You don't need to run at high plate current. Reduction will increase the life of the EL34's. The gain does not always drop when the output goes from A to AB. Its not like with transistors the transition is quite soft. This is because the plate current curves verses grid voltage is not linear but is more power like. UL also makes this more complicated. Sometimes the gain actually increases in the class 'B' part with only one valve driving. So the THD does not always get better as you increase bias current. 40ma may work just fine. I have used Hammond transformers they have good HF and LF characteristics.
Last edited:
It is basically made according to the Telefunken data sheet shown on page 1 here. Light loading, high voltage and low current.
He gets more tha 30W from a pair of EL84 the same way....(RM-10?)
50W abuse?
Philips/Mullard/Fapesa was happy with pulling 100W
out of a pair EL34
800V on plates, 400V on screens, 11k plate to plate impedance.
Biased very cold, it was meant as a PA amplifier, usually driving 70V or 100V lines.
Was very popular here, specially because FAPESA (our Philips branch) published full data on how to wind power and output transformers; they were mainly used in Commercial duty, often in small town so called "open radio" service: companies creating radio program type shows, with news, music and paid announcements, no RF involved, "transmitted" along high impedance wired lines stretching for many City blocks, easily over half a mile each on every direction.
Philips/Mullard/Fapesa was happy with pulling 100W
out of a pair EL34 800V on plates, 400V on screens, 11k plate to plate impedance.
Biased very cold, it was meant as a PA amplifier, usually driving 70V or 100V lines.
Was very popular here, specially because FAPESA (our Philips branch) published full data on how to wind power and output transformers; they were mainly used in Commercial duty, often in small town so called "open radio" service: companies creating radio program type shows, with news, music and paid announcements, no RF involved, "transmitted" along high impedance wired lines stretching for many City blocks, easily over half a mile each on every direction.
Attachments
baudouino,
You are correct. I was wrong.
Some topologies and some quiescent points actually give expansion in the AB region.
I forgot that fact.
For example, in Pentode mode, the transconductance goes way up at or near 0 (Zero) g1 volts.
Although the other tube is already cut off completely, the much higher transconductance
makes up for not having the off tube able to give less current (less than 0).
But normally, Pentode mode needs Lots of negative feedback to get the amplifier output impedance low (to get a reasonable damping factor).
You are correct. I was wrong.
Some topologies and some quiescent points actually give expansion in the AB region.
I forgot that fact.
For example, in Pentode mode, the transconductance goes way up at or near 0 (Zero) g1 volts.
Although the other tube is already cut off completely, the much higher transconductance
makes up for not having the off tube able to give less current (less than 0).
But normally, Pentode mode needs Lots of negative feedback to get the amplifier output impedance low (to get a reasonable damping factor).
JMFahey,
Thanks for the story on "Broadcasting" Audio over 70V or 100V lines over several city blocks.
I do not prefer Mercury Vapor rectifiers (like the 83 ones in the schematic you posted).
But they did do it properly, they used Real Choke Input B+ filters (no cap before the choke).
And they also used 2 power switches.
The first switch warmed up the 83 filaments, the fixed bias tube rectifier filaments, and the rest of the amplifier tubes, including the output tubes.
Then, Only After all the filaments were warm, the second switch was closed.
Sometimes engineers do get everything right (lots of times).
Just do not give any credit to the engineers of the [first] Tacoma Narrows bridge in the State of Washington, USA.
Every engineering student should be required to watch that very old Black and White movie.
It even recorded the concrete flexing way beyond the maximum stress limits, before it literally "exploded".
The Audio "Broadcast" reminds me of my college days, where a college AM "radio station" broadcast throughout a dorm building.
The method used a low power AM transmitter's RF signal that was connected to the power mains of the dorm building.
This was all long before the dorm residents had PCs, Cell Phones, etc. and no switching power supplies on the power mains.
Thanks for the story on "Broadcasting" Audio over 70V or 100V lines over several city blocks.
I do not prefer Mercury Vapor rectifiers (like the 83 ones in the schematic you posted).
But they did do it properly, they used Real Choke Input B+ filters (no cap before the choke).
And they also used 2 power switches.
The first switch warmed up the 83 filaments, the fixed bias tube rectifier filaments, and the rest of the amplifier tubes, including the output tubes.
