So...is anyone using the KT120's yet?

[es345]

Well, that depends at which condition you are measuring:

GU50 values (from Kytelabs web page )
Ua=800V, G2=250V, Ug1= -40V, Ia=50mA >>S=4,5mA/V

I have now done a quick measurement at my Powerstage with KT120

Ua=826V, Ug=621V,
Ug1=-85V, Ia(~95%Ik)=49mA
Ug1=-87V, Ia(~95%Ik)=41mA
>> resulting transconductance S=4 mA/V

the biggest difference (electrically) between the 2 tubes which I see:

GU50: max Pdiss at anode 40W; KT120: max Pdiss at anode 60W;
GU50: max Pdiss at G2 5W; KT120: max Pdiss at G2 8W;
GU50: max Voltage at G2 250V; KT120: max Voltage at G2 600(650) V

[kg]

yeah unfortunately because the kt120 is a "new design" tube there aren't any standards for it. so who actually knows what the specs are??

[es345]

... there is only that little spec mentioned somewhere above. So I helped myself looking to the max ratings of Ik,Pa and Ua, Pg2 and Ug2, put in some margin, calculated the respective OT, built a trial engine and startet bias adjustment with -100V...

I would call it "design by estimations" , but the results are not bad.

[imaradiostar]

A little bump- curious if anyone else has done anything with these monsters. I still have two very large output transformers begging to be used for something fun.

jamie

[jmaf]

A little bump- curious if anyone else has done anything with these monsters. I still have two very large output transformers begging to be used for something fun.

jamie

It seems like driving them is the whole issue. I'm also curious to see all-tube designs based on the KT120, as I've not done any experiments with them yet. In this thread there's an interesting Mosfet driver exemplified(and tested, according to the author), but I'd really like to see an all-tube kt120 amp. Perhaps later in the year I'll have time to experiment with them.

[imaradiostar]

I don't understand why grid drive is such a big deal. Just copy the UL Marshall Major output with a concertina driving 12AU7's or 6SN7's into the KT120 grids. No need for cathode follower drive- just big film caps driving 47k grid leak resistors.

The other option, of course, is to purchase one of the inexpensive Hammond or Triode Electronics phase splitter transformers and drive it with your pentode of choice. With multiple secondaries this would allow you to bias individual tubes and direct couple to the grid stopper resistors with no caps to charge up and grid block on transients.

jamie

[jmaf]

I thought about driving it with a transformer, using a el84 on the primary. The KT120 datasheet looks like it was photoshopped together from a bunch of pieces. It's a new tube yet the datasheet looks like it was made in 1930....so the only way to find out is to build these things and test. I wish I had the time.

[Gingertube]

Some notes on your posted schematic.

The Source Follower load resistors (100K) need to be reduced. I would aim for say 39K. This allows you to add grid stops of up to 10K for the output tubes and still meet the 50K max Rg1 specification.
I haven't looked up the data sheet but you did say that those output tubes are high gm - that means that grid stops will almost certainly be required to prevent parasitic oscillation.

The MOSFET source follower drain connection supply needs a GOOD quality filter cap added (non shown on your schematic). I use this schematic arrangement a lot in HiFi Tube Amps and the quality of that cap is quite noticable.
The supply voltage to the drains does not need to be high. If it is too high it just results in extra dissipation in the MOSFETs.
So how high should it be????
The problem with MOSFETs sounding good when used this way is the change in reverse capacitance with the audio signal as the Drain to Source Voltages changes - Drain is fixed, source follows the audio signal. That change in the reverse capacitance levels off and becomes negligable (or at least reasonable constant) when you make sure that there is at least 25 to 30 Volts left across drain to source on the most positive going peak of the audio signal on the mossfet source. As that point can swing up to 0V (or maybe a few volts over) then the drain voltage should be say +30 Volts minimum. More is OK but as stated above more just means more power dissipated in the MOSFETs. Aim for perhaps 40 to 50V

Not sure why you went to the sophistication of an SRPP input stage. Nothing wrong with it - it will be very clean BUT probably not required.

With the MOSFET source followers the splitter does not need to be grunty (the mosfets are a light load) - a 12AT7 (ECC81) will do the job and probably sound better than the 12AU7 (ECC82). The higher gain will result in better power supply rejection. This would also allow the use of more equal anode loads on the diff amp splitter, just use 100K for both.

AND finally,
The splitter (any differential amp stage actually) will sound better simply because the load on it (the MOSFET source followers) always remains in Class A and so the differential load remains more constant. Now that is a HIFi'ish consideration but probably is giving you benefits for a Bass Amp too.

