I have still not made it, do have chokes that say 235H/1100 ohms.
The simulation is just a jot together, to see what happens. Just on one tone, the parallel load resistor is not needed; in a sweep it is. In practice I do not think it is needed.
I started thinking about feedback after seeing the Elekit; here it is mucho better (DC balaced!). As well I have tried a feedback between the two cathodes, also good, but more delicxate to get richt. Why FB: I think the 300B I have at least, the carbon type, is not better than my old 10Y. Using the2K-LL2752 I get dynamics but not to great stage. I used all kinds of driver stages, good was 5751/CCS load, currently a two stage 2SN7 - indeed not comfortable.
What I will do is use another tube; in the past I have triode-strapped EF91, EF804, (strapped also low mu if I remember, not much better than the 18-20 dB of the ECC82) and they will give the output swing with ease, but do not have the zappas to provide some grid current. With a choke load that might be different. I expect the low resistance choke in itselfstops any grid blocking.
Please shed a little light on 'DHTs are not recommended for Shunt Cascode', I saw an article on the 3A5 used/implemented in this way. I also have some DH-pentodes.
But I am a bit wary of experimenting, I spent too much on that.
So >> a new run of the PCB will have my interest.
This can be made very near to zero, with a little shunt regulator; that's how I do it with my boards. The swing in the base current is very small, so the regulator need only be low-power, and the circuit complexity need not hold you back.
Member
Joined 2009
Paid Member
As I see it, to avoid the need to use a high load resistance for the cascode which eats up voltage / forces you to use lower bias current, you want a high transconductance in your tube - whereas most DHT’s have low transconductance but are very linear when used as a triode.
Hi transconductance tubes require the grid to be wound very close to the cathode and these techniques were perfect later in the tube era so you will be looking for modern space Age tubes for high transconductance, perhaps frame grid type
Hi transconductance tubes require the grid to be wound very close to the cathode and these techniques were perfect later in the tube era so you will be looking for modern space Age tubes for high transconductance, perhaps frame grid type
DHTs have gm that is too low. It's enough to make us look elsewhere. The advantage of DHTs show up with reactive loads (loudspeakers) and very high anode-voltage swings. We do not need either of those for this application, and the high-gm tubes offer high gain, and good compatibility with the vertical load-line.
Member
Joined 2009
Paid Member
But unfortunately, transconductance is usually a strong function of current, making it a non-linear property of all amplifying devices unlike mu where the non-linearity of the transconductance is balanced by the nonlinearity of the tube’s internal resistance. So whether using a cascode or a pentode you’ll still get some distortion. It can still be a very pleasing sound, dominated by second harmonic and monotonically falling higher orders. And importantly you can get the gain that you need to drive a low mu output triode.
The output spectrum shows the reality of the distortion. That's at 150V p-p, and it declines predictably at lower output.
Gm of some of these newer valves saturates at the working anode current, see the curves for gm in 6E5P or anode current in PC86 for example.
Moreover, you can drive a 300B directly from a DAC with this stage: no buffer, no linestage, no extra gainstage, as is often required. The order of harmonics is increased by adding these extra stages, which is usually audible.
With Es9018 DAC and the like, you can use the DSP volume control, and connect the DAC pins directly to the driver stage.
And the Shunt Cascode has no Miller effect, much kinder to opamp based DAC output stages.
With a
Gm of some of these newer valves saturates at the working anode current, see the curves for gm in 6E5P or anode current in PC86 for example.
Moreover, you can drive a 300B directly from a DAC with this stage: no buffer, no linestage, no extra gainstage, as is often required. The order of harmonics is increased by adding these extra stages, which is usually audible.
With Es9018 DAC and the like, you can use the DSP volume control, and connect the DAC pins directly to the driver stage.
And the Shunt Cascode has no Miller effect, much kinder to opamp based DAC output stages.
With a
How about the EL84 and the ECL variants of this pentode (I have a box . .)? They are know to be linear in triode but I can't get tye data of the use in this situation.
EL84 can be made to work, but the data-sheet gm of 10mA/V is given at an anode current of 50mA, and that is not helpful for heat dissipation. Lowering the anode current would make the gm too low, and the bias voltage relatively high.
Better to buy a few IF pentodes - 6F30 or EF184 are very useful, and the 15.6mA/V rated gm (at Ik=14.1mA) gives a raw gain of 420 (open loop) with a 27K output resistor. Or the 6Э5П (6E5P) for even more gain, and the distortion performance shown above.