Then, Only After all the filaments were warm, the second switch was closed.
Sometimes engineers do get everything right (lots of times).
Just do not give any credit to the engineers of the [first] Tacoma Narrows bridge in the State of Washington, USA.
Every engineering student should be required to watch that very old Black and White movie.
It even recorded the concrete flexing way beyond the maximum stress limits, before it literally "exploded".
The Audio "Broadcast" reminds me of my college days, where a college AM "radio station" broadcast throughout a dorm building.
The method used a low power AM transmitter's RF signal that was connected to the power mains of the dorm building.
This was all long before the dorm residents had PCs, Cell Phones, etc. and no switching power supplies on the power mains.
Its not from understanding the maths/graphs of the tubes in any detail. I just noticed on LTspice on the newly built mullard 5-20 thread that the NF is causing the sinewave on the drivers to become slightly flattened which indicates the gain is getting too high on the peaks. I know your knowledge is much deeper than mine.
baudouin0,
When a sine wave becomes flattened on both the top and the bottom crests,
That is what I call a foreshortening (mild clipping).
The dominant harmonic distortion in that case is 3rd Harmonic Distortion.
Global Negative feedback is not the cause of that.
Typically, that can occur in a push pull output stage, especially before corrections are applied (corrections such as either topology and operating point, or negative feedback).
When a sine wave becomes flattened on Either the top Or the bottom crests,
That is what I call a foreshortening (mild clipping) of just one direction.
The dominant harmonic distortion in that case is 2rd Harmonic Distortion.
Global Negative feedback is not the cause of that.
Typically, that can occur in a single ended output stage, especially before corrections are applied (corrections such as either topology and operating point, or negative feedback). The topology can be where the 2nd harmonic distortion of the driver cancels the output state 2nd harmonic distortion.
Unfortunately this correction only works when the load impedance of the loudspeaker is constant (so cancellation is possible at some frequencies, but not at many other frequencies).
Causes of the foreshortening typically are things like output tube(s) plate curves, output transformer saturation, etc.
Using negative feedback from either the output transformer primary or the output transformer secondary:
If the transformer core is saturated for any reason, then the negative feedback will make it saturate even more.
Too few laminations on SE and Push Pull OPTs (especially for very low frequencies), and un-balanced DC currents in the primary of a push pull transformer
are common causes of saturation.
When a sine wave becomes flattened on both the top and the bottom crests,
That is what I call a foreshortening (mild clipping).
The dominant harmonic distortion in that case is 3rd Harmonic Distortion.
Global Negative feedback is not the cause of that.
Typically, that can occur in a push pull output stage, especially before corrections are applied (corrections such as either topology and operating point, or negative feedback).
When a sine wave becomes flattened on Either the top Or the bottom crests,
That is what I call a foreshortening (mild clipping) of just one direction.
The dominant harmonic distortion in that case is 2rd Harmonic Distortion.
Global Negative feedback is not the cause of that.
Typically, that can occur in a single ended output stage, especially before corrections are applied (corrections such as either topology and operating point, or negative feedback). The topology can be where the 2nd harmonic distortion of the driver cancels the output state 2nd harmonic distortion.
Unfortunately this correction only works when the load impedance of the loudspeaker is constant (so cancellation is possible at some frequencies, but not at many other frequencies).
Causes of the foreshortening typically are things like output tube(s) plate curves, output transformer saturation, etc.
Using negative feedback from either the output transformer primary or the output transformer secondary:
If the transformer core is saturated for any reason, then the negative feedback will make it saturate even more.
Too few laminations on SE and Push Pull OPTs (especially for very low frequencies), and un-balanced DC currents in the primary of a push pull transformer
are common causes of saturation.
Last edited:
Sorry what I mean is that if the gain of the output tube increases near the peak then the NF will try to correct for this. So you will see the drive waveform peaks flattened slightly. In the same way if the output clips and the gain drops to near zero you will see the drive waveform try to correct by making the peaks here bigger. In the 5-20 case assuming the EF86 is linear the waveform on the plate is basically the error signal (well 85% is) for 17dB NF. So any distortion here is a good indication of non-linearities in the driver and output stages. I'am not modelling OPT non-linearities but they are significant at LF. Yes if the transformer saturates the NF will try and drive it harder. However that will result in a cleaner output up to the point of clipping if the NF is taken from the output side.
