Since you are DC coupled to the previous stage, the grids of the LTP are elevated a hundred volts or so above circuit common; therefore, grounding the CCS is just fine (all you typically need is between 5 and 20V across the CCS, more is slightly better to lower capacitance).
If the prior stage is AC coupled through a capacitor to the LTP, the grids are then at circuit common, and you will need the negative supply for the CCS.
A power transformer question... I would [currently] prefer to build this as a dual mono package (shared power transformer and chassis, everything else separate).
Would this Edcor transformer do the trick?
XPWR143
720V(360-0-70-360)@500mA, 6.3V(3.15-0-3.15)@5A, & 6.3V(3.15-0-3.15)V@5A
Thanks,
..Todd
Would this Edcor transformer do the trick?
XPWR143
720V(360-0-70-360)@500mA, 6.3V(3.15-0-3.15)@5A, & 6.3V(3.15-0-3.15)V@5A
Thanks,
..Todd
Excellent Taj! Thanks.
I'll make an attempt at R13.......Since we want 8ma through the 6GK5 (see above), we need about 370 ohms or so for R13. So either a fixed 390R or a 220R and a 500R trim pot wired in series same as you are showing below the 12AT7.
Also, you may want to show R11 &R12 horizontally right next to the component since they are grid stoppers, and (optionally) call out cc (carbon comp).
That transformer that you've spec'd looks good, although the cap and R values will change since you are pulling 2X the current through the PS. Also, you'll have to 1/2 wave rectify the bias since it shares the CT with the B+. I can model in PSUD if you'd like.
Last edited:
R19 connects to the 6GK5's anode, not pin 2 of the 'T7 socket.
Verify that the net KT88 grid to ground resistance does not exceed the limit contained in the data sheet. This is a "fixed" bias design and liberties must not be taken.
Peruse the 10M45S data sheet to determine the correct value for R13.
Taj: For some hands-on training, download and run PSUDII from here:
PSUD2
Take the existing PS model, and double the currents on the current taps, see if the B+ voltages and ripple voltage and current are reasonable. Also, you'll need a choke with lots of current capability, which puts you back to the enclosed Hammond 193 series.
Ask lots of questions if needed.
I'll also model the stereo version for fun.
Also, it looks like the max grid R is 100K for KT88 fixed bias with power >35W.
Does this need a pot there? Does the one on the LTP need one, or are fixed resistors okay.
Sure. Also, the filament windings have a CT. Can we use those to provide a bias, and lose the extra transformer for a stereo configuration?
..Todd
Stereo Power Supply
I just modeled the stereo power supply, and assuming that I have the current draw for each stage correct, it "barely makes it" with a 5H/26R Hammond 193P choke and the Edcor XPR143.
This choke appears to be the lowest R at a reasonable current rating for this PS.
Doubling the current makes the resistance of the choke the limiting factor at 360V supply; it barely makes a B+ of 463V when using 4 X 75ma for the KT88's. The PSUD model shows 378V as the unloaded voltage when you input 360V/500ma into the transformer properties box, assuming 5% regulation.
So there is no wiggle room for increasing the voltage even a little. It would be nice if Edcor had a 380-0-380 or so. If you want to run the KT88's hotter than 75ma, then B+ will reduce even more.
I used 75ma X 4 for the KT88's 6ma X 2 for the 12AT7's, and 8ma x 2 for the 6GK5's.
I just modeled the stereo power supply, and assuming that I have the current draw for each stage correct, it "barely makes it" with a 5H/26R Hammond 193P choke and the Edcor XPR143.
This choke appears to be the lowest R at a reasonable current rating for this PS.
Doubling the current makes the resistance of the choke the limiting factor at 360V supply; it barely makes a B+ of 463V when using 4 X 75ma for the KT88's. The PSUD model shows 378V as the unloaded voltage when you input 360V/500ma into the transformer properties box, assuming 5% regulation.
So there is no wiggle room for increasing the voltage even a little. It would be nice if Edcor had a 380-0-380 or so. If you want to run the KT88's hotter than 75ma, then B+ will reduce even more.
I used 75ma X 4 for the KT88's 6ma X 2 for the 12AT7's, and 8ma x 2 for the 6GK5's.