Better to buy a few IF pentodes - 6F30 or EF184 are very useful, and the 15.6mA/V rated gm (at Ik=14.1mA) gives a raw gain of 420 (open loop) with a 27K output resistor. Or the 6Э5П (6E5P) for even more gain, and the distortion performance shown above.
I have simulated a bit more. [But I will probably go for the ready boards in due time . . .] But I am very tempted to do the choke load because of the DC coupling that is then possible and the felexibility to the Rload.
Becaues of the DC nature of the 'output' of the driver with a Choke the output depends on the value of the choke. It has a rather high -3dB in the low frequencies -for instance 22-25 Hz.
With a parallel load resistor, say 25 K then it extends to below 10Hz.
I tested the circuit with a nuvistor (just for fun, Spice alows me to do it, same results for ECC88) and a rather big and unbypassed 5K Rk I got almost the same results as with a 6SN7 unbypassed 2k2, though I can't say for sure the harmonics are better: just 1 dB improvement of the second harmonics [@1 V in, 3V pp out, no feedback).
- At the edge of overload both drivers I have on paper (Choke vs CCS) also are showing the same high but neat degradation of harmonics.
- > Having a CCS at the output of the driver to get it to about zero DC requires a servo, to overcome real-life drift that the LTSpice does not show So let's say the CCS is impractical.
Feedback from driver-out to grid-in of the driver can be used at will.
With 220k to 12K the FB is about 6 dB.
The harmonics drop from 2=63, 3=75 (2Vpp, no Fb/220G) to 2=48, 3=81 (1Vpp 6dFb with 220k).
And in real life I am tempted to ditch my 300B in favour of a 10Y. I normally run at < 85 dB room noise. So that will be my next sim.
And then the next sim is: what about a 10 driving a 10Y? (needed: four Coleman reg boards. . a bonanza) I have seen the 10 behaves almost the same as the ECC82 in this circuit on paper.
Becaues of the DC nature of the 'output' of the driver with a Choke the output depends on the value of the choke. It has a rather high -3dB in the low frequencies -for instance 22-25 Hz.
With a parallel load resistor, say 25 K then it extends to below 10Hz.
The load resistor in practice can be chosen ad lib between practically 25K and 215K - there is almost no influence on the DC point of the driven grid. That means the amplification (mu x Rload) can be varied. So that is a big advantage over the 'traditional shunt-cascode' driver with its fixed resistor load (and a coupling cap).
I tested the circuit with a nuvistor (just for fun, Spice alows me to do it, same results for ECC88) and a rather big and unbypassed 5K Rk I got almost the same results as with a 6SN7 unbypassed 2k2, though I can't say for sure the harmonics are better: just 1 dB improvement of the second harmonics [@1 V in, 3V pp out, no feedback).
- At the edge of overload both drivers I have on paper (Choke vs CCS) also are showing the same high but neat degradation of harmonics.
- > Having a CCS at the output of the driver to get it to about zero DC requires a servo, to overcome real-life drift that the LTSpice does not show So let's say the CCS is impractical.
Feedback from driver-out to grid-in of the driver can be used at will.
With 220k to 12K the FB is about 6 dB. The harmonics drop from 2=63, 3=75 (2Vpp, no Fb/220G) to 2=48, 3=81 (1Vpp 6dFb with 220k).
And in real life I am tempted to ditch my 300B in favour of a 10Y. I normally run at < 85 dB room noise. So that will be my next sim.
And then the next sim is: what about a 10 driving a 10Y? (needed: four Coleman reg boards. . a bonanza
) I have seen the 10 behaves almost the same as the ECC82 in this circuit on paper.
Yes, 10Y or, in fact any DHT is much better saved for output stages, or line stage outputs.
The Shunt Cascode loads the triode with a vertical load-line, and this is not suitable for any DHT, really.
Look for more recent high gm triodes and pentode (triode connected), tuner triodes, high quality frame-grid bottles - these are made for the Shunt Cascode circuit.
The Shunt Cascode loads the triode with a vertical load-line, and this is not suitable for any DHT, really.
Look for more recent high gm triodes and pentode (triode connected), tuner triodes, high quality frame-grid bottles - these are made for the Shunt Cascode circuit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.