Member
Joined 2009
Paid Member
well then, you are a naughty boy popping into this thread to advertise ...
we want all your secrets
No secrets. I revealed all of them one by one, many times.
Nested feedbacks, video pentodes, MOSFET voltage regulators, active servo-damping.
What I advertise, I show end results that can be achieved if to do all properly. I don't think that potential customers can be found in this thread. I will go to some audiophile-accepted magazines when ready.
Nested feedbacks, video pentodes, MOSFET voltage regulators, active servo-damping.
What I advertise, I show end results that can be achieved if to do all properly. I don't think that potential customers can be found in this thread. I will go to some audiophile-accepted magazines when ready.
from 845 datas
Ia=80mA
Ua=1150V
-Ug=175V
___________
dUa=120V
dUg=25V
dIa=80mA
___________
mu=4.8
Ri=1500 ohm
S=3.2 mA/V
___________
Cdyn=Cgk+(I-AI+1)Cga
Cga=13.5pF
Cgk=6pF
Cak=6.5pF
with mju cca to abs(-A) and 1pF from socket
Cdyn=91pF cca
for minimum phase shift
131KHz@-3db and @20KHz-0.25db
Rgen<=13350 ohm
round Cdyn to 100pF
for131KHz@-3db and @20KHz-0.25db
Rgen<=12150 ohm
that is for minimum generator resistance
(300B will have (with Ri about 800ohm) 15X less of needed value)
...
But the second and maybe more demanding issue will be large and undistorted as can be, voltage input to grid of 845 of about 350Vp-p. For full power or close to.
...
...
that is for minimum generator resistance
(300B will have (with Ri about 800ohm) 15X less of needed value)
...
But the second and maybe more demanding issue will be large and undistorted as can be, voltage input to grid of 845 of about 350Vp-p. For full power or close to.
...[/QUOTE]
How does this match up?
that is for minimum generator resistance
(300B will have (with Ri about 800ohm) 15X less of needed value)
...
But the second and maybe more demanding issue will be large and undistorted as can be, voltage input to grid of 845 of about 350Vp-p. For full power or close to.
...[/QUOTE]
How does this match up?
Looks like you had a second midlife crisis....
Mine was a 31’ boat with 2-300hp Yamahas
Back to opt question, look into a gm70, cheaper and easier to drive.
Im using old fashioned IT coupling with great results but iron is expensive
Last edited:
you are still missing the whole point, grid current is high, look at Tubelab SE, I read his post, he got 5% THD at 42 watts which is exactly what the tube can do, he measured:
around 70ma grid current peaks to obtain that, otherwise it will distort, especially in the bass department...
so a 800ohm drive or lower is required with possibility to send at least 50ma of current to the hungry grid.
around 70ma grid current peaks to obtain that, otherwise it will distort, especially in the bass department...
so a 800ohm drive or lower is required with possibility to send at least 50ma of current to the hungry grid.
There is another way to look at these kinds of amplifiers (845's, 300B's etc), that doesn't involve a numbers race for power output (expressed in Watts per Channel - misleading at best) or thermal efficiency. If either is your goal, find a better solution. This ain't it.
But, if your goal is an amplifier that's truly monotonic for distortion vs. signal level, has adequately low source impedance for practical loudspeakers, and can drive typical 88dB SPL/1W/1M speakers to live acoustic sound levels (say, 105 dB SPL peaks at near-ish listening position) at adequately low distortion, these can work.
The most difficult requirement is monotonicity - not at all even in a tiny bit trivial.
It doesn't actually require Kilovolt 50-watters to do this (trioded sweep valves are great), but they're fun in a boatanchor sense.
All good fortune,
Chris
But, if your goal is an amplifier that's truly monotonic for distortion vs. signal level, has adequately low source impedance for practical loudspeakers, and can drive typical 88dB SPL/1W/1M speakers to live acoustic sound levels (say, 105 dB SPL peaks at near-ish listening position) at adequately low distortion, these can work.
The most difficult requirement is monotonicity - not at all even in a tiny bit trivial.
It doesn't actually require Kilovolt 50-watters to do this (trioded sweep valves are great), but they're fun in a boatanchor sense.
All good fortune,
Chris
Hi Chris, one of my favorite amplifier is the Creek Destiny II, now an obsolete design, it had a lot of details and feedback. (surprisingly they use opamp in the high gain input stage, which can be bypassed with a pre-amp or high power source) their CD players too have opamps which again has plenty of detail... this is how much potential that amplifier had.
