Stability has not been an issue yet though I've only clocked about 100hrs sofar.
I used a CCS on each of the output cathodes which should keep stability out of the equation until the output valves drift to far and the CCS can't hold things together. The CCS should adjust with both input drift and output drift without any adjustment. The negative rail (-300V comes up instantly with silicon rectification) and the +B comes up with a time delay.
I had real serious issues with getting it to work initially because the CCS's kept popping. Turned out that the instantanious switch on situation was putting about 300V across the CCS. In the end I had to build a ring of three CCS with a Horizontal deflection transistor (600V rated) as the main controlling element. No issues since and its had a few hundred on off cycles. One time I did think one had popped when things started to buzz badly. Turned out that the line voltage was down about a third and the CCS had fallen out of regulation, once the voltage came back up all was fine
Shoog
I used a CCS on each of the output cathodes which should keep stability out of the equation until the output valves drift to far and the CCS can't hold things together. The CCS should adjust with both input drift and output drift without any adjustment. The negative rail (-300V comes up instantly with silicon rectification) and the +B comes up with a time delay.
I had real serious issues with getting it to work initially because the CCS's kept popping. Turned out that the instantanious switch on situation was putting about 300V across the CCS. In the end I had to build a ring of three CCS with a Horizontal deflection transistor (600V rated) as the main controlling element. No issues since and its had a few hundred on off cycles. One time I did think one had popped when things started to buzz badly. Turned out that the line voltage was down about a third and the CCS had fallen out of regulation, once the voltage came back up all was fine
Shoog
ray_moth said:
26db of gNFB is way too much. I did two designs, one with PP 807s and another with PP 6BQ6GTBs. The 807 design uses ~7.0db of local NFB, and 6.0db of gNFB. The 6BQ6 design has adjustable gNFB that maxes out at 12db, but no local NFB. With that design, the full 12db of gNFB is definitely tending to a solid statey sound. The 807 design, when running with 20db of gNFB sounded absolutely horrible -- as bad as, or even possibly worse, than any SS amp. Do the open loop design right, and you'll need nowhere near 26db of gNFB.
There may have been more need for extreme gNFB back in those days. Back before the Thiele method of designing speeks, speek design was very much a black art, and a lot of the early speeks did some pretty funkey things with impedance over the audio band. The extra damping would have been needed to tame that.
Do it right, and you won't need high gNFB in pentode operation, but the topology will be hugely different. You'll be needing good regulation for the screens. I include active regulation to supply screen voltage to the finals. In both cases, these were hollow state active regulators, but there's no reason you couldn't solid state that instead.
An idea I was pondering...
What about using tv toobs, like lets say 6fw5's (I have a lot) as finals, screen driven. And drive it with pentodes and DC coupled source followers. Then use the UL tap (20%, 40%, whatever) to feed partial feedback, not to the finals, but back to the driver pentodes? Would that work out well? Even if the final stage was thinly biased?
What about using tv toobs, like lets say 6fw5's (I have a lot) as finals, screen driven. And drive it with pentodes and DC coupled source followers. Then use the UL tap (20%, 40%, whatever) to feed partial feedback, not to the finals, but back to the driver pentodes? Would that work out well? Even if the final stage was thinly biased?
hey-Hey!!!,
I have one similar amp. E-Linear with a pentode driver, but sandwiched in between is a triode cathode follower for hitting A2. It works quite well.
cheers,
Douglas
I have one similar amp. E-Linear with a pentode driver, but sandwiched in between is a triode cathode follower for hitting A2. It works quite well.
cheers,
Douglas
Thanks to all who answered my question, sorry to have taken a while to repond.
On the subject of partial NFB plate-plate NFB, as explained by John Broskie, I've read that article many times and I've tried modeling it with LT Spice more times than I can remember. One thing always eludes me: how to get enough current in the driver tube so it can swing enough to drive the OP tube without cutting off.
A better solution for me seems to be to have three differential stages, the first being a LTP splitter, and use cross-coupled NFB from the OP tube plates to the IP tube plates. I have a suspicion, though, that any kind of P-P NFB will work properly only in Class A.
I disagree with Miles (sorry) about the amount of NFB needed for pentodes. My problem is damping, because my speakers need plenty. They sound too "boomy", even with my EL34 triode-mode PP amp, unless I add either x-coupled or global NFB.
On the subject of partial NFB plate-plate NFB, as explained by John Broskie, I've read that article many times and I've tried modeling it with LT Spice more times than I can remember. One thing always eludes me: how to get enough current in the driver tube so it can swing enough to drive the OP tube without cutting off.
A better solution for me seems to be to have three differential stages, the first being a LTP splitter, and use cross-coupled NFB from the OP tube plates to the IP tube plates. I have a suspicion, though, that any kind of P-P NFB will work properly only in Class A.
I disagree with Miles (sorry) about the amount of NFB needed for pentodes. My problem is damping, because my speakers need plenty. They sound too "boomy", even with my EL34 triode-mode PP amp, unless I add either x-coupled or global NFB.
ray_moth said:
I don't see your disagreement: Miles said that the whole negative feedback not necessary must be a global one through the whole power amp to get the similar result, if I understood him right.
Miles said:
So he's talking about a lot less than 26dB in either case. I don't dispute that Miles likes the sound he's getting, I just think this is a case of "YMMV".
I been messin with LTSpice, trying to figure the best way to abuse
a pair of Pentodes in AB1... But not going so far as to define exactly
how "fixed" bias should be generated. Here's some simulated 50W
output stages. Assuming a pair of 6L6's and 450V B+.
The best of the batch seems to be in Pentode mode, with output
windings fed back to cathodes. This was clean well beyond the
rated output, without needing an absurd amount of drive.
The oddball here is Screen Drive. I had problems getting this one
to work well for me. Till I experimented to see what happens if I
dump 1W into G1. And run with G2 screen mostly in the negative,
and never higher than the lowest dip of the plate... G2 negative
(Space Charge Mode?) looks very promising, at least in the sim.
You may need to put the BeamTet symbol in your Misc Folder.
Same folder as the default Triode and Pentode symbols. Its
pins have been renumbered to work with the spice models.
a pair of Pentodes in AB1... But not going so far as to define exactly
how "fixed" bias should be generated. Here's some simulated 50W
output stages. Assuming a pair of 6L6's and 450V B+.
The best of the batch seems to be in Pentode mode, with output
windings fed back to cathodes. This was clean well beyond the
rated output, without needing an absurd amount of drive.
The oddball here is Screen Drive. I had problems getting this one
to work well for me. Till I experimented to see what happens if I
dump 1W into G1. And run with G2 screen mostly in the negative,
and never higher than the lowest dip of the plate... G2 negative
(Space Charge Mode?) looks very promising, at least in the sim.
You may need to put the BeamTet symbol in your Misc Folder.
Same folder as the default Triode and Pentode symbols. Its
pins have been renumbered to work with the spice models.
Drive goes positive, but never exceeds cathode, so its never in A2.
There is 1V of 120Hz ripple on the power supply to investigate for
possible PSRR problems? Good news, there don't appear to be any.
Then again, its only a sim...
And only a stage, not a complete amplifier...
Oops.. I left input set on 40Hz in the Zip above to tweak the Henries!
Also the sim time was set rather long to capture that. 1KHz will show
much better behavior. And the sim time can be very short, as filtering
caps for PS are not modeled, DC stabilizes rather quickly.
There is 1V of 120Hz ripple on the power supply to investigate for
possible PSRR problems? Good news, there don't appear to be any.
Then again, its only a sim...
And only a stage, not a complete amplifier...
Oops.. I left input set on 40Hz in the Zip above to tweak the Henries!
Also the sim time was set rather long to capture that. 1KHz will show
much better behavior. And the sim time can be very short, as filtering
caps for PS are not modeled, DC stabilizes rather quickly.
Attachments
Ray,
Let's start over at why NFB is used, to begin with. NFB is used for 2 reasons. Reducing O/P impedance, so that the damping factor is acceptable, and reducing distortion.
Regulating g2 B+ at a fraction of anode B+, along with good small signal circuitry, can yield a reasonably satisfactory distortion profile, without any NFB being used.
The damping factor needed depends on the speakers being used. Lowther driver based horn speakers will require very little loop NFB being applied. OTOH, AR3s or AR9s are going to need a LARGE amount of loop NFB being applied, to obtain the requisite voice coil control.
Let's start over at why NFB is used, to begin with. NFB is used for 2 reasons. Reducing O/P impedance, so that the damping factor is acceptable, and reducing distortion.
Regulating g2 B+ at a fraction of anode B+, along with good small signal circuitry, can yield a reasonably satisfactory distortion profile, without any NFB being used.
