It's been a long time since I had any experiments on the bench and I decided it would be interesting to see what a type 30 could do. So I wired up a simple gain stage with a 10M45 as a plate load and captured a few numbers. Nothing really surprising except for the poor HF frequency response. Later on I plan to use a plate resistor and see how it behaves. Anyhow here's a basic schematic. The filament is lit with DC.
1 volt RMS input
I'm starting to suspect the 100k input impedance of my analyzer is causing the rolloff. While verifying the 50KHz reading I put my bench meter in parallel and it loaded it down further to 5 volts.
1 volt RMS input
| 1KHz | 0.30% THD | 8.4 Vo |
| 10KHz | 0.39% THD | 8.1 Vo |
| 20KHz | 0.66% THD | 7.5 Vo |
| 50KHz | 1.7% THD | 5.5 Vo |
I'm starting to suspect the 100k input impedance of my analyzer is causing the rolloff. While verifying the 50KHz reading I put my bench meter in parallel and it loaded it down further to 5 volts.
astouffer,
I hope you did not tear down your circuit yet.
Try changing the 1k CCS stopper resistor to 250 Ohms.
Then re-measure the high frequency rolloff.
Please report back to us if the high frequency response improves.
The 30 grid to plate capacitance is large, 6pF. If the gain is close to u (9.2). The Miller Effect capacitance is 55.2 pF.
That is almost as much as a typical 300B stage's Miller Effect capacitance.
Drive the 30 grid from a medium impedance, or the 55pF will cause high frequency roll off.
I hope you have happier measurement results.
I hope you did not tear down your circuit yet.
Try changing the 1k CCS stopper resistor to 250 Ohms.
Then re-measure the high frequency rolloff.
Please report back to us if the high frequency response improves.
The 30 grid to plate capacitance is large, 6pF. If the gain is close to u (9.2). The Miller Effect capacitance is 55.2 pF.
That is almost as much as a typical 300B stage's Miller Effect capacitance.
Drive the 30 grid from a medium impedance, or the 55pF will cause high frequency roll off.
I hope you have happier measurement results.
Last edited:
I think You right?
Dont forget that at the input of analizer You have
Cak = 2.1p + 0.7p of socket. But that is minor value to add on the cable capacitances...
Internal resisistance Ri=10300 ohms for 30. And the load is huge so in parallel it is almost the same
Ri value for output to the analizer?
.
But maybe to check source generator to input of the 30?
From the datasheet of 30
Cgk=3p + 0.7p (socket contacts) = 3.7pF
Cga=6p + 0.7p (socket contacts) = 6.7pF
-A in in Your configuration of Load is very close to mju of the 30
mju=9.3 so -A=8.9 (or even more)...
Cdynamic = Cgk + [1+abs(-A) x Cga] = 3.7pF + (9.9 x 6.7pF) = 70pF cca.
XCdynamic = 1 / (2 x PI x Fo x Cdynamic)
XCdynamic(50KHz)=45473 ohm
The equivalent resistance of the source, generator in this case should be less than XCdynamic at Fo.
Source with Req=XCdynamic will form an RC -3db filter @ Fo
...
With Your source resistance of 100K there is larger -db loss
cheers
Last edited:
Its not the Zin of your analyzer. The Hi Z of your circuit, and the interconnect cable and the input capacitance of the analyzer forms a first order low pass filter, rolling off the HF response.
Just insert the source or cathode follower between the output of your circuit and the input of your analyzer to isolate the hi Zout from the capacitances.
Art
Just insert the source or cathode follower between the output of your circuit and the input of your analyzer to isolate the hi Zout from the capacitances.
Art
The 50kHz roll-off means cca. 3μs time constant. The 10kohm internal resistance of tube and 300pF load capacitance create this 3μs. Where is the 300pF in this circuit? I don't think CCS (it seems too much). If you take a 47kohm resistance instead CCS, you can see (or not) the change of hi-freq. roll off. If the loading capacitance is too big (for example shielded cable of analyser), the roll-off's freq. rises only a little bit. If the roll-off freq. rises dramatically, the CCS is the guilty.
This is a triode with 10k inherent resistance. This determines the small signal roll-off. The parallel 100k can't show essential effect.