It is still my benchmark for how much resolution can be obtained in audio. It is equal or better to the best tube amps out there. It is not as smooth as Accuphase class A.
I strongly suggest to build something with the Tubelab driver. The only critic I can make to the tubelab is that with a sufficiently big power supply the output valve, 300B or 845 can be destroyed with over-current. Some current sensing could be incorporated to reduce the grid bias and prevent the valve from damage.
It is still my benchmark for how much resolution can be obtained in audio. It is equal or better to the best tube amps out there. It is not as smooth as Accuphase class A.
I strongly suggest to build something with the Tubelab driver. The only critic I can make to the tubelab is that with a sufficiently big power supply the output valve, 300B or 845 can be destroyed with over-current. Some current sensing could be incorporated to reduce the grid bias and prevent the valve from damage.
Last edited:
Can you explain this some? I have built 3 tube amps (two push-pull and one SE) and my measurements all show distortion that increases monotonically with level.
I took some really detailed measurements (small increases in level) on one Class-AB push-pull KT88 amp trying to catch it in the act of jumping up in distortion as my level pushed past the A-B boundary. I had just read Douglas Self's book and saw all of his results making similar measurements on Solid-State amps and was really excited to try to replicate the results on a tube amp. What I found was that standard push-pull AB amps were better than I imagined. Maybe because the gm curves of the tubes are much more complementary than the gm curves of solid-state devices?
Do you have results that show that standard tube amp topologies are not monotonic? I'd love to see them and learn more on this.
You can obtain very good transition to class AB (mono... whatever) choosing the right bias, it might not be the best sounding bias, nor be the lowest THD, it is a matter of trade off.
Now I learned something new, it seems that real triodes like 845 etc have a smoother hysteresis transition thd to class AB. This explains why I can get a very smooth and enjoyable sound from triodes in class AB.....
My UL amps have more detail and resolution, at the cost of less smooth transition....
Triodes are just more enjoyable, there is no doubt about it, they can also be more 'shooty' in the midrange...
For some reason, triodes + vinyl is so good... UL is less discriminating of the source. Triodes with CD is not the same.
Now I learned something new, it seems that real triodes like 845 etc have a smoother hysteresis transition thd to class AB. This explains why I can get a very smooth and enjoyable sound from triodes in class AB.....
My UL amps have more detail and resolution, at the cost of less smooth transition....
Triodes are just more enjoyable, there is no doubt about it, they can also be more 'shooty' in the midrange...
For some reason, triodes + vinyl is so good... UL is less discriminating of the source. Triodes with CD is not the same.
FWIW, none of these amps that I built have triode finals and I took detailed measurements at 50mA, 60mA, and 70mA bias. None of the three showed the expected increase in distortion at the point where devices started switching off at any bias level. What I'm saying is that I went in search of this phenomenon and couldn't measure it. I can only guess that tubes gm curves just are inherently very complementary and work very well for push-pull use over a wide range of overlaps.
I'm really kind of puzzled by your use of the term 'hysteresis' above. Not sure what you mean by that.
Last edited:
ellipse curve at the transition point.
This is what I read in theory books. I think someone is going to jump in and explain to us...
What I understand is that the higher bias = less IMD at lower power, but higher , harsher class AB.
Lower bias, less class A but better class AB
the recipe for a powerful class AB is a higher load impedance, or some cathode feedback, a lower B+ voltage but plenty of drive and current with fix bias....
If you have lower load impedance, your better with a less powerful amp... higher bias, leaner power supply...
Class AB power supply is the main and biggest problem, my next amplifier is a KT77 with fully regulated power supply.... I am thinking about it since 4 years... I still cannot make my mind on how much dissipation I should give in the power supply, I am aiming for 40 watts, at 16 hz with 'hopefully beautiful square waves in the 20hz... I already have all the parts and bits
This is what I read in theory books. I think someone is going to jump in and explain to us...
What I understand is that the higher bias = less IMD at lower power, but higher , harsher class AB.
Lower bias, less class A but better class AB
the recipe for a powerful class AB is a higher load impedance, or some cathode feedback, a lower B+ voltage but plenty of drive and current with fix bias....
If you have lower load impedance, your better with a less powerful amp... higher bias, leaner power supply...
Class AB power supply is the main and biggest problem, my next amplifier is a KT77 with fully regulated power supply.... I am thinking about it since 4 years... I still cannot make my mind on how much dissipation I should give in the power supply, I am aiming for 40 watts, at 16 hz with 'hopefully beautiful square waves in the 20hz... I already have all the parts and bits
Last edited:
Member
Joined 2009
Paid Member
no arguing with your experience but this seems awfully generalized.
My experience has been different. I've found pentodes to be more problematic (single ended) than triodes in terms of achieving smooth midrange. But even then, the sound was also dependent on other components and the topology so I would not assume this to always be true.
I suspect that you refer to the load-line. An idealized purely resistive load produces a single straight load-line on a set of plate curves in which the gradient of the line is derived from the value of the resistive load. But a real world speaker has inductance and capacitance which causes the load line to widen out into something resembling an ellipse. A sinewave input signal with a resistive load will cause the operating point to slide up and down the load-line. In the case of an elliptical load-line the operating point will move around the ellipse (clockwise or anti-clockwise depending on wether it is primarily inductive or capacitive). I've not measure this to see it for myself, I read it in Merlin's book (page 155) - well worth purchasing your own copy.
The venture into high grid current is brief and infrequent on normal music even when cranked to the point of touching clipping on loud peaks. Grid current can be an issue under continuous testing especially with square waves. Grid CURRENT is not the reason for tube damage, AVERAGE grid POWER DISSIPATION is. You can hit an 845 with very brief 100 mA pulses as long as the average power dissipation is kept below the point of damage.
There are two ways to fix this with the PowerDrive driver circuit, but either will reduce the maximum power output by reducing the point at which the output tube clips due to lower peak positive drive voltage.
The simplest and safest way to limit grid power is to lower the voltage on the drain of the mosfet to a value such that the grid voltage never exceeds the point of damage even if the mosfet were to fail to a short.
A more useful method is to place a resistor in series with mosfet's drain such that the drive when the amp is fed a square wave at the clipping level is at a safe, but maximum level. Add a capacitor from the drain to ground to allow for brief excursions into the "danger zone." The capacitor should be sized experimentally depending on the users choice of music and volume level.
It is also possible to use a CCS in series with the drain to limit peak grid current, and / or a Zener diode from drain to ground to limit peak grid voltage.
The new TSE-II board uses a Zener and a capacitor to limit grid power. I will be rebuilding my 845SE amp using one of the new ISE-II boards running a 45 as the driver.
Yes 42w was also the max value I got out of my 845SE in A2. You do need beefy driver stages (and a stiff DC path!) for successful A2.
I now run mine at lower dissipation to prolong life, which gives about 25w. It still runs into A2 to get that figure.
It would have been easier to use the Tubelab approach but I wanted an amp that if a flash-over happened, no damage would occur, apart from a possible valve swap.. Nothing religious about that.
Thank you for the answer Tubelab, p.s. the post is so outstanding, the amps so mouth watering, that my computer went into restart, it was too much. I have been a avid reader of your website, I especially liked your CFB, UL, Triode and transformer experiments and teachings.
Samsdad, the ticket to listening Beethoven on 845s is there. I would apply some GNF, voila.
Samsdad, the ticket to listening Beethoven on 845s is there. I would apply some GNF, voila.
Last edited:
As far as I know, I don't need anyone to explain anything to me on this. I've drawn composite curves for push-pull output stages in both Broskie's application and in Excel according to RDH4's process. It is amazing how linear and evenly-spaced the curves are for a standard push-pull output stage of reasonable bias current. I went through the exercise for KT88s, both operated as 'pentode' and 'triode' and various other tubes. All were astoundingly linear for anything above 50mA bias. There is some real magic in push-pull transformer drive and tubes. I didn't see any advantage to lower bias except tube life. Linearity seems to always increase at all points with increased bias current.
All of these configurations (pentodes, pseudotriodes, and real triodes) will handle elliptical load lines very well. Old-school triodes have beautiful, evenly-spaced curves for sure but high-gm power tubes like a KT88 operated as a pentode with plate-grid feedback (cathode feedback or shunt) will have far better curves (more evenly-spaced and lower rp) and less distortion into an elliptical load.
I have attached a graphic with 300B curves (yellow) KT88 triode-connection curves (blue) and KT88 20% feedback curves (magenta).
Attachments
The art in design of power amps is to find such "waterfall" that fools imagination the best. Distortions must be as close to distortions of our own perceptions as possible, so saturation would be perceived not as distortions, but as increased loudness. I.e. the amp must be as clean as possible on nominal listening level, while on peaks distortions and their order would raise such a way to avoid shrill, audible saturation.
I am pretty close to that, but still use my own subjective listening to judge. I am going to assemble an easy-tweakable amp, and after many listening tests with audiophiles find the best conditions.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.