Attachments
This is soooo nutsy. There is no issue at all to running KT88's at 400V. There is a barely perceptible decrease in power( at the power-measuring resistor ), and effectively nothing as far as the ears are concerned.
cheers,
Douglas
Taj,
Your schematic looks good, though I have found a slightly different working point for the 12AT7. But I did not notice where you started from, i.e. the 6GK5's anode voltage - no problem.
I give the following, sticking close to the original EF86 - 12AT7 - KT88's idea. Also, I have not included the 10M45S CCS simply because setting up working points and basic limiting measurements do not require it, and there is not one ready nearby. I have no problem with it; in my experiment the 'tail' current required was 4,8mA.
But first! My face was RED again. I have to divert slightly from previous given values. Not an excuse, but at least an explanation: The #$@&*^% data sheets for EF86 differ from source to source. (I even found a double diode-pentode one for an EF86; the EF86 is only a pentode.) So copying a few I rested with the Mullard sheets - seven of them. Print not altogether sharp but useful. Then ... I also ran through several EF86 examples to find one conforming with the above data. The one I previously used did not quite! That my bad for not testing first. Not to discourage users; at least with an established circuit and readings one can determine that. I would say that deviation of some +/-5% would not upset matters. (Others on this?)
So, referring to the last circuit in my possession (Draft - Rev. 0.5, 30 January 2010 by TAJ), the following: (And after finishing this it was quite long, so I took out the steps followed and only left results. I imagine details might be valuable for 'freshmen", a more complete description might follow as a tutorial.)
Firstly the gain: As mentioned before the prime element determining gain is R15. (Making R16 = 0 for now; see later). It can be either bypassed or unbypassed. Cutting out detail, I found that unbypassed an input sensitivity of 38mVp is required, and bypassed some 40% of that only. To my notion, with say 20dB of global NFB, that gives a basic sensitivity of 380mVp. I would regard that as sufficient (it could even be less sensitive), so I stuck with R15 unbypassed.
One can now make a simplification. It is unnecessary to have R16 as a low value; one can just as well eliminate it and connect NFB directly to the EF86 cathode, i.e.top of R15. So, R17//C13 go directly to pin 3 of the EF86, with R15 to ground; no R16. We will come back here.
I cheated somewhat; I do not have KT88s for testing! But from data and experience I took the maximum Vg1 signal to be 50Vp; in fact that appears to give >60W output, but there is still the output transformer (OPT) efficiency. Thus, first starting with 120Vdc on the 12AT7 cathodes, R21 = R22 = 56K and next stage G1 resistors 100K, and adjusting the 12AT7 current, there is some difficulty for it to give a clean 50Vp (100Vpp) on both anodes. Overload (peak rounding = 3rd harmonic distortion) visibly (scope) sets in at about 110Vpp.
Taking Vcc(12AT7) = 100V (see below), one can get a good 70Vp per anode, with which I stuck. This was achieved with 100Vc on the 12AT7 cathodes, at a tail current of 4,8mA, leaving some 323Vdc on each 12AT7 anode. That is what the CCS current will have to be set at; easiest is to set it so that the specified Va appears. (The equivalent tail resistance used in my set-up was 22K.)
[Referring to previous discussion, both calculation and measurement (as well as the scope trace can be detected) shows an unbalance of 8% on the 12AT7 anodes without further ado - i.e. with the resistor of 22K instead of the CCS. This is almost good; an extra resistor of 4,3K in series with R22 would give balance. This is now regarding the use of a resistor for tail.]
One still needs to get Va(EF86) to just below 100V (bias for the 12AT6 is about 3V) to set the 12AT7 cathodes. I found that possible with R13 = 100K; R14 = 560K; R15 = 560 ohm. Finally calculating for global NFB of about 20dB, R17 = 47K.
A hint: First build: Leave R17 off and omit C12, cutting gain right down. Dc voltages can then be checked and set without fear of instability. When that is done, reconnect C12, which should give a slight hiss at the loudspeaker. One can bypass the input by a say 10K resistor; that could be normal for the output impedance of a pre-amp; probably less.
R17. Having the OPT output the wrong way round when connecting NFB is not a good idea (meaning very possible damage somewhere from the very loud squeal that would result). First try for polarity with R17 at least the feedback factor larger - say 1 megohm. The amplifier should not oscillate, but a definite increase/decrease in output noise will be noticed. A decrease means negative feedback, which indicates the way in which the transformer secondary should be wired.
Time is getting on and the post may exceed the alotted length, so until further. Setting NFB stability is next, also frequency response - that involves R12, C11, C13
IMPORTANT THOUGH: These values are given as a 'chopping block'. I would like experienced members to kindly check. Also, many members here have 'been there, done that'. The way forward is to peruse and come up with discussion, please! I think I know my way round tubes - but so do many others!
Your schematic looks good, though I have found a slightly different working point for the 12AT7. But I did not notice where you started from, i.e. the 6GK5's anode voltage - no problem.
I give the following, sticking close to the original EF86 - 12AT7 - KT88's idea. Also, I have not included the 10M45S CCS simply because setting up working points and basic limiting measurements do not require it, and there is not one ready nearby. I have no problem with it; in my experiment the 'tail' current required was 4,8mA.
But first! My face was RED again. I have to divert slightly from previous given values. Not an excuse, but at least an explanation: The #$@&*^% data sheets for EF86 differ from source to source. (I even found a double diode-pentode one for an EF86; the EF86 is only a pentode.) So copying a few I rested with the Mullard sheets - seven of them. Print not altogether sharp but useful. Then ... I also ran through several EF86 examples to find one conforming with the above data. The one I previously used did not quite! That my bad for not testing first. Not to discourage users; at least with an established circuit and readings one can determine that. I would say that deviation of some +/-5% would not upset matters. (Others on this?)
So, referring to the last circuit in my possession (Draft - Rev. 0.5, 30 January 2010 by TAJ), the following: (And after finishing this it was quite long, so I took out the steps followed and only left results. I imagine details might be valuable for 'freshmen", a more complete description might follow as a tutorial.)
Firstly the gain: As mentioned before the prime element determining gain is R15. (Making R16 = 0 for now; see later). It can be either bypassed or unbypassed. Cutting out detail, I found that unbypassed an input sensitivity of 38mVp is required, and bypassed some 40% of that only. To my notion, with say 20dB of global NFB, that gives a basic sensitivity of 380mVp. I would regard that as sufficient (it could even be less sensitive), so I stuck with R15 unbypassed.
One can now make a simplification. It is unnecessary to have R16 as a low value; one can just as well eliminate it and connect NFB directly to the EF86 cathode, i.e.top of R15. So, R17//C13 go directly to pin 3 of the EF86, with R15 to ground; no R16. We will come back here.
I cheated somewhat; I do not have KT88s for testing! But from data and experience I took the maximum Vg1 signal to be 50Vp; in fact that appears to give >60W output, but there is still the output transformer (OPT) efficiency. Thus, first starting with 120Vdc on the 12AT7 cathodes, R21 = R22 = 56K and next stage G1 resistors 100K, and adjusting the 12AT7 current, there is some difficulty for it to give a clean 50Vp (100Vpp) on both anodes. Overload (peak rounding = 3rd harmonic distortion) visibly (scope) sets in at about 110Vpp.
Taking Vcc(12AT7) = 100V (see below), one can get a good 70Vp per anode, with which I stuck. This was achieved with 100Vc on the 12AT7 cathodes, at a tail current of 4,8mA, leaving some 323Vdc on each 12AT7 anode. That is what the CCS current will have to be set at; easiest is to set it so that the specified Va appears. (The equivalent tail resistance used in my set-up was 22K.)
[Referring to previous discussion, both calculation and measurement (as well as the scope trace can be detected) shows an unbalance of 8% on the 12AT7 anodes without further ado - i.e. with the resistor of 22K instead of the CCS. This is almost good; an extra resistor of 4,3K in series with R22 would give balance. This is now regarding the use of a resistor for tail.]
One still needs to get Va(EF86) to just below 100V (bias for the 12AT6 is about 3V) to set the 12AT7 cathodes. I found that possible with R13 = 100K; R14 = 560K; R15 = 560 ohm. Finally calculating for global NFB of about 20dB, R17 = 47K.
A hint: First build: Leave R17 off and omit C12, cutting gain right down. Dc voltages can then be checked and set without fear of instability. When that is done, reconnect C12, which should give a slight hiss at the loudspeaker. One can bypass the input by a say 10K resistor; that could be normal for the output impedance of a pre-amp; probably less.
R17. Having the OPT output the wrong way round when connecting NFB is not a good idea (meaning very possible damage somewhere from the very loud squeal that would result). First try for polarity with R17 at least the feedback factor larger - say 1 megohm. The amplifier should not oscillate, but a definite increase/decrease in output noise will be noticed. A decrease means negative feedback, which indicates the way in which the transformer secondary should be wired.
Time is getting on and the post may exceed the alotted length, so until further. Setting NFB stability is next, also frequency response - that involves R12, C11, C13
IMPORTANT THOUGH: These values are given as a 'chopping block'. I would like experienced members to kindly check. Also, many members here have 'been there, done that'. The way forward is to peruse and come up with discussion, please! I think I know my way round tubes - but so do many others!
Follow:
Douglas posted simultaneously. He is quite correct. But it then needs to be mentioned that my above post worked with V+ = 460V. The 12AT7 would struggle to do an undistorted anode signal of 100Vpp at lower h.t. I do not suggest one keeps the h.t. up to just please the phase inverter. This must however be re-considered, starting with a fresh Vg1 for the KT88s etc. Later on that as well. (Bed time down here!)
Douglas posted simultaneously. He is quite correct. But it then needs to be mentioned that my above post worked with V+ = 460V. The 12AT7 would struggle to do an undistorted anode signal of 100Vpp at lower h.t. I do not suggest one keeps the h.t. up to just please the phase inverter. This must however be re-considered, starting with a fresh Vg1 for the KT88s etc. Later on that as well. (Bed time down here!)
I'm all ears.....does it make sense to lower the B+ a little farther? It would certainly reduce the start-up stress on the PS caps a bit. I'm just trying to stay close to TubeMack's initial target values....Per Johan's post above, would we need to change some values around the 12AT7 when using a lower B+?
Sometimes, the needs of the LTP rule the roost. We had a similar experience with "El Cheapo". The 'T7 triode sounds good with an IB of 3 mA. and a 200 to 220 V. plate to cathode differential.
The "check valve" technique used in EC to decouple "finals" from small signal circuitry minimizes voltage drop. The "check valve" consists of a small resistance (100 Ω in the EC case) in series with a UF4007. In EC, decoupling network losses are less than 2 V.
hey-Hey!!!,
IMO, application of type 12AT7 is an interesting exercise provided it doesn't get in the way. With lower B+ it is certainly running near its limits. For a LTP, type 6H6Pi is a very sweet tube, and not very expensive either...let alone when its performance is accounted for. Type 5687 is another *EXCELLENT* LTP valve. GE brands have sublime section-section matching.
cheers,
Douglas
Yes, the "super triodes" (5687, 6Н30П, and ECC99) are superb tubes and make good diff. gain blocks. However, none of them have the HD spectrum exhibited by the 'T7 that blends so very well with PP "finals".
FWIW, I can see a 'T7 section voltage amplifier and an ECC99 LTP as the small signal complement in Mullard style circuitry.
FWIW, I can see a 'T7 section voltage amplifier and an ECC99 LTP as the small signal complement in Mullard style circuitry.
If our mental exercise was just to accommodate that stereo power transformer, then forget about it, I'll look for a better fit. What would be ideal for that? Or would doing anything different upset the 'fit' we have with the 12AT7 and the KT88s? I could just as easily build two complete monoblocks into a single chassis with separate power transformers. The price would end up nearly the same (except shipping) and no mods would be necessary.
I'm still trying to figure out load lines so I can better understand Johan's excellent (and much appreciated) walk-through, it's getting a little clearer. I won't tackle understanding these wacky high voltage power supplies until after I get some more amplification circuitry figured out. I'm an old dog, these are new tricks.
..Todd
I'm still trying to figure out load lines so I can better understand Johan's excellent (and much appreciated) walk-through, it's getting a little clearer. I won't tackle understanding these wacky high voltage power supplies until after I get some more amplification circuitry figured out. I'm an old dog, these are new tricks.
..Todd
Last edited:
Downloaded! Thank you for all the proffesional work!
( nice job on the other schematic, I'm sure there will be plenty of interest in both!)I would like to keep the Watts up if at all feasible. 50W + would be great, but only if doable. I have already built a 35W tube amp, and would like to try some speakers with the new blocks that will be pushing things a little.
I'm bitting my nails waiting for Johan's next post. Bring it on! This thread is exceeding all expectations.....
Last edited: