well, i can do the simulation in real life.
I have a transformer with symmetrical cathode feedback of roughly 100 watts, and the hammond 100 watts, which doesn't CF, it runs at about 13 db of GNF. The drivers is a 6sn7 gain to 6sn7 LTP with CCS, I have another version with a ECC99 LTP with a pentode CCS with negative supply, to be honest if I had to redo the amp I would use a 6sl7 modified para to a 6sn7 CF drivers which can drive 4x kt120 to like 500 watts without distortion.
I can easy introduce some feedback with the anode to input grid and see... the amp starts to have more THD at 50 watts then gradually clips to something like 120 watts.
I just want to carry tests at around 10 to 20 watts, I don't like testing at high power on nice transformers and hear them scream
its quite a big amp with big chokes, just 50uf + 10H choke + 100uf to power tubes which are kt120 , voltages are around 700V and 70 mV bias, didn't bother measure it.
SpreadSpectrum, why would we still make UL transformers if the solution is simply to use local anode feedback? what is the advantage of UL? why UL was developed in the first place if it was so easy to get a better result?
I have a transformer with symmetrical cathode feedback of roughly 100 watts, and the hammond 100 watts, which doesn't CF, it runs at about 13 db of GNF. The drivers is a 6sn7 gain to 6sn7 LTP with CCS, I have another version with a ECC99 LTP with a pentode CCS with negative supply, to be honest if I had to redo the amp I would use a 6sl7 modified para to a 6sn7 CF drivers which can drive 4x kt120 to like 500 watts without distortion.
I can easy introduce some feedback with the anode to input grid and see... the amp starts to have more THD at 50 watts then gradually clips to something like 120 watts.
I just want to carry tests at around 10 to 20 watts, I don't like testing at high power on nice transformers and hear them scream
its quite a big amp with big chokes, just 50uf + 10H choke + 100uf to power tubes which are kt120 , voltages are around 700V and 70 mV bias, didn't bother measure it.
SpreadSpectrum, why would we still make UL transformers if the solution is simply to use local anode feedback? what is the advantage of UL? why UL was developed in the first place if it was so easy to get a better result?
Last edited:
UL got rid of the (regulated) screen grid supply.
Shunt Schade also has problems with cheap OTs that have poorly matched coupling (to secondary) of the two plate windings. Using the UL taps for the Schade pickups should work better (lowest dist.) if stability problems don't develop. (assuming the UL taps are closest to the secondary winding)
Shunt Schade also has problems with cheap OTs that have poorly matched coupling (to secondary) of the two plate windings. Using the UL taps for the Schade pickups should work better (lowest dist.) if stability problems don't develop. (assuming the UL taps are closest to the secondary winding)
Last edited:
Because UL is simpler and cheaper. It was good enough and it saved them a few bucks.
Well, you are right, it is not ultra-linear, far from it.
I do like how it improves the sound. I am willing to try removing the UL connections and make the change to see if I like it.
I will also use that CFB transfo and listen and measure it versus the 1650R from hammond (+ 2x 300bx power transfo 2x 193J chokes)+, I got some data from the hammond so far...
Bandwidth is -3 db at 10 watts 6hz to 24.477khz, this is not very good, it was my first amplifier built like 15 years ago after liking so much the tube4hifi design. I am still shocked to hear how much neutral the kt88 are vs the kt77 and el34 which are becoming unacceptable to my ears, like I mean the sound on the kt77 is very analog and el34 have some charms, but the kt88 are so much better.
I tried IMD, at 10 watts, nothing with 20hz, 120hz, just regular THD at 500 and 5000, but at 2000, and 2001 hz, this is the hardest test there is an IMD at -52 db, then some THD and IMD products 2nd at -75 db, 3rd at -55.9, 4 th and 5 th at -75.7, nothing major, just good performance there which you expect form the kt120,
THD is aroud 0.17% at 1kz 20watts, 8khz it is 0.34% expected from the reduced bandwidth, same at 16khz nothing major, bass at 100hz is 0.2% at 20watts, 0.1% at 5 watts. at 1khz 5 watt it is 0.05%.
I do like how it improves the sound. I am willing to try removing the UL connections and make the change to see if I like it.
I will also use that CFB transfo and listen and measure it versus the 1650R from hammond (+ 2x 300bx power transfo 2x 193J chokes)+, I got some data from the hammond so far...