The damping factor needed depends on the speakers being used. Lowther driver based horn speakers will require very little loop NFB being applied. OTOH, AR3s or AR9s are going to need a LARGE amount of loop NFB being applied, to obtain the requisite voice coil control.
Duncan warns his model for G1 current is only an approximation.
So this sim run could be completely and totally invalid.... Seems
plausible??? Drive voltage swing is high, but very little current.
Also not sure if the Miller effect is modeled to G2 swings or not?
1W of continuous dissipation at G1 seems effective to replace
several Watts of discontinuous dissipation at G2, when screen
voltage would traditionally have been much higher. The plate
here would have to dip below 60V before drawing a G2 current.
More than enough plate swing for the targeted 50W output.
I am sure that allowing any G2 current would be a disaster.
Positive G1 could then focus beams directly onto G2! And I am
confident the Spice model is unaware of that effect. So I did
not care to press the design into the realm of G2 dissipation.
The FFT looks pretty decent for not having any NFB. Perhaps
G2 driven negative is asserting a constant Mu, like a Triode?
So this sim run could be completely and totally invalid.... Seems
plausible??? Drive voltage swing is high, but very little current.
Also not sure if the Miller effect is modeled to G2 swings or not?
1W of continuous dissipation at G1 seems effective to replace
several Watts of discontinuous dissipation at G2, when screen
voltage would traditionally have been much higher. The plate
here would have to dip below 60V before drawing a G2 current.
More than enough plate swing for the targeted 50W output.
I am sure that allowing any G2 current would be a disaster.
Positive G1 could then focus beams directly onto G2! And I am
confident the Spice model is unaware of that effect. So I did
not care to press the design into the realm of G2 dissipation.
The FFT looks pretty decent for not having any NFB. Perhaps
G2 driven negative is asserting a constant Mu, like a Triode?
Attachments
hey-Hey!!!,
I measured the E-Linear W6m's and they ran to 60Watts w/o making any nasty clipping. I suspect this is due to the high tap percentage putting the brakes on the driver stage a bit early. Since 40% is looking good for the finals, a new set of the 16431 Peerless irons will have to be re-made with additional lower taps. Given the reputed complexity I am sure Heyboer will be ever so glad to hear from me...🙂 Let's see, get rid of the 70V winding, separate the 4R coils so I can get 1 Ohm and paralleled copper for 4 Ohm...sort of like the OEM 20-20 Plus was done.
Also, has the tube-cad been modified to allow pentode finals? or did you do the model on triodes Ray?
cheers,
Douglas
I measured the E-Linear W6m's and they ran to 60Watts w/o making any nasty clipping. I suspect this is due to the high tap percentage putting the brakes on the driver stage a bit early. Since 40% is looking good for the finals, a new set of the 16431 Peerless irons will have to be re-made with additional lower taps. Given the reputed complexity I am sure Heyboer will be ever so glad to hear from me...🙂 Let's see, get rid of the 70V winding, separate the 4R coils so I can get 1 Ohm and paralleled copper for 4 Ohm...sort of like the OEM 20-20 Plus was done.
Also, has the tube-cad been modified to allow pentode finals? or did you do the model on triodes Ray?
cheers,
Douglas
Hi Douglas,
I don't have Tube-Cad, I have LT Spice.
Incidentally, I think I've found a good solution for a 40w pentode PP amp with good damping and low distortion. It's actually a 54w design but if it's only driven as far as 40w there seems to be a significant improvement in the distortion generated. It models well on LT Spice, both as a stand-alone single channel amp with fixed input voltages and as a fully-integrated stereo amp with SS regulated power supply.
The new design consists of:
Stage 1: 6SL7 LTP sp[litter, with transistor cascode CCS in the tail and single transistor CCS in each plate load. This is direct coupled to -
Stage 2: 6SN7 differential amp, also with transistor cascode CCS in the tail and single transistor CCS in each plate load. This is indirectly coupled to -
Stage 3: EL34 pentode PP, using fixed bias and an OPT giving 3.5k plate-plate load.
There is one global NFB loop, from the secondary of the OPT to the grounded grid of the 6SL7.
I don't have Tube-Cad, I have LT Spice.
Incidentally, I think I've found a good solution for a 40w pentode PP amp with good damping and low distortion. It's actually a 54w design but if it's only driven as far as 40w there seems to be a significant improvement in the distortion generated. It models well on LT Spice, both as a stand-alone single channel amp with fixed input voltages and as a fully-integrated stereo amp with SS regulated power supply.
