Given the same operating voltages and idle currents Ultralinear will allways give better tube longevity. When signal is applied, as anode current increases (on the appropriate half cycle), that is, as the anode is pulled lower in voltage then the screen voltage will drop too, thus limiting the rise in screen current.
In Pentode Mode there is no mechanism limiting screen current. You have to do that yourself by adding screen resistors. Big screen resistors give noticable audio compression which can sound OK for guitar but is NOT what you want for HiFi.
So for HiFi you want screen resistors (mounted hard up against the tube socket screen pin) to give you a grid stop function but you want only as much resistance as required
to suppress any parasitic oscillation. If you need more to protect the tube then you are operating at too high a B+ voltage and/or too high an idle current and you should fix that instead.
Of the tubes listed I have run KT88, EL34, 6V6 and EL84 in HiFi Amps and prefer Ultralinear Mode to Pentode Mode for all of those. I have not tried the others.
I've run 6L6, EL34, 6V6 and EL84 in guitar amps. Only my 6V6 amp had switchable UL/Pentode Mode and I found (for guitar) I preferred Pentode Mode.
Cheers.
Ian
In Pentode Mode there is no mechanism limiting screen current. You have to do that yourself by adding screen resistors. Big screen resistors give noticable audio compression which can sound OK for guitar but is NOT what you want for HiFi.
So for HiFi you want screen resistors (mounted hard up against the tube socket screen pin) to give you a grid stop function but you want only as much resistance as required
to suppress any parasitic oscillation. If you need more to protect the tube then you are operating at too high a B+ voltage and/or too high an idle current and you should fix that instead.
Of the tubes listed I have run KT88, EL34, 6V6 and EL84 in HiFi Amps and prefer Ultralinear Mode to Pentode Mode for all of those. I have not tried the others.
I've run 6L6, EL34, 6V6 and EL84 in guitar amps. Only my 6V6 amp had switchable UL/Pentode Mode and I found (for guitar) I preferred Pentode Mode.
Cheers.
Ian
Last edited:
How long would you expect a set (quad/pair) of new production KT88/KT90/KT120's to last in ultra-linear mode using regular capacitor coupling from a phase splitter? There is 47k ohms worth of resistance through the bias circuit from the control grid to ground. The B+ in ultra-linear is 550 volts.
As long as you have a max Rg1 of 47K for EACH TUBE then I would expect a minimum of 5 years life. Most failures I've seen have been due to thermal run away from too high an Rg1 value. That is why on all new designe I use cathode followers or mosfet source followers, one per output tube, direct coupled to the output tube grid and with the bias voltage then applied the follower grid/gate.
Getting Rg1 to a suitable low value is the most important thing you can do for tube reliability. Because the Rg1 is also a load on the Phase Inverter (unless you use a mosfet source follower or cathode follower buffer) it is also the one thing which is most often gotten wrong.
If you do a search you may find a few posts with schematic traces I did on an ARANOV LS9100 which is a quad 6550C in Ultralinear, 100 Ohm screen resistors, +530V. It had suffered a thermal run away, the usual problem of too high an Rg1 value. I selected Winged C Svetlana 6550C for match AND low grid current too try and make that more reliable .
Cheers,
Ian
Getting Rg1 to a suitable low value is the most important thing you can do for tube reliability. Because the Rg1 is also a load on the Phase Inverter (unless you use a mosfet source follower or cathode follower buffer) it is also the one thing which is most often gotten wrong.
If you do a search you may find a few posts with schematic traces I did on an ARANOV LS9100 which is a quad 6550C in Ultralinear, 100 Ohm screen resistors, +530V. It had suffered a thermal run away, the usual problem of too high an Rg1 value. I selected Winged C Svetlana 6550C for match AND low grid current too try and make that more reliable .
Cheers,
Ian
GEC 4 x kt88
Gingertube
I have made several amplifiers using the GEC multiple pair KT88 circuit using 4 Tubes. Attached.
A member on another thread was very critical of this circuit due to the Cathode resistors on the 6SN7’s being 47k he claims it should only be 15k and the 47K grid resistors for the KT88’s.Later circuits used 39K.
I wondered what values you used on your cathode followers to drive KT88’s? Do you have a circuit?
I found the output tubes 6550’s on this amp last a long time and due to the mains running high in West Australia at times the plate voltage on the 6550's was 580 volts.
Phil
Gingertube
I have made several amplifiers using the GEC multiple pair KT88 circuit using 4 Tubes. Attached.
A member on another thread was very critical of this circuit due to the Cathode resistors on the 6SN7’s being 47k he claims it should only be 15k and the 47K grid resistors for the KT88’s.Later circuits used 39K.
I wondered what values you used on your cathode followers to drive KT88’s? Do you have a circuit?
I found the output tubes 6550’s on this amp last a long time and due to the mains running high in West Australia at times the plate voltage on the 6550's was 580 volts.
Phil
Attachments
550v B+ might be too high for ultralinear operation in a KT88. I know typical 6550 wouldn't like it. In ultralinear mode the plate and the screen is pretty much getting the same voltage so when that voltage exceeds the screen rating in the tube manual, expect shortened life. In pentode operation you can always choose a lower voltage point for the screen independent of the plate voltage and that would always help tube life. High screen voltage is one of the major causes of short tube life. If you must use ultralinear then choose a voltage point that's safe for both the plate AND the screen. There's a reason many vintage commercial tube amps using EL34 or 6550/KT88, do not exceed 450V. (Marantz runs EL34 at 425v in UL.) If you must milk power then use high plate voltage and lower screen voltage, preferably regulated.
The sonic difference or preference between pentode and ultralinear is another topic.
Last edited:
If you look at old KT88 data sheets you will get the big clue.
http://www.triodeel.com/kt88p1.gif
Cathode Bias:
Max Rg1 with Pa <= 35W 470K
Max Rg1 with Pa >35W (42 W max) = 270K
Fixed Bias:
Max Rg1 with Pa <= 35W = 220K
Max Rg1 with Pa > 35W (max 42W) = 100K
So for a parallel pair of KT88 you want Rg1 to be less than 100K||100K = 50K IF you are biasing them at maximum idle power.
If you drop the idle current so that you are idling at 35W or less the max Rg1 can be 220K || 200K = 110K max.
For 5 tubes on each side of the push pull that would mean 220K/5 = 44K max if baising tubes at <=35W. 47K is a little above that so what I would be suggesting is that for a single 47K with 5 tubes then you would want to bias each tube at about 32 Watts
Guitar Amp Guys (the most abused Rg1 values known to man) sort of know this, or at least they know that they should bias their amps at not more than 70% of maximum anode dissipation rating. That is, they know to do this but probably don't understand why.
Was the arc due to thermal runaway? - YES almost certainly.
Cheers,
Ian
http://www.triodeel.com/kt88p1.gif
Cathode Bias:
Max Rg1 with Pa <= 35W 470K
Max Rg1 with Pa >35W (42 W max) = 270K
Fixed Bias:
Max Rg1 with Pa <= 35W = 220K
Max Rg1 with Pa > 35W (max 42W) = 100K
So for a parallel pair of KT88 you want Rg1 to be less than 100K||100K = 50K IF you are biasing them at maximum idle power.
If you drop the idle current so that you are idling at 35W or less the max Rg1 can be 220K || 200K = 110K max.
For 5 tubes on each side of the push pull that would mean 220K/5 = 44K max if baising tubes at <=35W. 47K is a little above that so what I would be suggesting is that for a single 47K with 5 tubes then you would want to bias each tube at about 32 Watts
Guitar Amp Guys (the most abused Rg1 values known to man) sort of know this, or at least they know that they should bias their amps at not more than 70% of maximum anode dissipation rating. That is, they know to do this but probably don't understand why.
Was the arc due to thermal runaway? - YES almost certainly.
Cheers,
Ian
Last edited:
This may be possible assuming the condition you mentioned (same voltage/idle current).
I have noticed when compared UL / pentode circuits that optimum linearity in UL circuit requires higher idle current than in pentode.
So in practice the case about longevity is not that straightforward.
Right now I am running KT90 quads in my amplifiers. The current revision has 100k ohm resistors R17, R19, R22, and R100. In the third revision I plan on swapping those for 22k. A maximum of 5K + 18k (through the bias network) + 22k yields 45k ohms. Should those net to ground equal 45k or just the R17, R19, R22, and R100?
As it stands the circuit produces very low distortion (< 0.3% at 1k at 138 watts) with a continuous sine wave into an 8 ohm resistive dummy load. Should I lower B+? KT88 screens can take up to 600 volts and KT90, higher. The objective for future revisions is to ensure long tube life to avoid breaking the bank of maintenance costs.
As it stands the circuit produces very low distortion (< 0.3% at 1k at 138 watts) with a continuous sine wave into an 8 ohm resistive dummy load. Should I lower B+? KT88 screens can take up to 600 volts and KT90, higher. The objective for future revisions is to ensure long tube life to avoid breaking the bank of maintenance costs.
Last edited:
Stu Hegeman on UL as told by Bob Carver: "As two hours went by, he admitted to having been caught up in the ultra-linear fad of his day, that it was the biggest blunder of his career, and he did it only once and would NEVER design an amp with an ultra-linear output stage ever again.
And he didn't. The Citation V was pure pentode, as was the subsequent Lafayette 550 and everything else he designed from then on. I asked how come. He explained that when the plate pulls the top of the output transformer winding towards ground, the ultra-linear tap pushes the screen grid so low that it renders the tube unable to drive difficult loads. In addition, he pointed out that the normal idle potential on the screen grid regularly exceeded a safe voltage, often causing output tubes to blow up. The tube manufacturers hated it, but had to go along or lose market share, and so changed the specification for screen voltage in order to allow ultra-linear output stages. I wonder if they changed the tube design, or just changed the screen voltage specification."
More here
And he didn't. The Citation V was pure pentode, as was the subsequent Lafayette 550 and everything else he designed from then on. I asked how come. He explained that when the plate pulls the top of the output transformer winding towards ground, the ultra-linear tap pushes the screen grid so low that it renders the tube unable to drive difficult loads. In addition, he pointed out that the normal idle potential on the screen grid regularly exceeded a safe voltage, often causing output tubes to blow up. The tube manufacturers hated it, but had to go along or lose market share, and so changed the specification for screen voltage in order to allow ultra-linear output stages. I wonder if they changed the tube design, or just changed the screen voltage specification."
More here
I've been leaning toward Pentode and Triode more than UL. For sweep tubes, screen is limited to about 200V, mostly Pentode Mode; for 6550, and KT88's Triode mode but not at excessive B+ I have UL-Triode switches, but leave it on Triode mode....
Recommend MosFet drivers on the grids, very little extra components...
Recommend MosFet drivers on the grids, very little extra components...
Well, have a look to early documents about UL ("an Ultralinear Amplifier",Haefler and Keroes 1951). As a starting point you will see separate windings for screen and plate with 2 separate supplies for screen and plate. To save costs and to simplify the OT, in a second step they propose to merge the 2 windings into one winding with a tap for the screen.
I have built a 200W UL PP Evaluation power stage with KT120 (~600V Screen supply , 800V plate supply), has worked fine. Here http://www.diyaudio.com/forums/tubes-valves/239572-about-gu50-pp-ul-separate-g2-winding-supply.html I have described shortly how this can be done with a GU50 including some results.
HG
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.