I don't have the uTracer, the graph I posted was just a link to Dr. Dekker's site. But you should not run a trace with the screen voltage higher than the plate voltage. In fact, to verify your settings, you should start by duplicating a Ep-Ip characteristic from a datasheet. The traced curves should match that of the datasheet closely.
Do you have a plate current limiting resistor enabled in uTracer (or added externally), that way, there should be no problem with starting the trace at "0V". I am sure you will get this sorted out soon...
It's true that you don want an ordinary tube circuit have a screen significantly higher then the anode.
But the uTracer is pulsed, 1ms measuring and then pause, thus the
overheating is not really a problem.
It does however gives an opportunity to measure regimes that only
abnormal loads will create.
May have missed my earlier posting by time overlap:
Jazbo8's illustration is quite nice. However, something is fishy there too. The pentode knee starts at 160 mA (k = 0), but at k = 0.4 it has dropped to 80 mA. This would imply only 25% as much power output for 40% UL.
Is the uTracer measuring screen grid current and then combining that, at k effectiveness, with the plate current in software? Ie, display Ip + k*Iscrn
Jazbo8's illustration is quite nice. However, something is fishy there too. The pentode knee starts at 160 mA (k = 0), but at k = 0.4 it has dropped to 80 mA. This would imply only 25% as much power output for 40% UL.
Is the uTracer measuring screen grid current and then combining that, at k effectiveness, with the plate current in software? Ie, display Ip + k*Iscrn
Last edited:
I use a 120 Ohm stopper resistor + a few ferrite beads on the g2 lead. And, 1000 Ohm stopper + a few ferrite beads on the g1 lead, for the Tekronix tracer here, to stop oscillations. Even works for frame grid tubes.
But the curves Jazbo8 posted have a different problem. The knee currents are too low. This looks like the screen grid current is not getting combined (at k effectiveness) with the plate current for UL mode.
But the curves Jazbo8 posted have a different problem. The knee currents are too low. This looks like the screen grid current is not getting combined (at k effectiveness) with the plate current for UL mode.
I don't think so, it is showing Ia only, Ik is actually higher, so the output power is a bit higher than the 25% estimate. But in order to have a similar Po as the pentode, the plate voltage and the load impedance need to be altered, so the operating conditions for the UL connection would not be the same as the pentode.
So then we can not tell linearity for UL mode from the uTracer curves....
I think I will have to rig something up for the Tek tracer here to find out. It has separate input jacks for external Vert. or Horiz. display. Using a current probe, I could combine Ip + k Ig2 by turns ratio thru the probe. Not much room in the current probe gap for multi turns with insulation though.
Ohh, except how to set Vg2 at some fraction of B+ - Vp? That's where I got hung up before. Have to use an attenuated Mosfet driver for Vg2. Too much trouble for UL.
The uTracer needs to provide a software display option for UL mode.
I think I will have to rig something up for the Tek tracer here to find out. It has separate input jacks for external Vert. or Horiz. display. Using a current probe, I could combine Ip + k Ig2 by turns ratio thru the probe. Not much room in the current probe gap for multi turns with insulation though.
Ohh, except how to set Vg2 at some fraction of B+ - Vp? That's where I got hung up before. Have to use an attenuated Mosfet driver for Vg2. Too much trouble for UL.
The uTracer needs to provide a software display option for UL mode.
Last edited:
Not sure if it worth going through the trouble, since Langford-Smith already did the experiments back in the '50's. Also since Ig2 tracks Ia, adding it basically shifts the curve family to the left a little bit. Using Ia should be close enough to estiamte the THD in any case...
Ig2 only tracks Ip if Vg2 is kept at a constant % of Vp. But UL does not do that. It uses B+ - Vp. Very different. Ig2 is quite spiky in UL, since Vp descends well below Vg2. You can not infer linearity from Ip alone.
Going swiftly through the 31 pages of this thread I see nobody actually voicing an opinion about SOUND and how one likes UL or not.
It's all technical mumbojumbo without a single subjective listening preference.
Well here's one for future reference:
I LOVE how Single-Ended UL sounds compared to my SS amps, it has dimension and balls, more upfront and "in the room" presence sound.
It's all technical mumbojumbo without a single subjective listening preference.
Well here's one for future reference:
I LOVE how Single-Ended UL sounds compared to my SS amps, it has dimension and balls, more upfront and "in the room" presence sound.
Last edited:
well sadly my UL amp has been out of commission for the last 18 years, but when it ran i loved it.
Ig2 does get spiky when Vp gets low, but when Ia and Ig2 are added together, the resultant characteristic still resembles that of the Ia alone, doesn't it?
"Ig2 does get spiky when Vp gets low, but when Ia and Ig2 are added together, the resultant characteristic still resembles that of the Ia alone, doesn't it?"
Check page 6 for 6L6GC:
http://frank.pocnet.net/sheets/093/6/6L6GC.pdf
They don't look very similar to me. I think if you put a small SE OT in the output tube UL screen currents, to listen on headphones to the screen current, you won't want it added back in.
For triode mode they are similar, since they have the same voltages then. UL just turns nice kinkless pentodes into tetrodes. I suppose, ideally, the sum would give a flat top plate curve like a Mosfet.
Check page 6 for 6L6GC:
http://frank.pocnet.net/sheets/093/6/6L6GC.pdf
They don't look very similar to me. I think if you put a small SE OT in the output tube UL screen currents, to listen on headphones to the screen current, you won't want it added back in.
For triode mode they are similar, since they have the same voltages then. UL just turns nice kinkless pentodes into tetrodes. I suppose, ideally, the sum would give a flat top plate curve like a Mosfet.
Last edited:
Not sure what I am suppose to look at... Perhaps I did not phrase the question properly, I meant to ask - shouldn't the composite Ik = Ia + Ig2 characteristic, which is the current seen by the OPT be similar to the Ia characteristic, which is the one shown by uTracer? Are you expecting them to be completely different and that's why you think that we can not judge the linearity of the UL connection looking at just the Ia characteristic... 🙂
Well, on page 6, the Ig2 curve is completely different shape to the Ia curve. I don't think that could sum to the same Ia (scaled) curve. But they might sum to a flat top curve like a Mosfet.
Also, the OT does not see Ik, but rather Ia + k x Ig2 due to the tapping.
An interesting example are the dual control tubes like 9KC6 or 6LE8. One can adjust the g3 voltage to minimise g2 current, and the plate curves square right up like a Mosfet.
(9KC6 and 6LE8 are very much under-appreciated tubes by the way, one can tune their plate curves to the shape you like. And they are on the $1 list besides. 9KC6 is a 12GN7 with a dense g3, putting +10V to +13V on g3 turns it into a square knee/flat top Mosfet effectively )
Also, the OT does not see Ik, but rather Ia + k x Ig2 due to the tapping.
An interesting example are the dual control tubes like 9KC6 or 6LE8. One can adjust the g3 voltage to minimise g2 current, and the plate curves square right up like a Mosfet.
(9KC6 and 6LE8 are very much under-appreciated tubes by the way, one can tune their plate curves to the shape you like. And they are on the $1 list besides. 9KC6 is a 12GN7 with a dense g3, putting +10V to +13V on g3 turns it into a square knee/flat top Mosfet effectively )
Last edited:
True, but sometimes the answer may gotten through measurements:
Back to the uTracer, i have looked at a number of possibilities; what is consistent is that the step in Ia is between 2 data points, regardless of the data point spacing; what my current thinking is that there may well be some oscillation in the tube (that is either on or off), that gives a step response. So my thinking is the tube is more susceptible to oscillation when Va is low and Vs is high (this needs verification) This may explain why the lower Va start range setting influences the appearance of the step; if Va is not low enough no oscillation will start.
Back to James' point; if this is correct (and I will put a probe onto the system) then there is no reason why this may not happen under real audio conditions, but it appears somewhat random and only at larger signals, but will possibly lead to audio manifestations. Also "equivalent tube" types such as the 6CA7's behave quite differently to EL34's I have checked, so there are many variables that influence this phenomena.
Yup, you caught it - the OT does not see Ik, but rather Ia + k x Ig2. So we get this, close enough?🙂
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
This is a calculated summation then? It is looking quite OK.
So the horrific (earlier) kinks are gone now. We would likely have to resort to FFT analysis to make any detailed decisions on distortion with such clean curves now.
The drop in the "knee" of the curves with increasing %UL will require a higher primary Z OT to get similar power out compared to the pentode case. So switches that flip from pentode, to UL, or triode mode in the same amp likely give misleading comparisons sound-wise.
So the horrific (earlier) kinks are gone now. We would likely have to resort to FFT analysis to make any detailed decisions on distortion with such clean curves now.
The drop in the "knee" of the curves with increasing %UL will require a higher primary Z OT to get similar power out compared to the pentode case. So switches that flip from pentode, to UL, or triode mode in the same amp likely give misleading comparisons sound-wise.
Last edited:
Yes, they are calculated with Ia+Ig2 x k. We still do not know why there are kinks in uTracer graph posted earlier, I think it is just a software or setup glitch. The distortion with the UL connection is low but Langford-Smith did note the distortion characteristic under overload condition is not as good as pentode or triode. And Hegeman also pointed out that when overdriven, the UL amplifier failed miserably - precisely due to the spikes that you mentioned earlier, i.e., Ig2 shoots through the roof, the basically craps out if not destroying the tubes in the process. That is one reason that you seldom see UL output stages in guitar amps, they would be blown to smithereens...
By overload condition, you mean lowish Zload?
I would have thought that low Zload issues would be changing monotonically across P - UL - T rather than U shaped.
Well, lets see, triode certainly doesn't like too low a Zload. But if you vary Zload, the load line already being almost perp. to the triode lines should have a minimized variation versus grid V. While the load line for a pentode is close to parallel with the plate curves, so one would expect to see more variation with change in load line slope.
Seems like UL might actually be best under that analysis, since one could get the curves exactly perp. to the load line. But that analysis is ignoring the constant Mu effect for constant current. One could consider all the cases as just triodes with different Mu factors.
Maybe a better analysis just sees the pentode end of the scale as a current source, and the triode end as a voltage source. UL falling in between as an impedance. A matched impedance would seem least sensitive to load variation (versus power out). But maybe the Z matching condition is too extreme for baseline distortion. They all want high Zload for least initial distortion. In which case (nornally high Zload) the triode would be least sensitive. (its Rp largely shunting out the Zload variation)
I would have thought that low Zload issues would be changing monotonically across P - UL - T rather than U shaped.
Well, lets see, triode certainly doesn't like too low a Zload. But if you vary Zload, the load line already being almost perp. to the triode lines should have a minimized variation versus grid V. While the load line for a pentode is close to parallel with the plate curves, so one would expect to see more variation with change in load line slope.
Seems like UL might actually be best under that analysis, since one could get the curves exactly perp. to the load line. But that analysis is ignoring the constant Mu effect for constant current. One could consider all the cases as just triodes with different Mu factors.
Maybe a better analysis just sees the pentode end of the scale as a current source, and the triode end as a voltage source. UL falling in between as an impedance. A matched impedance would seem least sensitive to load variation (versus power out). But maybe the Z matching condition is too extreme for baseline distortion. They all want high Zload for least initial distortion. In which case (nornally high Zload) the triode would be least sensitive. (its Rp largely shunting out the Zload variation)
Last edited:
- Status
- Not open for further replies.