.pdf
I tried a 2.5K series plate resistor with a 6JT8 triode to see if that would improve the linearity, and it mostly just reduces the current levels by voltage reduction. By the time it actually linearizes anything, nothing will be left. I was hoping that a modest resistance might work.
I'll try a pentode screen resistor next.
That 6sp3-ev does look interesting, although it is difficult to tell just how straight the Ip versus Vg1 curves are with such a vertical scale on the datasheet. Everything looks like a vertical line.
http://frank.pocnet.net/sheets/113/6/6S3PE.pdf
I tried a 2.5K series plate resistor with a 6JT8 triode to see if that would improve the linearity, and it mostly just reduces the current levels by voltage reduction. By the time it actually linearizes anything, nothing will be left. I was hoping that a modest resistance might work.
I'll try a pentode screen resistor next.
Last edited:
Well I tried a 5K pot in series with the screen grid (triode wired) of a 6JT8 pentode section, and it is not much help at linearizing the curve. Mostly just attenuates it.
However, I did make a quite useful discovery. By lowering the filament voltage to 4.7 V instead of 6.3 V, the top part of the curve linearizes out. I can even make it curve the other way with lower filament V. So my earlier guess about cathode saturation effect was right on the money. Probably can fix most any tube now. Filament nightmares ahead though, every amplifier tube optimising at a different filament V likely. I guess one could use diode/rectifier strings (both ways for AC) in series.
But the filament V may have to be raised slowly as the tube ages and emission drops.
I'm going to try a higher value series pot with the screen grid again, not triode wired this time. Then I can adjust the screen V source up further to compensate for the pot V loss. Maybe that will offer more correction effect.
However, I did make a quite useful discovery. By lowering the filament voltage to 4.7 V instead of 6.3 V, the top part of the curve linearizes out. I can even make it curve the other way with lower filament V. So my earlier guess about cathode saturation effect was right on the money. Probably can fix most any tube now. Filament nightmares ahead though, every amplifier tube optimising at a different filament V likely. I guess one could use diode/rectifier strings (both ways for AC) in series.
But the filament V may have to be raised slowly as the tube ages and emission drops.
I'm going to try a higher value series pot with the screen grid again, not triode wired this time. Then I can adjust the screen V source up further to compensate for the pot V loss. Maybe that will offer more correction effect.
Last edited:
Here is a fairly coarse plot of the 6s3p-ev and a trioded 6J11P for comparison:
Attachments
Yes, Don that's a good observation about cathode saturation.
It is notable that triode-connected pentodes of a small physical size also work well ( and sound well), if their design operating current is moderate: say 12 to 20mA.
Presumably, the small physical size, or - the small heater power (PC86 is 3,8V 300mA from memory) - drives the cathode saturation threshold. This certainly fits my observation of good candidates for Shunt Cascode.
This behaviour has already been well evaluated - by the makers of LTSpice models for various valves, at least. The Russian 6Э5П (6E5P), which Ale Moglia spotted as a choice for Shunt Cascode, models this well, and shows very low distortion in the circuit.
Mona, you can use a resistor, if you prefer. But if you do that, the power supply must be regulated and very well filtered, and the cascode voltage must also be tracking-regulated (voltages must track together).
The power supply-noise-rejection is very poor in this case, so care must be taken.
On the other hand, you can use a resistor to get lower noise, if the supply is really quiet. I use this to advantage in a Shunt Cascode moving-coil stage, using JFET 2SK369 in place of the triode.
I have a few of these - bought because they look like the Russian PC86.
Had no time to test, sadly.
But you are right - high gm is all advantage, certainly up to 20mA/V.
Above that, the gain may make stabilising the circuit too difficult. The 6C45П may be a bad candidate - it is hard enough to work with, as it is!
Rod, if I understand this circuit correctly, the gain is set by the combination of the valve gm and the resistor from pnp collector to ground, so the higher gm allows you a lower output impedance for a given gain-assuming the fixed voltages allow you the necessary voltage swings. Have you found this hybrid circuit more unstable than a regular triode-triode cascode?
Yes, the gain is simply the (effective) gm x Rout.
And you are right, if you want low Zout for high drive applications, you can select high gm valves combined with lower value of Rout.
The precaution here:
- calculate the idle current required to set the idle output voltage at 50% of the desired output swing;
- Then check that the PNP transistor's safe-operating area is within limits (a signal-transistor like MPSA92 is desirable, with low capacitance, & moderate Hfe;
Stability gets more difficult with increasing physical size of the whole gain stage, and with increasing gain. The excellent tuner triodes are designed for UHF operation, so they will oscillate at any frequency, if you don't keep things under control!
I experimented with ordinary (series) cascodes only enough to find:
- that a bipolar transistor on the top measures and sounds better than any triode;
- that the sound does not compare well with Shunt Cascode.
- I had no trouble with stability, though. But typically, the gain is forced to be low, because the output resistor must carry the anode current. In fact, the series-cascode has many drawbacks, and only one advantage: it's easy to make.
And you are right, if you want low Zout for high drive applications, you can select high gm valves combined with lower value of Rout.
The precaution here:
- calculate the idle current required to set the idle output voltage at 50% of the desired output swing;
- Then check that the PNP transistor's safe-operating area is within limits (a signal-transistor like MPSA92 is desirable, with low capacitance, & moderate Hfe;
Stability gets more difficult with increasing physical size of the whole gain stage, and with increasing gain. The excellent tuner triodes are designed for UHF operation, so they will oscillate at any frequency, if you don't keep things under control!
I experimented with ordinary (series) cascodes only enough to find:
- that a bipolar transistor on the top measures and sounds better than any triode;
- that the sound does not compare well with Shunt Cascode.
- I had no trouble with stability, though. But typically, the gain is forced to be low, because the output resistor must carry the anode current. In fact, the series-cascode has many drawbacks, and only one advantage: it's easy to make.
It does work, and I tried this setup (en route to developing the Shunt Cascode version):
Vsupply = 475V
Va = 175V
Ia = 14mA
Upstairs = BU408 (NPN television line output type, very burly)
downstairs = PC86
Rout = 15K
Gain ~= 200
This works, but power supply rejection is terrible. The BU408 sounded better than any triode upstairs, indicating that gm is what counts in this position.
Shunt Cascode allows much more swing, - and sounds notably better.
I really meant with triode upstairs and down. In the circuit you describe, what is the load seen by the PC86? I do like the look of the shunt cascode; even for those allergic to silicon (not me), I cannot see that the transistor does anything except clamp the anode voltage and dissipate some power. Many of the textbooks do not cover thoroughly enough 2-transistor configurations. The only one which I possess which covers the folded cascode is "Analysis and Design of Analogue Integrated Circuits" by Gray et al.
With the lower device seeing a load of nearly zero ohms, the load-line follows the curves shown in this thread - because Va is effectively constant. You can see how good this is by lookking at the measured curves, kindly supplied by Adalin.
With a carefully chosen triode, and correct idle conditions of voltage & current, the transfer characteristic is very linear - AND high gain. This is true for series or shunt cascodes, but the series type has poor supply noise-rejection, and EMI-rejection, and lower range of gains.
With a carefully chosen triode, and correct idle conditions of voltage & current, the transfer characteristic is very linear - AND high gain. This is true for series or shunt cascodes, but the series type has poor supply noise-rejection, and EMI-rejection, and lower range of gains.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.