A couple of us usual crackpots have been in discussion on the I&A forum trying to guide (yeah, right) a newbie-ish who was thinking about building a “big” amplifier for bass guitar using sweep tubes. Something about the size of my big 6550 monoblocks. In principle its quite possible, and pretty sure I could make it work as a demo. I’ve built a 135 watter around a quad of 26DQ5’s and it works perfectly. The only trouble I had with it was a vintage tube in the phase splitter with h-k leakage. Once corrected I had zero trouble. BUT some of us have bigger plans on the horizon. Bigger tubes, more of them, and therefore more opportunities for problems. The biggest of which is probably keeping the bias balanced. I don’t want to have to rely on precarious adjustment over time. These big sweeps aren’t necessarily matched, and probably drift over time.
I was thinking of an auto bias which would allow the cathode currents to increase under drive (unlike many approaches), but keep them in balance during the bias-up procedure and while in operation. The tubes would be individually fixed biased (since I know this works with random sweeps), and work in pairs. The idea would be to be able to have multiple pairs running simultaneously, however. Each with their own bias control. Id have to figure out the gain structure, and where the dominant pole needs to be (pretty low). I had something simplified auto-biasing a pair of un-matched type TIP transistors on the bench years ago, with large emitter resistors which somewhat limited the open loop gain, but it worked. Deliberately upsetting one side by injecting base current resulted in an increase in the other to match, until it ran out of headroom. And the error voltage would follow. I’m sort of out of actually building anything for the moment due to recent injury, and just looking back at this because I planned to use something like this on The Big One, and The Big One has come up yet again. Might not go anywhere - but I figured this needs its own thread if I want to explore it. Ive got several pairs (dozens) of “disposable” 20 watt class sweep tubes to play with, when I finally get to put this to hardware.
I was thinking of an auto bias which would allow the cathode currents to increase under drive (unlike many approaches), but keep them in balance during the bias-up procedure and while in operation. The tubes would be individually fixed biased (since I know this works with random sweeps), and work in pairs. The idea would be to be able to have multiple pairs running simultaneously, however. Each with their own bias control. Id have to figure out the gain structure, and where the dominant pole needs to be (pretty low). I had something simplified auto-biasing a pair of un-matched type TIP transistors on the bench years ago, with large emitter resistors which somewhat limited the open loop gain, but it worked. Deliberately upsetting one side by injecting base current resulted in an increase in the other to match, until it ran out of headroom. And the error voltage would follow. I’m sort of out of actually building anything for the moment due to recent injury, and just looking back at this because I planned to use something like this on The Big One, and The Big One has come up yet again. Might not go anywhere - but I figured this needs its own thread if I want to explore it. Ive got several pairs (dozens) of “disposable” 20 watt class sweep tubes to play with, when I finally get to put this to hardware.
Attachments
This could work well with some additions.
- you must keep bias low before heted up
- during heatup wiat a minute or so before attempting to adjust bias
- ignore bias except for abnormal current during play ( as protection )
- some circuit to detect silence and thus enable adjustment
- maybe remember bias settings at power off to restart faster.
Probably not for auto bias. The intention here is to manually SET the bias, but keep the other side in balance. In principle, manual adjustment of the other side would not be needed, but getting the pot set to minimize error voltage will maximize the headroom for additional correction. The LEDs should tell me when I’m in middle of the window - when the op amp is only making a small correction at quiescent. If the gm on one side is weak it will introduce a DC offset under drive. The op amp can then correct this by upping the bias in the weak side - on the fly. And reduce it back slowly at idle.
You could do that.. but you would have to periodically mute the amplifier to get cathode current measurements. If you only measure during start up its pointless if the amp stays on for long times.