Bandwidth is -3 db at 10 watts 6hz to 24.477khz, this is not very good, it was my first amplifier built like 15 years ago after liking so much the tube4hifi design. I am still shocked to hear how much neutral the kt88 are vs the kt77 and el34 which are becoming unacceptable to my ears, like I mean the sound on the kt77 is very analog and el34 have some charms, but the kt88 are so much better.
I tried IMD, at 10 watts, nothing with 20hz, 120hz, just regular THD at 500 and 5000, but at 2000, and 2001 hz, this is the hardest test
there is an IMD at -52 db, then some THD and IMD products 2nd at -75 db, 3rd at -55.9, 4 th and 5 th at -75.7, nothing major, just good performance there which you expect form the kt120, THD is aroud 0.17% at 1kz 20watts, 8khz it is 0.34% expected from the reduced bandwidth, same at 16khz nothing major, bass at 100hz is 0.2% at 20watts, 0.1% at 5 watts. at 1khz 5 watt it is 0.05%.
Last edited:
One needs to look at the full picture. Ultimately the customer's preference is the determining factor. UL ultimately wimps out when the screen V gets too low. This should make for soft clipping.
Schade is more likely to clip hard, depending on how aggressive the driver tube is. UL makes sense for smaller Amps. For big Amps, Schade may make better sense, but it is hardly the last word in linearity. A big Amp is likely to put more loop gain into a more complex local N Fdbk to the driver scheme, if it is big enough to never clip.
And the customer may just like a certain "tube" sound anyway. Linearity doesn't seem to get much respect in reviews, just in sales brochures.
Schade is more likely to clip hard, depending on how aggressive the driver tube is. UL makes sense for smaller Amps. For big Amps, Schade may make better sense, but it is hardly the last word in linearity. A big Amp is likely to put more loop gain into a more complex local N Fdbk to the driver scheme, if it is big enough to never clip.
And the customer may just like a certain "tube" sound anyway. Linearity doesn't seem to get much respect in reviews, just in sales brochures.
You can't just add a resistor. It will mess up the DC conditions at the grid. You either have to add a capacitor in series (I don't like this approach, it won't handle overload well) or try to pull the DC voltage down at the grid by some means. It costs you extra supply rails to do it right, plus it requires a screen supply that the UL does not. It's significantly more expensive to a production operation. You don't do it unless you think you will make your money back and then some.
I built an amp like this and blogged about it here. I also used the Hammond 1650R for that amp and got 60kHz small signal bandwidth. I didn't measure at 10 Watts, or we could compare if plate-grid feedback gets better bandwidth. I never put feedback around the output transformer on that amp, so it is kind of comparable to a 300B push-pull open-loop amp, but the output stage is more linear than a 300B and has a lower rp.
By big amps, do you mean over 100W, .... 200W 1% THD whatever it means...
The more you type the more questions I have!
Please don't get upset by my ignorance, if you can tell me how does the screen gets too low in a clipping situation?
I believe from my readings that UL removes some THD caused by the transformer, reduces the tube output impedance to stabilize it, at the cost of gain, it divides the tube inside into smaller sections which accelerate electrons according to what they should do.
One big advantage of kt88 is it is so easy to drive, i didn't calculate, does Schade fb reduces more the gain of the output to obtain an equal reduction of THD, is it to a 1:1 ratio or is it more advantageous, I know that UL is a good solution.
60khz at what power? the amp is just there behind me, it is easy to measure it....
I love those 6bl7 drivers and the design, (ahh look at this : A 6F8G 6SN7 6N8P 6H8C 5692 ST TUBES IN ORIGINAL BOX NOS US EMERSON Smoked Glass | eBay) I can just plug back the kt88 into the amp and measure the bandwidth at the same power.
my LTP driver is a Motorola, special edition of the 6sn7, it is just very linear... it just goes through a cap, no transistors there, I find that with adequate tubes juice transistors are not required, as long as it is done properly.
The more you type the more questions I have!
Please don't get upset by my ignorance, if you can tell me how does the screen gets too low in a clipping situation?
I believe from my readings that UL removes some THD caused by the transformer, reduces the tube output impedance to stabilize it, at the cost of gain, it divides the tube inside into smaller sections which accelerate electrons according to what they should do.
One big advantage of kt88 is it is so easy to drive, i didn't calculate, does Schade fb reduces more the gain of the output to obtain an equal reduction of THD, is it to a 1:1 ratio or is it more advantageous, I know that UL is a good solution.
60khz at what power? the amp is just there behind me, it is easy to measure it....