The new design consists of:
Stage 1: 6SL7 LTP sp[litter, with transistor cascode CCS in the tail and single transistor CCS in each plate load. This is direct coupled to -
Stage 2: 6SN7 differential amp, also with transistor cascode CCS in the tail and single transistor CCS in each plate load. This is indirectly coupled to -
Stage 3: EL34 pentode PP, using fixed bias and an OPT giving 3.5k plate-plate load.
There is one global NFB loop, from the secondary of the OPT to the grounded grid of the 6SL7.
Ray,
Something I have borrowed from SS design which should be easy to check with your modelling.
Instead of "degraded" CCS in each anode try using Current Mirrors for the Anode Loads of the differential stages. They will provide reasonably high AC impedance loads while not "fighting" the tail CCS since they don't care what current they operate at, they simply try to keep current each side the same.
Cheers,
Ian
Something I have borrowed from SS design which should be easy to check with your modelling.
Instead of "degraded" CCS in each anode try using Current Mirrors for the Anode Loads of the differential stages. They will provide reasonably high AC impedance loads while not "fighting" the tail CCS since they don't care what current they operate at, they simply try to keep current each side the same.
Cheers,
Ian
Hi Ian,
Yes, I feel sure you're right. I'll have to try something better than degraded single-transistor CCSs as the plate loads. The original reason for degraded CCS goes away when we deliberately create a difference between cathode current and the sum of the plate currents.
The cathode current has to be set slightly higher than the sum of the plate currents. The surplus current is used to feed high-value resistors (around 2.2 Meg) that are connected in parallel with the plate load CCSs. This sets the plate voltages to suit the required conditions.
In fact under these conditions, the more accurate the CCSs in the plates, the better. There will be no conflict between CCSs, so no need for downgrading the plate load CCSs to let the cathode CCS "win".
Yes, I feel sure you're right. I'll have to try something better than degraded single-transistor CCSs as the plate loads. The original reason for degraded CCS goes away when we deliberately create a difference between cathode current and the sum of the plate currents.
The cathode current has to be set slightly higher than the sum of the plate currents. The surplus current is used to feed high-value resistors (around 2.2 Meg) that are connected in parallel with the plate load CCSs. This sets the plate voltages to suit the required conditions.
In fact under these conditions, the more accurate the CCSs in the plates, the better. There will be no conflict between CCSs, so no need for downgrading the plate load CCSs to let the cathode CCS "win".
ray_moth said:
Hi Ray,
I have built that amp( save for the CCS plate loads ). It switched to a 6SN7>>6BL7 front end shortly after...before getting shelved for the E-Linear pentode/cascode work. It could have done service as an integrated with the SL/SN combo. Now it is running 6AC7's and I doubt I'll get after it anytime soon.
cheers,
Douglas
P.S To run thr model, you'll need most of the files I've included. The
IRF820 in the PS uses a special circuit element that looks like a N-MOSFET with extra connection to the source-gate. If it doesn't work, you may need to use one of the existing symbols.
IRF820 in the PS uses a special circuit element that looks like a N-MOSFET with extra connection to the source-gate. If it doesn't work, you may need to use one of the existing symbols.
To run in reasonable time on my system, I had to dumb down the
circuit to just one channel, and completely fake the power supply.
My butchery may have ruined it??? Still seems to spice pretty good.
Was 1VAC in the speaker circuit to emulate a back EMF?
We can see GNF religiously track the input with back EMF disabled.
Your circuit must have had an absurd amount of gain before that
tracking correction is applied...
All the same its clear, GNF here makes for one very clean sinewave.
I'm not sure what it will do for music? And transients that may clip...
Like an op-amp with an iron hysteresis loop in the feedback path.
We may need a less idealistic model of an OPT to really know...
circuit to just one channel, and completely fake the power supply.
My butchery may have ruined it??? Still seems to spice pretty good.
Was 1VAC in the speaker circuit to emulate a back EMF?
We can see GNF religiously track the input with back EMF disabled.
Your circuit must have had an absurd amount of gain before that
tracking correction is applied...
All the same its clear, GNF here makes for one very clean sinewave.
I'm not sure what it will do for music? And transients that may clip...
Like an op-amp with an iron hysteresis loop in the feedback path.
We may need a less idealistic model of an OPT to really know...
- Status
- Not open for further replies.