The CCS load can maximize the amplification to mu value. The distortion will near minimum, but some cancellation trick (SRPP-like mu-follower, but not total CCS implementation) is better (not better sounding, but smaller distortion).
An integrated CCS like the 10M45 has an equivalent impedance in the 1 to 5 megohm range. Anything you connect across it like the input resistance and capacitance of the next stage will have a much greater effect on the total performance that if the plate was loaded with a resistor of 100K ohms or so. The performance with the CCS in place of the resistor should always be better than the resistor itself.
Want to see some really strange results? Stick a 10M45 in place of the resistor in a pentode gain stage, and buffer its output so there is a nearly infinite load on the tube. Gain goes into the 1000X plus region but large tubes like the 6SJ7 with wound grids become microphones. The right tube for this job in the tiny little 6AK5. Apply a small amount of local negative feedback and medium gains at THD's in the 0.1 to 0.3% range are possible. Experimentation is ongoing.
Want to see some really strange results? Stick a 10M45 in place of the resistor in a pentode gain stage, and buffer its output so there is a nearly infinite load on the tube. Gain goes into the 1000X plus region but large tubes like the 6SJ7 with wound grids become microphones. The right tube for this job in the tiny little 6AK5. Apply a small amount of local negative feedback and medium gains at THD's in the 0.1 to 0.3% range are possible. Experimentation is ongoing.
Ha, I thought about that a while ago, but did not try it, decided that a pentode is almost a CCS, and it means they would be fighting each other. Very interesting. Shows that one should try things even if they feel weird 🙂
In theory the gain would be infinite. There might be a single infinitesimal point where the pentode and the CCS are set to exactly the same current which should put the plate at about half the supply voltage. Any perturbation in the current through either device would push the plate voltage to the B+ rail or to the cathode voltage.
In reality there are no perfect pentodes, and no perfect CCS circuits. This creates a very high gain amp whose gain is dependent on the Gm of the tube and the equivalent impedance of the plate load. This is the CCS circuit's AC impedance in parallel with the load impedance which is often another tube circuit. Both of these usually have a frequency dependent component.
Initial experiments were for guitar amp applications where some distortion can be tolerated especially if it came with buckets full of gain. It is also possible to configure a mosfet or 10M45 chip so that it is wrapped around the plate load resistor in a circuit sometimes called a gyrator since it operates somewhat like a large inductor.
Look at the input stage of this ultra low budget guitar amp. Mosfet Q1 buffers the plate voltage and puts it on the top of the load resistor. A resistor with the same AC voltage on each end flows no AC current, so its impedance is very high. The pot (R9) is 2 meg ohms which is in series with the parallel resistance of R6 and R7 at AC. This variable resistance is the load seen by the pentode for all important frequencies (82 Hz to 5 KHz). One knob adjusts the gain from about 50 to about 1000 with a fairly low Gm tube. The 18FW6 is a 6AU6 with an 18 volt 100 mA heater. Extremely high Gm tubes can be quite unstable in this circuit.
In another crazy experiment looking at getting triode curves from a TV sweep tube whose low screen grid voltage rating ruins the fun, I created a circuit I called UNSET. Here you apply local feedback to the control grid with a simple two resistor voltage divider. You then stand a P channel mosfet on its head to drive the cathode. Triode curves with low distortion result.
Combine that with the buffered CCS mentioned here in the same tube and you get gains in the 40 to 70 range with THD's in the 0.1 to 0.5% range and it's mostly second harmonic. The breadboard shown here uses that concept for a driver that can push a pair of TV sweep tubes to 35 watts in class A SE with a THD of 1.2%.
I'll try almost any crazy electrical idea at least once, and if it doesn't blow up, keep pushing it until it does. UNSET came about from connecting a Pch and Nch mosfet pair on all three grids, and the cathode of a small sweep tube and playing with it.
Attachments
I don't have any immediate plans for using these tubes but the recent thread about high frequency heating and the modest current requirements gave me an idea. The filament only draws 60mA at 2v. Plenty of opamps will output enough current to drive the filament directly. It would be pretty easy to make a little phase shift oscillator circuit running at 200Khz, or more interesting the subsonic frequency of 4Hz as someone had mentioned.