I love those 6bl7 drivers and the design, (ahh look at this : A 6F8G 6SN7 6N8P 6H8C 5692 ST TUBES IN ORIGINAL BOX NOS US EMERSON Smoked Glass | eBay) I can just plug back the kt88 into the amp and measure the bandwidth at the same power.
my LTP driver is a Motorola, special edition of the 6sn7, it is just very linear... it just goes through a cap, no transistors there, I find that with adequate tubes juice transistors are not required, as long as it is done properly.
Unfortunately, I did not write the level down. It was something small. I didn't spend more time measuring because I was giving the amp away and needed to get it done. There are 10kHz square waves with level into 8 Ohms for a visual comparison on the blog post I linked.
The 6BL7 performed a little better at large swing than a modern production 6SN7 did, so that's why I used it for the driver. I set the output tube gain to ~3, so it needs more voltage swing than a 300B to drive to clipping.
I used the p-channel mosfet because it makes it so that you can direct-couple to only two resistors. It was extremely elegant and simple. If you use a tube, it will take more components or a coupling cap to make the DC conditions work out. An antimatter tube might work if you can get your hands on one.
The 6BL7 performed a little better at large swing than a modern production 6SN7 did, so that's why I used it for the driver. I set the output tube gain to ~3, so it needs more voltage swing than a 300B to drive to clipping.
I used the p-channel mosfet because it makes it so that you can direct-couple to only two resistors. It was extremely elegant and simple. If you use a tube, it will take more components or a coupling cap to make the DC conditions work out. An antimatter tube might work if you can get your hands on one.
"Big Amp" is likely to be relative to the type of speakers to be driven, the size of the room, the type of music etc. But yeah, 100 to 200 Watt is probably a Big Amp for most home situations. Maybe not for a dance scene or a rock concert.
The screen voltage on a UL configured tube gets pulled down from B+ as the tube plate pulls down the plate winding. (auto transformer action at the 40% tap) It could end up almost 40% reduced when the plate is pulling down maximum strength (near clipping).
If you look at the alternate plate curves (versus grid 2 voltage) on many tube datasheets, you will see the maximum current the tube can draw reduces as grid 2 voltage reduces. When the tube needs the most current slam, UL cuts off the air supply. While pentode mode holds the screen V up at max current conditions.
This is OK as long as the design takes this into account. Somewhat reduced max power out for the same tube.
TV Sweep tubes take this to the opposite extreme, with a big cathode that can supply more current than the tube can rightly supply, without warming up the plate to a toasty red Tubelab glow if maintained.
UL is close to making the tube into an adjustable triode. (versus % tapping)
But triodes are well known for their difficulty in pulling the plate down to low voltages. (the electrons just aren't attracted to low voltage) On the other hand, many beam pentodes can pull down to 50 or 60 Volts (the "knee" voltage), and some can even get down to 30 V or so. (the steady grid2 V allows this)
Transistors/Mosfets can pull down to 0.1 V often. Inability to pull down the plate to low voltage means loss of power output (reduced voltage swing), and increased heating of the device. Schade gives you full pull-down to the tube knee voltage (due to screen V staying constant). But the penalty is more abrupt clipping.
UL by emulating a triode does indeed lower distortion over a bare pentode. Approaching that of a triode. It is just that most later tubes were designed with grid 1 optimized for best gm, rather than low distortion. Ie, grid1 is kept close to the cathode, causing some grid wire proximity effects. (leading to roughly square law V to I transfer) Grid 2 (or plate for triodes) is far enough away from the cathode to avoid grid wire proximity effects, leading to classical 3/2 power transfer. This mis-tracking of power laws between the grids causes the triode mode curves to have an obvious rollover appearance at higher plate voltages. Now the older DHT triodes typically had more spacing between grid 1 and cathode, giving a better tracking of grid 1 and grid 2 (or plate), so less rollover of the triode curves (which cause 2nd H dist mainly). UL would work wonders on an old DHP.
Schade has the advantage (for later tubes) in that it injects the N Fdbk signal through the same electrode as which caused the forward distortion in the first place. Enhancing the tracking. And producing nice linear (no roll-over) triode curves. So less 2nd harmonic dist., and helping with higher orders too. With lower knee voltages (per pentode curves) as well to improve efficiency. But making clipping harder.
Putting the driver stage in the loop as well can increase the loop gain for N Fdbk (Schade or otherwise) to lower the distortion even further. With even sharper clipping. Win some, lose some.
The screen voltage on a UL configured tube gets pulled down from B+ as the tube plate pulls down the plate winding. (auto transformer action at the 40% tap) It could end up almost 40% reduced when the plate is pulling down maximum strength (near clipping).