Edit:
One last thing - the Mosfet protection Zeners need to be directly from gate to source NOT as they are currently drawn in the schematic.
In addition the MOSFETs will possibly need gate stoppers just like the output tube need grid stoppers. I would put in 1K stoppers in the MOSFET gates (protection zener connection right on the gate not on the othetside of the gate stopper resistors). If the MOSFETs are mounted on heatsinks remote from the rest of the circuit then the gate stoppers and the protection zeners should be right at the MOSFET not at the end of any wires connecting to the rest of the circuit..

Hope there is something of value in this for you.

Cheers,
Ian

[es345]

Ian,

thanks for your feedback.

indeed I changed already several things at that evaluation platform. Some comments to your proposal:

- Source follower load resistor: can be changed, but is it necessary? The resistance of the source follower is much lower and is in parallel from AC point of view. Feedback would be great.
- grid stoppers are now in main reason to put them in was to limit the cathode current in overdrive: without them the peak curent was going up to 1A - which will kill the tube over the time
- power supply has good enough filtering - no hum observed
- supply voltage of the source follower is reduced to ~100V
- the connection of the Zener was a simple drawing mistake, should have sent an updated version
- I will test the gridstoppers at the mosfet gates

Then the biggest change - my latest evaluation:

I change all tubes from triode to pentode (EF86) - no longer SRPP - to get higher gain without more stages
Result: now I can drive it directly with my bass up to overdrive.

I have attached the current status including the power supply. The power supply is multirail to be ableto serve differnt configurations.

Thanks again

Hans- Georg

[Gingertube]

Hans,
On the source follower load resistors.
You are quite correct that the AC Impedance from output tube grids to 0V will be low enough due to the low output impedance of the follower.
My understanding of positive and negative grid currents in output tubes is not expert BUT I believe that to prevent thermal run away from bias perturbations from negative grid current then the DC impedance needs to also meet that 50K Ohm maximum specification.
In my HiFi Amps I run "Ring of Two" transistor current sources rather than a resistor load and these amps have the "blackest" background (lowest noise) and best fine detail of any amp I've ever heard. Not a huge consideration for a Musical Instrument Amp.
My ideas are unfortunately "corrupted" by my HiFi Amp design experience but every amp design change which took more effective control of the output tube grids by providing lower DC resistance and AC impedance back to 0V was sonically better.

I don't run this scheme in my guitar amp as I only use a pair of 6V6 outputs and for those at least, I ditched the diff amp splitter and went back to as concertina, I just liked it better - a caveat on that, I use a concertina splitter using a high mu tube, in my case a 6SL7. The high mu in the concertina tends to give the 2nd and even harmonic distortion overload characteristic I wanted for what is a "Rock and Blues" Amp. A "shreader" metal amp might have been different.

Off for the next 4 days - will check back in next week.

Cheers,
Ian

[kg]

Hans,
On the source follower load resistors.
You are quite correct that the AC Impedance from output tube grids to 0V will be low enough due to the low output impedance of the follower.

agreed.

My understanding of positive and negative grid currents in output tubes is not expert BUT I believe that to prevent thermal run away from bias perturbations from negative grid current then the DC impedance needs to also meet that 50K Ohm maximum specification.

that's actually the ONLY thing that needs to be monitored: the DCR.

My ideas are unfortunately "corrupted" by my HiFi Amp design experience but every amp design change which took more effective control of the output tube grids by providing lower DC resistance and AC impedance back to 0V was sonically better.

agreed, ime as well.

as far as the DCR of the output grids, one thing we must insure is that the "top" of the source follower is connected to a low Z supply. low Z implies not only low reactance but low DCR as well.

let's trace the path of an electron which appears at the output grid back to its own cathode.. it would pass through:

the 10k stopper
the buz50a (looking at datasheets if the buz50 is biased at say 50mA the gm is still very high, in the 100mmho range, netting an effective Z of ~10 ohms)
5x reverse biased 20v zeners (the zeners should be well into the avalanche region and of low impedance)
a forward biased diode drop
a 100vac transformer winding (i don't know the dcr of the tranny winding, but i'm sure it's in the 10s of ohms)
another forward biased diode drop
chassis ground
a 10r cathode resistor in the output stage

so adding up all the impedance between g1 and k, we are still well under the design max grid return values. in fact the g1 stopper dominates as the largest factor, by at least a factor of 10 if not 100.

to address your concern about the source follower load resistor... as the current loop described above illustrates, the value of that resistor never enters the equation at all! we've got what is often described as an "active pull up" driver.

BUT (of course there's a "but")

that loop describes what happens during a positive grid current scenario: conventional current flow INTO the grid of the tube, electrons coming OUT of the grid. this occurs on signal peaks, when the grid is pulled high. this is what gives you greater power output capacity, fast recovery from grid current, lack of grid blocking distortion, etc.

what happens when the reverse is true? a negative grid current scenario?

it's not the same loop!

what we'll have is a situation in which electrons are being sucked INTO the grid pin. those electrons have to come from somewhere... the only place they can be sourced is (as you correctly pointed out) THROUGH the source resistors.

10k stopper
100k Rs
diode drop
full wave doubler
diode drop
ground

that's a LOT less attractive than the positive peak case... no matter what, the ability of the circuit to supply electrons to the output grids will be reduced by at LEAST 110k (and that's a best case scenario). this is due to the "passive pull down" nature of the resistive load. the loops are not symmetrical, and neither is the ability to effectively sink and source current.

to make matters worse, in a runaway scenario this is exactly the failure mode that occurs--high ionization current in the tube (from age, overcurrent operation, outgassing, poor initial vacuum hardness, etc) causes positive ions to be attracted to the most negative element available: almost invariably the negatively biased control grid. the ions impact and "steal" electrons from the grid, making it more positive, which then leads to increased cathode current, which leads to higher ionization, and so on until red plates (or worse).

so you have every right to be concerned about THIS current loop, arguable the more important one in terms of longevity and stability!

what can we do?

well, as ian suggests, reduce the Rs. that is the first thing.

second, reconsider the negative rail feeding those source resistors, as the impedance of that rail is in series with them. i suggest trying to make a circuit WITHOUT a doubler. set your negative rail value according to your desired source follower current value, along with the (lower) Rs value. you have some idea what you want the top of that resistor to be, voltage wise, since that's the output tube bias. as far as source follower currents, within reason, pulling more current through the followers will be a "good thing." i'd shoot for at least 50mA. juggle your V, I, and R around until the values make sense. it may be helpful to refer back to my earlier posted sketch of a very low Z negative supply.

third, in ian's design he uses a current sink which creates an "active pull down," continually pumping electrons into the source whether they go through the upper device OR end up in the tube fighting the ionization current. this is a "better," if more complicated, solution, and is very similar to the one i employed in the BAGA driver (pentode connected el84 follower pullups, 15mA constant current sink pulldowns).

hth
ken

[es345]

Ian, Ken,

thanks for your explanations and suggestions.

I just have changed as a first step the source resistor to 27K (now ~7 mA source current).
Up to now I haven't recognized a change in behaviour, but I am just starting to test and measure.

Ken,

regarding the negative rail, just to align the understanding:

Bias is about -90V, so the negative signal peak without overdrive is somewhat about -180V.
To secure stable operating of the MOSFET I need more than -200V.
Taking into account the voltages at that NT I had to do voltage doubling. I will need more time to put in the stabiliziation circuit.
In addition I will measure how much the voltage is changing between no drive and overload.

cheers

Hans- Georg

[es345]

... well, have done some more tests after changing the source resistor to 27K.

the first topic is to observe the behaviour when the grid is starting emission.
The setup is easy, it is described in the attachment. basically after pulling out the power tubes I have emulated a grid emission of about 3 mA. Interesting to see that the bias is changing not more than 0,1V.

The reasoning:
- assume no grid emission. Then about 7 mA is flowing through the source resistor.
- assume grid emission of 3 mA. Then 4 mA are still going through the MOSFET.
If you look to the embedded data sheet you can see that 500mA change in current correlates to 0,5 V change in gate source Voltage. Consequently the voltage change resulting from a current change from 7 mA to 4mA is very small

Conclusion:
as long as some current is flowing through the MOSFET, the situation does not differ substantially from the case of grid conductance.

2nd topic was to measure the grid emission:
I have measured after 20 minutes full load - with and without load , both tubes. The maximum measured grid emission was 1.65 uA.
Would be interesting to know
- if someone has done more measurements about that topic?
- if someone has experience with worst case measurements, when a tube is starting to die?

3rd topic was to measure the change of the negative rail voltage between idle and overdrive up to rectangle
you can see the results in the attachment. Conclusion for me is that I will stabilize the bias even that it seems not too critical.

Conclusion:

Looking to the excel you see that the minimum current through the source - this happens when the signal has its negative peak -

- is about 960 times higher than the highest measured grid emission current with a source resistor of 27K
- is about 260 times higher than the highest measured grid emission current with a source resistor of 100K

Both cases do not differ regarding the behavior of the MOSFET. Therefor I could not observe a change behaviour.

Change to 27K increases the margin - done -
Stabilizing the Bias - still to be done -

Cheers

[imaradiostar]

Which begs the question- why not just use a multi-winding transformer, a low-Z bias supply and a bias adjustment per transformer secondary?

jamie

[es345]

... for sure that will work. With a high quality transformer you will reach comparable results.

It's just the question what you prefer. From my experience the MOSFET source follower solution works well and it might be the cheaper one.

Hans- Georg