Last edited:
I just looked back at the printout, and I realized that it's not great. The loadline looks fine, but the screen voltage section is supposed to be greyed out, as that's for pentode mode. It just looks like a class AB loadline to me. Is there anything wrong with a 70 percent bias at 400v? This is with a 5k P-P load and 40 percent screen taps. It seems to be well within what the datasheet rates the tubes for.
baudouin0,
I there is a super knowledgeable reader of this thread . . .
Please read my statements below, then if I am wrong, then
Either speak up now, or forever hold your peace.
When negative feedback drives a saturated transformer even harder to make up for the saturation, will only make it saturate even more, then the negative feedback will drive it harder yet, in an endless loop attempting to make up for the saturation.
That is a loosing battle.
I am talking about when the negative feedback is sampled from the output transformer signal.
Either the negative feedback is taken from the primary, or the negative feedback is taken from the secondary.
In both cases, negative feedback only makes a saturated transformer even more saturated.
Both IM and THD will increase.
I there is a super knowledgeable reader of this thread . . .
Please read my statements below, then if I am wrong, then
Either speak up now, or forever hold your peace.
When negative feedback drives a saturated transformer even harder to make up for the saturation, will only make it saturate even more, then the negative feedback will drive it harder yet, in an endless loop attempting to make up for the saturation.
That is a loosing battle.
I am talking about when the negative feedback is sampled from the output transformer signal.
Either the negative feedback is taken from the primary, or the negative feedback is taken from the secondary.
In both cases, negative feedback only makes a saturated transformer even more saturated.
Both IM and THD will increase.
Correct exactly I don't disagree with anything you have said there. The core saturation is soft however. So for small amounts of core saturation initially the drive will succeed and the output will be more linear than with no NF. However as you say its a loosing battle so when the output stage clips the output waveform will clip more suddenly. This actually will be more unpleasant than with no NF as it generates more HF harmonics. So before the output stage runs out of drive the NF will drop the distortion like any other non-linearity. But after that it will make it much worse.
Last edited:
...Back to the question of bias point is that when you drive the output stage at full power almost certainly you will exceed the plate dissipation of the EL34 for fixed bias. Cathode bias will back off and allow some protection and create crossover distortion instead - nice. I think something in hand is good for the tube so having a lower quiescent current helps dissipation on music (higher peak to mean ratio) rather than sine waves and increases life. Certainly for my 8x6550 amp its also good for the environment and not quite as good as a room heater. If the plates are glowing its too hot.
Last edited:
Up to you. All I am saying is that the lowest THD does not always occur when biased hot, and biasing hot does reduce life. If you can keep the tubes cooler most of the time that is a good thing. Obviously plenty enough bias to overcome crossover distortion but what's the correct point well that's a good question. Depends on a number of things valve, HT, UL position. I tend to look in ltspice - but its only as good as the models supplied. I think you will struggle with load lines. I believe most valves get destroyed by exceeding the screen dissipation rather than plate. Sorry to give you a rather confusing answer. The 6550's I ran at 40ma @ 430v with 40% UL with good results.
Last edited:
Hmm okay, thanks. My current thinking is that a 5k transformer with 40 percent screen taps and a B+ of 400 volts should yield low distortion biased at about -35 volts, or about 64% of max dissipation.
Another line with a 6.6k transformer and a B+ of 450v gets me low distortion at only 50 percent max dissipation, so maybe that's the way to go?
Mostly what I need to find out is which transformer from Edcor to order, the 6.6k or the 5k one.
I don't see myself using this amp constantly, so tube life probably won't be a huge issue.
I guess I could probably order either. I'll look at some amp designs to see what's most popular.
Let me know what you think, thanks!
Another line with a 6.6k transformer and a B+ of 450v gets me low distortion at only 50 percent max dissipation, so maybe that's the way to go?
Mostly what I need to find out is which transformer from Edcor to order, the 6.6k or the 5k one.
I don't see myself using this amp constantly, so tube life probably won't be a huge issue.