If you look at the alternate plate curves (versus grid 2 voltage) on many tube datasheets, you will see the maximum current the tube can draw reduces as grid 2 voltage reduces. When the tube needs the most current slam, UL cuts off the air supply. While pentode mode holds the screen V up at max current conditions.
This is OK as long as the design takes this into account. Somewhat reduced max power out for the same tube.
TV Sweep tubes take this to the opposite extreme, with a big cathode that can supply more current than the tube can rightly supply, without warming up the plate to a toasty red Tubelab glow if maintained.
UL is close to making the tube into an adjustable triode. (versus % tapping)
But triodes are well known for their difficulty in pulling the plate down to low voltages. (the electrons just aren't attracted to low voltage) On the other hand, many beam pentodes can pull down to 50 or 60 Volts (the "knee" voltage), and some can even get down to 30 V or so. (the steady grid2 V allows this)
Transistors/Mosfets can pull down to 0.1 V often. Inability to pull down the plate to low voltage means loss of power output (reduced voltage swing), and increased heating of the device. Schade gives you full pull-down to the tube knee voltage (due to screen V staying constant). But the penalty is more abrupt clipping.
UL by emulating a triode does indeed lower distortion over a bare pentode. Approaching that of a triode. It is just that most later tubes were designed with grid 1 optimized for best gm, rather than low distortion. Ie, grid1 is kept close to the cathode, causing some grid wire proximity effects. (leading to roughly square law V to I transfer) Grid 2 (or plate for triodes) is far enough away from the cathode to avoid grid wire proximity effects, leading to classical 3/2 power transfer. This mis-tracking of power laws between the grids causes the triode mode curves to have an obvious rollover appearance at higher plate voltages. Now the older DHT triodes typically had more spacing between grid 1 and cathode, giving a better tracking of grid 1 and grid 2 (or plate), so less rollover of the triode curves (which cause 2nd H dist mainly). UL would work wonders on an old DHP.
Schade has the advantage (for later tubes) in that it injects the N Fdbk signal through the same electrode as which caused the forward distortion in the first place. Enhancing the tracking. And producing nice linear (no roll-over) triode curves. So less 2nd harmonic dist., and helping with higher orders too. With lower knee voltages (per pentode curves) as well to improve efficiency. But making clipping harder.
Putting the driver stage in the loop as well can increase the loop gain for N Fdbk (Schade or otherwise) to lower the distortion even further. With even sharper clipping. Win some, lose some.
Last edited:
Also, misleading "Ultra Linear" name that increased sales.
Plus, some winders routinely add taps when interleaving. Cheap and efficient.
Much less. Because there are 2 resistors, and the feedback is parallel, so input resistance of the output stage is the first resistor in parallel with the second one divided by the amplification factor. It is not so simple and cheap when "designers" learn "rules of dumbs" copying-pasting from textbooks. However, with operational amplifiers they started to teach how to use feedbacks routinely. What on this forum was named "Schade Feedback", is similar to opamp-based inverting amplifier.
An externally hosted image should be here but it was not working when we last tested it.
But nobody even tried to add to an opamp or any "special audio" invention an "Ultra-Linear feedback input" with a transfer curve different from normal inputs, because it is insane and useless.
Changed from which value to 1.5 Ohm?
If it is a cathode bias resistor, it looks like 100 times lower than needed, depending on G2 voltage.
I did a mistake, wavebourn, now I fixed it.
The transformer is connected both UL and CFB, but it has a lot of 2n harmonics, I have no idea what is going on.
it is fixed bias, the 1.5R is only to take the current reading... looks like there is some problems the way I wired that transformer.
second harmonics are super high, like it is not working in PP... weird
The transformer is connected both UL and CFB, but it has a lot of 2n harmonics, I have no idea what is going on.
it is fixed bias, the 1.5R is only to take the current reading... looks like there is some problems the way I wired that transformer.
second harmonics are super high, like it is not working in PP... weird
Last edited:
If it is fixed bias, then the transformer winding in series with the cathode resistor can not increase idle current. It can decrease it, but absolutely insignificantly, due to DCR of the winding.
On 70 mA KT120 supposed to have pretty high second order harmonic. It is a push-pull tube for class AB amps. If you want it to run in more linear region, consider lowering G2 voltage.
On 70 mA KT120 supposed to have pretty high second order harmonic. It is a push-pull tube for class AB amps. If you want it to run in more linear region, consider lowering G2 voltage.
Attachments
Last edited: