Up till now I've been strictly a solid state guy. However, a considerable supply of VT components have come my way. This would include power and impedance matching xfmrs, PS filter chokes, and, of course a nice selction of VTs. Since I've never heard a VT amp, I'd like to give this a go. So far, I've got 807s, which would be the finals in a PP topology. So here are some design ideas I've had:
- Solid state front end, or VTs all the way?
- Circuit topology: Duplicate the usual solid state topology: LTP -> VAS -> Driver -> Finals? I figure that a 12BY7A would make a good VAS. Or try something more "conventional"?
- Parallel feedback for each final? Would that help tame down the tetrode "harshness"?
- Any other ideas?
[/list=1]
Thanks in advance.
807s are an excellent choice. Be sure to have a look at the STC Application Note for the 807. It is quite extensive.
Unless you're going into class *2, or using screen drive, it is probably best to use valves all the way (plus it looks better - very important
)
It shouldn't be necessary to have a four stage amplifier. Three stages should be plenty - input differential pair, drivers and output; or some variation thereof. Getting a huge amount of gain doesn't really help that much to reduce distortion because the amount of global NFB you can apply is limited by the odd stuff the OPT does at the extremes. Applying local feedback from the anode to the grid of each output valve, or from the anode to the grid of the driver of the other output valve may well help tame "tetrode harshness", for the lack of a better term.
Unless you're going into class *2, or using screen drive, it is probably best to use valves all the way (plus it looks better - very important
)It shouldn't be necessary to have a four stage amplifier. Three stages should be plenty - input differential pair, drivers and output; or some variation thereof. Getting a huge amount of gain doesn't really help that much to reduce distortion because the amount of global NFB you can apply is limited by the odd stuff the OPT does at the extremes. Applying local feedback from the anode to the grid of each output valve, or from the anode to the grid of the driver of the other output valve may well help tame "tetrode harshness", for the lack of a better term.
Depending on how you configure your output stage, you might even be able to get away with something even simpler. If you only need, say, 40V peak of drive, a 6SL7-based LTP will swing that easily. For a 2.5V input sensitivity (roughly 3.7V peak), you'll even have enough gain left to put in 8-9dB of feedback, if you desire.
Interpose cathode followers between LTP and output stage and you've got yourself a pretty nice amp.
Interpose cathode followers between LTP and output stage and you've got yourself a pretty nice amp.
I agree, I made a bunch of 807 PP pentode config with global NFB and they sound very nice. By now, I'm trying to enhance those results adding ultralinear feedback with separate winding and separate suppressor grid power supply. I aim to less power but better performance.
I think this scheme could be a good start to aproach tubes.
Cheers
Larry
I think this scheme could be a good start to aproach tubes.
Cheers
Larry
Miles Prower said:
Miles,
check this out:
http://www.bonavolta.ch/hobby/fr/audio/pacific.htm
It is a schema that was published by .Jean Hiraga in the French avantgarde magazine "L'Audiophile".
The design was described in the no 43 edition of 1988 with a single FET in the input2SK170BL. the divider was for the cascode 108k/2,6k.
Now the follwing ising to be interesting - it shows the cunningness of Mr Hiraga.
By changing slightly the bias fpor the cascode the content of the harmonics can be tweeked.
The stage gives some 30 V - enough IMO for the 807.
It is very responsive and needs a well laid out circuit. The following is mentioned with care. The earth for the zeners (all 4 - both channels have their own zeners) go to the same and single point. Or there will be
So have fun.
For a 2.5V input sensitivity (roughly 3.7V peak), you'll even have enough gain left to put in 8-9dB of feedback, if you desire. Interpose cathode followers between LTP and output stage and you've got yourself a pretty nice amp.
This is pretty much the topology I had in mind. So this design won't require the kind of inverse feedback that solid state needs. So no VAS.
That leaves:
LTP -> Drivers (Comon plate) -> 807s (Class AB(1) )
I'm going for an input sensitivity of 1.0V(rms)
Unless you're going into class *2, or using screen drive, it is probably best to use valves all the way (plus it looks better - very important )
So I guess it's best to leave the silicon out. Would it make any difference to use solid state in the power supply (silicon diodes, voltage regulators)?
This is pretty much the topology I had in mind. So this design won't require the kind of inverse feedback that solid state needs. So no VAS.
That leaves:LTP -> Drivers (Comon plate) -> 807s (Class AB(1) )
I'm going for an input sensitivity of 1.0V(rms)
Unless you're going into class *2, or using screen drive, it is probably best to use valves all the way (plus it looks better - very important )
So I guess it's best to leave the silicon out. Would it make any difference to use solid state in the power supply (silicon diodes, voltage regulators)?
> Solid state front end, or VTs all the way?
Are you trying to start a fight?
Sand driving vacuum was tried and widely rejected. When you spend some quality time with tubes, you will see that this is a field for the ultra-geeks, or for cheap gitar amps.
Much "tube sound" IS in the outputs; but the spice is in the preceeding stages. Don't sand-state until you try pure tube.
> Circuit topology: Duplicate the usual solid state topology:
LTP? VAS? Please don't use such foul language around bottles: they hate it.
Tubes (and their power and chassis) are expensive. Tube amps need to be simpler than sand amps.
Tubes give lower gain but also lower distortion than transistors. Like 5%, not 25% open-loop. The sand-head notion of making a LOT of gain to apply a LOT of NFB is misplaced: a zero-NFB tube is very listenable, and as Jason says, VAS-y stuff is pointless when the output transformer gives so much phase shift that even 20dB of negative feedback is difficult. One stage with voltage gain of 70-300 is often all you can use.
LTP practically forces a negative supply rail, very awkward when we only had hollow rectifiers. And it is not necessary, in a speaker amp, to have two high-impedance inputs, nor is the even-order distortion of a non-push-pull feedback comparator any real problem. And a LTP gives half the voltage gain per bottle of a grounded-cathode amp. LTP tube amps exist: most Fender guitar amps use a wacky LPT driver. But this is conciously low-NFB and also a lot about the sound in gross-overdrive, a range HiFi amps (should) never run in.
Given 6L6/807s, build the Standard Amp: hi-Mu voltage amplifier and concertina phase splitter. Old Reliable.
Used in Pentode, 6L6/807 has high output impedance, while modern HiFi speakers expect low source impedance. For this reason, NFB is commonly used. Usually overall, sometimes just back to the driver. Alternatively, you can strap Pentodes as Triodes, which is equivalent to internal NFB and gives low output impedance at the cost of high drive voltage and lower maximum power. 807/6L6 was invented to save us from the hard inefficient work of flogging power out of triodes; however even the 6L6's promoters felt triodes would remain the standard of fidelity. They were probably right: we spent 1938 through 1978 trying to get pentode Watts/Dollar with something like triode sound, and then triodes started sneaking back. OTOH, there is nothing like the pounding sound of a hard-flogged light-NFB pentode amp: perfect for some types of music.
Are you trying to start a fight?
Sand driving vacuum was tried and widely rejected. When you spend some quality time with tubes, you will see that this is a field for the ultra-geeks, or for cheap gitar amps.
Much "tube sound" IS in the outputs; but the spice is in the preceeding stages. Don't sand-state until you try pure tube.
> Circuit topology: Duplicate the usual solid state topology:
LTP? VAS? Please don't use such foul language around bottles: they hate it.
Tubes (and their power and chassis) are expensive. Tube amps need to be simpler than sand amps.
Tubes give lower gain but also lower distortion than transistors. Like 5%, not 25% open-loop. The sand-head notion of making a LOT of gain to apply a LOT of NFB is misplaced: a zero-NFB tube is very listenable, and as Jason says, VAS-y stuff is pointless when the output transformer gives so much phase shift that even 20dB of negative feedback is difficult. One stage with voltage gain of 70-300 is often all you can use.
LTP practically forces a negative supply rail, very awkward when we only had hollow rectifiers. And it is not necessary, in a speaker amp, to have two high-impedance inputs, nor is the even-order distortion of a non-push-pull feedback comparator any real problem. And a LTP gives half the voltage gain per bottle of a grounded-cathode amp. LTP tube amps exist: most Fender guitar amps use a wacky LPT driver. But this is conciously low-NFB and also a lot about the sound in gross-overdrive, a range HiFi amps (should) never run in.
Given 6L6/807s, build the Standard Amp: hi-Mu voltage amplifier and concertina phase splitter. Old Reliable.
Used in Pentode, 6L6/807 has high output impedance, while modern HiFi speakers expect low source impedance. For this reason, NFB is commonly used. Usually overall, sometimes just back to the driver. Alternatively, you can strap Pentodes as Triodes, which is equivalent to internal NFB and gives low output impedance at the cost of high drive voltage and lower maximum power. 807/6L6 was invented to save us from the hard inefficient work of flogging power out of triodes; however even the 6L6's promoters felt triodes would remain the standard of fidelity. They were probably right: we spent 1938 through 1978 trying to get pentode Watts/Dollar with something like triode sound, and then triodes started sneaking back. OTOH, there is nothing like the pounding sound of a hard-flogged light-NFB pentode amp: perfect for some types of music.
Sue me; I like cathode coupled (LTP) phase splitters. Put a CCS in the tail and apply NFB from the O/P trafo to the non-inverting grid. There's more than 1 way to make up for the lowish gain. Avoiding global NFB (the loop is around only 2 stages) and phase shift oscillation makes sense to me. There are several ways a "high" voltage gain block can be made that's very linear without trotting NFB out.
FWIW, I think FET voltage followers are OK. Otherwise, I have no use for "sand" in the signal path.
FWIW, I think FET voltage followers are OK. Otherwise, I have no use for "sand" in the signal path.
Are you trying to start a fight?
No, just trying something new and different.
Tubes give lower gain but also lower distortion than transistors. Like 5%, not 25% open-loop. The sand-head notion of making a LOT of gain to apply a LOT of NFB is misplaced: a zero-NFB tube is very listenable, and as Jason says, VAS-y stuff is pointless when the output transformer gives so much phase shift that even 20dB of negative feedback is difficult. One stage with voltage gain of 70-300 is often all you can use.
What can I say? After all, I am a "sand-head". That's why I figured I'd run this by you before committing to any particular topology. OK, I understand: less inverse feedback; ditch the VAS; hollow state front end.
Given 6L6/807s, build the Standard Amp: hi-Mu voltage amplifier and concertina phase splitter. Old Reliable.
My preliminary research into this pointed out a big problem with concertina type phase splitters: wildly assymmetrical levels of THD between sections. The LTP and cathodyne don't seem to have this problem. That's why I had it narrowed down to one or the other.
OTOH, there is nothing like the pounding sound of a hard-flogged light-NFB pentode amp: perfect for some types of music.
My kind of music anyway.
807s are an excellent choice. Be sure to have a look at the STC Application Note for the 807. It is quite extensive.
TNX audiousername
No, just trying something new and different.
Tubes give lower gain but also lower distortion than transistors. Like 5%, not 25% open-loop. The sand-head notion of making a LOT of gain to apply a LOT of NFB is misplaced: a zero-NFB tube is very listenable, and as Jason says, VAS-y stuff is pointless when the output transformer gives so much phase shift that even 20dB of negative feedback is difficult. One stage with voltage gain of 70-300 is often all you can use.
What can I say? After all, I am a "sand-head". That's why I figured I'd run this by you before committing to any particular topology. OK, I understand: less inverse feedback; ditch the VAS; hollow state front end.
Given 6L6/807s, build the Standard Amp: hi-Mu voltage amplifier and concertina phase splitter. Old Reliable.
My preliminary research into this pointed out a big problem with concertina type phase splitters: wildly assymmetrical levels of THD between sections. The LTP and cathodyne don't seem to have this problem. That's why I had it narrowed down to one or the other.
OTOH, there is nothing like the pounding sound of a hard-flogged light-NFB pentode amp: perfect for some types of music.
My kind of music anyway.
807s are an excellent choice. Be sure to have a look at the STC Application Note for the 807. It is quite extensive.
TNX audiousername
Miles Prower said:
Erm, but the cathodyne and the concertina is the same thing. Possibly another problem is that this type of phase splitter needs to produce the peak to peak drive voltages at the same time, which may or may not be problematic depending on the valve and the supply voltage. This article may also be of interest.
A possible choice of valve may be one of the single envelope triode-pentodes like ECL86/6GW8 (or preferably, a cheaper heater variant like 14GW8), or maybe ECL84/6DX8. You'll get a high µ triode and a fairly beefy cathodyne from the triode-strapped pentode section. You can see some of the triode-strapped curves here.
Miles Prower said:TNX audiousername
You're most welcome
Assuming you're driving a symmetrical load (nearly always the case in a push-pull amp), this is not correct; Kirchoff's Law enforces equal voltages on both halves. So distortion, frequency response, and source impedance are identical for both halves. The big disadvantage is that the split load does not have gain. That's why I tend to use constant-current diff amps (or LTPs if you prefer that terminology).
Here's what the original article said:
He was describing a design I've seen before: one common cathode stage driving one of the PP finals, plus another common cathode stage through a voltage divider that drives the other PP final. (I've seen that done in some solid state designs for logic gates or other digital apps.) He advised either using a cathodyne (used as a TTL driver) or LTP, since either topology doesn't give that distortion mismatch. He measured different levels of distortion, and that imbalance can't be a good thing.
He was describing a design I've seen before: one common cathode stage driving one of the PP finals, plus another common cathode stage through a voltage divider that drives the other PP final. (I've seen that done in some solid state designs for logic gates or other digital apps.) He advised either using a cathodyne (used as a TTL driver) or LTP, since either topology doesn't give that distortion mismatch. He measured different levels of distortion, and that imbalance can't be a good thing.
> asymmetrical distortion seems to be an incurable plague of phase inverters
I could argue: asymmetrical distortion is NOT necessarily bad for the ear. But in fact nobody has used that type splitter in a looong time.
"Concertina", Cathodyne, etc is an emitter-follower with a collector resistor (except we say cathode and plate) and two outputs.
And while you would think one tube making two outputs could not swing two grids as hard as two plate-loaded tubes, plate-loaded is often limited by nonlinearity and Cathodyne is mighty linear. Plate-loaded, THD tends to 5% when peak output is 20% of supply voltage, say 60V peak for 300V supply. If you want less than 5% THD you have to take less swing or get more clever. The Cathodyne working at 300V can make two 60V peak outputs at under 2% THD, often well under.
There are many ways to skin cats, and many great amps did it different. But for a starter, I suggest wiring a cathodyne and flogging the heck out of it. I feel that even Great Designers Of The Past neglected it in favor of more complicated ideas that are not necessarily better.
BTW, I do a lot of sand-head stuff too. I grew up with tubes, learned the strange world of vacuumless devices, I've worked-with and made money on both, and mostly I find sand-state perfectly respectable if you understand it. But when I design with tubes, I think letting sand in is a form of cheating, and possible pollution of vacuum simplicity. Oh, I may use an emitter follower, current-source, even a chip servo where it clearly solves a problem the best way. (I have a sneaky suspicion that a hi-volt BJT would make a great Cathodyne.) And I always thought vacuum rectifiers were wasteful pigs (though silicon rectification has evils also).
Someday I may regress to transformer coupling. It died because R-C coupling "works so much better for the money." But tubes and transformers go together well. If you can find and afford good transformers.
FWIW: I DO love the long-tail. Done it several times in several ways. One of my favorite little amps was a 12AX7 LTP with like -450V tail-supply (I distrust active current sources). Dunno if the LTP was key to the sweetness, or the TV Tuner twin-triode tubes I used for output.
The next step after the triode volt-amp/ cathodyne is the Williamson, the true Williamson with triode output not the UltraLinear output sometimes used with a Williamson frontend.
I could argue: asymmetrical distortion is NOT necessarily bad for the ear. But in fact nobody has used that type splitter in a looong time.
"Concertina", Cathodyne, etc is an emitter-follower with a collector resistor (except we say cathode and plate) and two outputs.
And while you would think one tube making two outputs could not swing two grids as hard as two plate-loaded tubes, plate-loaded is often limited by nonlinearity and Cathodyne is mighty linear. Plate-loaded, THD tends to 5% when peak output is 20% of supply voltage, say 60V peak for 300V supply. If you want less than 5% THD you have to take less swing or get more clever. The Cathodyne working at 300V can make two 60V peak outputs at under 2% THD, often well under.
There are many ways to skin cats, and many great amps did it different. But for a starter, I suggest wiring a cathodyne and flogging the heck out of it. I feel that even Great Designers Of The Past neglected it in favor of more complicated ideas that are not necessarily better.
BTW, I do a lot of sand-head stuff too. I grew up with tubes, learned the strange world of vacuumless devices, I've worked-with and made money on both, and mostly I find sand-state perfectly respectable if you understand it. But when I design with tubes, I think letting sand in is a form of cheating, and possible pollution of vacuum simplicity. Oh, I may use an emitter follower, current-source, even a chip servo where it clearly solves a problem the best way. (I have a sneaky suspicion that a hi-volt BJT would make a great Cathodyne.) And I always thought vacuum rectifiers were wasteful pigs (though silicon rectification has evils also).
Someday I may regress to transformer coupling. It died because R-C coupling "works so much better for the money." But tubes and transformers go together well. If you can find and afford good transformers.
FWIW: I DO love the long-tail. Done it several times in several ways. One of my favorite little amps was a 12AX7 LTP with like -450V tail-supply (I distrust active current sources). Dunno if the LTP was key to the sweetness, or the TV Tuner twin-triode tubes I used for output.
The next step after the triode volt-amp/ cathodyne is the Williamson, the true Williamson with triode output not the UltraLinear output sometimes used with a Williamson frontend.
Now that's very interesting. At this point, I'll give both a try and see how they compare.
I'm just the opposite. I have a couple of sand designs that work just great. I got the idea of trying VTs since I was acquired a whole bunch of components from helping a neighbor clean out an attic. Lots of stuff loose, and probably for some ham rig project that never got built. So why not? Google was helpful in getting somewhat up to speed concerning hollow state. So now I have some basis to go on. Of course, I'll have to build a few in order to see what's really going on. Does sound promising, though.
Perhaps fast recovery diodes would help?
Use high PIV Silicon Carbide (SiC) Schottky diodes and the reverse recovery spike (switching noise) disappears. ALL PN junction diodes, no matter how fast, exhibit a reverse recovery spike. Faster is better.
Assuming switching noise is not a problem, the instant on character of SS diodes can cause cathode stripping, particularly when the B+ rail voltage is high. There are several ways to slow the B+ rise down. A simple scheme is the use of negative temperature coefficient (NTC) thermistors. NTC devices work well, but you loose protection once they have heated up. Don't turn the amp off and back on quickly.
What I had in mind for the power supply was a "soft start" sequencer that would apply heater voltage in a couple of steps: 50% -> 100%, then apply the full V(pp). Since you can't just ditty-bop down to the neighbourhood drug store for new VTs any more, it makes sense not to hit the cold heaters with the full-on heater voltage. No sense blowing them unnecessarily.
This ought to do something about that cathode stripping problem as well.
This ought to do something about that cathode stripping problem as well.
You can always employ transformer-based phase splitting.
Why? I thought that all the trouble folks went through to find electronic methods was for the purpose of getting away from iron and its nonlinearity. I have one of these "interstage" phase spilt xfmrs. The core X-section area is 161.3E-6 sq. m. Yuckers!
There is no such thing as "too much iron"
Larger core X-Section? I already have a Stancor OPT that looks like it has quite a generous amount of core. Bigger than some power xfmrs.
Why? I thought that all the trouble folks went through to find electronic methods was for the purpose of getting away from iron and its nonlinearity. I have one of these "interstage" phase spilt xfmrs. The core X-section area is 161.3E-6 sq. m. Yuckers!
There is no such thing as "too much iron"
Larger core X-Section? I already have a Stancor OPT that looks like it has quite a generous amount of core. Bigger than some power xfmrs.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.