I guess I could probably order either. I'll look at some amp designs to see what's most popular.
Let me know what you think, thanks!
Not only will you not be using this amp constantly, you won't be using it at full power either. I would be surprised if average power over any 30 second period was more than 10 watts.
Tubes are very different from semiconductors in that the plate has a very large thermal mass, so you can get away with excessive plate dissipation for short periods of time. It is very common for amateur radio RF amplifiers to greatly exceed the plate dissipation ratings with key down, the idea being that both CW (Morse code) and SSB voice are not 100% duty cycle, and as long as it averages out over 10 seconds or so it doesn't cause a problem. Of course, this becomes a problem for "brick on the key" digital modes.
FWIW, I have tested both the 50W 6.6K Edcor and 50W 6.6K Hammond output transformers. The Edcor is flat with low distortion down to 20 Hz, which is significantly better than the Hammond was. That said, the Hammond was slightly better behaved at HF. IMO, sub-30 Hz performance is not a huge deal in a tube amp, but better HF performance makes it easier to tweak the feedback loop. Fit and finish was about the same. All of that said, the extra cost of the Hammond isn't worth it, IMO. Hammond can be bought through Mouser / DigiKey, so if you're really impatient they're the way to go for 6L6s.
The most common choice for EL34s seems to be around 5K primary impedance, but 6.6K may give slightly better performance at the cost of a small amount of power. Dynaco used 4.3K primaries on the ST70, and that seems to have worked out quite well.
Also, don't get too caught up in the whole frenzy over NOS tubes. New production tubes have come a long way in the last 20 years, and NOS tubes have gotten ridiculously expensive. I've had no shortage of NOS tubes arc, short or fail catastrophically in some way or another. Even after reactivation some of them still choose to pursue their dream of becoming a Thryatron. I always buy sets of new production tubes that have had a 24 hour burn in, and I can't recall having any issues with them.
Tubes are very different from semiconductors in that the plate has a very large thermal mass, so you can get away with excessive plate dissipation for short periods of time. It is very common for amateur radio RF amplifiers to greatly exceed the plate dissipation ratings with key down, the idea being that both CW (Morse code) and SSB voice are not 100% duty cycle, and as long as it averages out over 10 seconds or so it doesn't cause a problem. Of course, this becomes a problem for "brick on the key" digital modes.
FWIW, I have tested both the 50W 6.6K Edcor and 50W 6.6K Hammond output transformers. The Edcor is flat with low distortion down to 20 Hz, which is significantly better than the Hammond was. That said, the Hammond was slightly better behaved at HF. IMO, sub-30 Hz performance is not a huge deal in a tube amp, but better HF performance makes it easier to tweak the feedback loop. Fit and finish was about the same. All of that said, the extra cost of the Hammond isn't worth it, IMO. Hammond can be bought through Mouser / DigiKey, so if you're really impatient they're the way to go for 6L6s.
The most common choice for EL34s seems to be around 5K primary impedance, but 6.6K may give slightly better performance at the cost of a small amount of power. Dynaco used 4.3K primaries on the ST70, and that seems to have worked out quite well.
Also, don't get too caught up in the whole frenzy over NOS tubes. New production tubes have come a long way in the last 20 years, and NOS tubes have gotten ridiculously expensive. I've had no shortage of NOS tubes arc, short or fail catastrophically in some way or another. Even after reactivation some of them still choose to pursue their dream of becoming a Thryatron. I always buy sets of new production tubes that have had a 24 hour burn in, and I can't recall having any issues with them.
If 6.6k seems reasonable, which it does to me, that's what I'm going to go for, as from Edcor I can get it with 4, 8, and 16 ohm taps instead of just 8 ohms that the 5k one can. The 6.6k loadline I was looking at had low distortion at a cooler bias, so it should be easier on the tubes in the long run, especially if using fixed bias.
I guess that I'll work backward from the power stage to design an amp! Any reason to use a LTP like the Mullard design I've seen instead of a Cathodyne phase splitter? I'm more familiar with the cathodyne design.
Thanks!
I guess that I'll work backward from the power stage to design an amp! Any reason to use a LTP like the Mullard design I've seen instead of a Cathodyne phase splitter? I'm more familiar with the cathodyne design.
Thanks!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.