Moving average filter in software could work, or trigger the ADC to take a cathode current measurement when both anode voltages are more or less equal
I worked in a large Motorola plant in Florida for over 40 years. In the early years (late 70's through 90's) the Motorola ham radio club held a free flea market in the parking lot twice a year. There was a father / son pair that showed up every time and sold tubes. They were also at all of the local hamfests. I made guitar amps at the time, so I got to know them and visited their warehouse in Boynton Beach (before the move to Orlando). They were Stan, and his father Irving of ESRC. One day Stan called me and asked me if I wanted a large quantity of JUNK Chinese KT88's cheap. These were early vacuum tubes from Shuguang and, yes, they sucked big time. I got over 100 for somewhere between $1 and $2 each. Many would spark out or glow red or blue on initial power up. I was making "Turbo Champ" SE guitar amps at the time and a Sovtek 6550WE could make about 10 watts in SE pentode mode on about 430 volts. Only the strong Chinese tubes survived. I had one of the Chinese fireworks spark out so violently that the glass shattered! I had fun blowing up all the junk tubes and eventually wound up with 20+ good ones that worked without issue. I still have 5 or 6 that will work at 100 watts per pair with no drama. They are my test mules for experiments that could end badly.
Somewhere in the late 80's I traded my very first car, still in running condition for an H.H. Scott 272 integrated tube amp and matching tuner. After replacing two broken resistors this thing kicked all the Carver and Phase Linear stuff right out of my stereo rack, so I decided to make tube HiFi amps......long before Tubelab was a dream.
I made guitar amps ever since childhood, made DIY conputers dating back to the SWTPC 6800 system in 1975, tinkered with ham radio, built fast cars and blew up engines, but also had my own color darkroom and shot a lot of film. Somewhere in a photography magazine in the 90's I think, I found a link to a photography web site, which led me to this:
http://www.normankoren.com/Audio/TENA.html
Of course I built one, or something somewhat like it. I cut many corners using junk parts like guitar amp quality OPT's, and Mil surplus power transformers, but it worked, and worked well enough that I eventually sold the Scott stuff for far more than my rusty old 1949 Plymouth was worth. The notable thing here is the autobias stuff that works similar to the way you discuss, one tube is the master and the others track it. His amp even uses cathode feedback with auto bias sensing on the cathode. Not sure how well it works, but it did not work well with a cheap OPT, so my version had 4 pots.
I plan to use a microcontroller to control the bias. Yes, some code writing will be involved, but I have some knowledge there, and some history. Back in 2007 I used some technology I had developed at Motorola for use in cell tower amps to get better than 50% efficiency in a SE triode vacuum tube amplifier. It does this by dynamically adjusting the B+ voltage on a cathode follower output stage to keep it just above saturation. I entered this in a design contest for Circuit Cellar magazine and Microchip where it won a prize and a magazine article publish. The article from 2009 is included here.
The schematic of the amp is in figure three. There is a tap of the audio signal from the driver tube plate through R8. This lets the dsPIC chip look for several successive similar samples to find a moment of silence during which idle currents are measured and adjusted as needed. I plan to adapt some of this old tech and code to a Teensy module which is a 32 bit Arduino compatible part that runs at about 600 MHz.
Somewhere in the late 80's I traded my very first car, still in running condition for an H.H. Scott 272 integrated tube amp and matching tuner. After replacing two broken resistors this thing kicked all the Carver and Phase Linear stuff right out of my stereo rack, so I decided to make tube HiFi amps......long before Tubelab was a dream.
I made guitar amps ever since childhood, made DIY conputers dating back to the SWTPC 6800 system in 1975, tinkered with ham radio, built fast cars and blew up engines, but also had my own color darkroom and shot a lot of film. Somewhere in a photography magazine in the 90's I think, I found a link to a photography web site, which led me to this:
http://www.normankoren.com/Audio/TENA.html
Of course I built one, or something somewhat like it. I cut many corners using junk parts like guitar amp quality OPT's, and Mil surplus power transformers, but it worked, and worked well enough that I eventually sold the Scott stuff for far more than my rusty old 1949 Plymouth was worth. The notable thing here is the autobias stuff that works similar to the way you discuss, one tube is the master and the others track it. His amp even uses cathode feedback with auto bias sensing on the cathode. Not sure how well it works, but it did not work well with a cheap OPT, so my version had 4 pots.
I plan to use a microcontroller to control the bias. Yes, some code writing will be involved, but I have some knowledge there, and some history. Back in 2007 I used some technology I had developed at Motorola for use in cell tower amps to get better than 50% efficiency in a SE triode vacuum tube amplifier. It does this by dynamically adjusting the B+ voltage on a cathode follower output stage to keep it just above saturation. I entered this in a design contest for Circuit Cellar magazine and Microchip where it won a prize and a magazine article publish. The article from 2009 is included here.
The schematic of the amp is in figure three. There is a tap of the audio signal from the driver tube plate through R8. This lets the dsPIC chip look for several successive similar samples to find a moment of silence during which idle currents are measured and adjusted as needed. I plan to adapt some of this old tech and code to a Teensy module which is a 32 bit Arduino compatible part that runs at about 600 MHz.
Well, I’m considering a years worth of board revisions, throwing things, and cussing a computer program to be too much work for just keeping the DC out of the core of a power toroid being used as a 300+ watt output transformer. I’d trust Nyquist to tell me where the poles are when they can be calculated, and not some random function of code execution time. My coding skills just aren’t going to get good enough to do anything timing critical. I don’t want to have to meter and adjust it every other time I power it up, or after several hours at war volume - just trust that it will stay balanced. Matched and burned-in power tubes aren’t in the cards, considering the sources of these things.
I did manage to scrounge up 24 of the same 21EX6 (the GE/Raytheon/Arcturus variety), plus a few oddballs one of which looks like the Raytheon 6HJ5 inside. That’s what I was going to use for prototyping the power stage and getting the servo up and running. At $4 a pop, OCTAL and still being able to push out half an amp apiece without g2 over 150V they seemed the best choice for early experiments. I simply couldn’t obtain enough of my first choice (26DQ5). Don’t want to go directly to those LW6’s or booger up a bunch of compactron sockets “playing”.
I did manage to scrounge up 24 of the same 21EX6 (the GE/Raytheon/Arcturus variety), plus a few oddballs one of which looks like the Raytheon 6HJ5 inside. That’s what I was going to use for prototyping the power stage and getting the servo up and running. At $4 a pop, OCTAL and still being able to push out half an amp apiece without g2 over 150V they seemed the best choice for early experiments. I simply couldn’t obtain enough of my first choice (26DQ5). Don’t want to go directly to those LW6’s or booger up a bunch of compactron sockets “playing”.
Assuming that I actually build the "big one" instead of just talking about it, I will have the same DC balance issues with my big Plitron audio output toroids. Experiments have determined that the distortion at LF is highly dependent on the DC offset between the two sides of the OPT. Absolute zero is not the best spot, but it is very low, a few mA and the two transformers are not exactly the same.
My choice of tubes in the final amp will likely be the 26HU5 unless I find that 4 'LW6s will do the job where it will probably take 6 X 26HU5s per channel to make 500 WPC for extended periods of time. I have about 25 LW6s (26 or 36 don't remember which) but they are not all the same. I have almost 50 26HU5's and most are identical GE's.
All development work will be done with some small sacrificial sweep tubes that I have lots of. I know that I got a box of at least 100 6DG6s with a few 6W6s mixed in. I have maybe 30 6EZ5s. These have the same pinout as the usual audio tubes like the 6V6, 6K6 and 6L6GC. If you connect a plate cap to pin 3, then the 6BQ6GT, 6BQ6GA(bigger), 6CU6(same as the 6BQ6GA), 6DQ6(even bigger), and the 6GW6 will work in the same socket. I have a box full of the 6BQ6GA/6CU6 somewhere. I have no 21EX6's and only one old RCA 6EX6. It doesn't look like it would live long in one of my experiments. The plates in the BQ6GA/6CU6s are bigger.
Either way you look at managing a big tube amp that runs on 600+ volts, the early experiments are going to make some dead parts. Back in about 1990 I bought a big batch of guitar amp quality 6600 ohm to 0-4-8-16, "80VA" OPTs. I set out to see how much power I could squeeze through one. Somewhere north of 150 watts on about 750 volts my Radio Shack quality dummy load bank went open. This caused the OPT to instantly burst into flames leaving a really big burnt spot on my workbench that remained until I sold the house in 2014. There were also parts of a fat electrolytic cap embedded in the ceiling, but that was a different non audio Dumm Blonde story.
I never got to finding out how many of those OPT's and sweep tube pairs could be connected together to make a Marty McFly guitar amp. I still have a bunch, so maybe it's time.......
My choice of tubes in the final amp will likely be the 26HU5 unless I find that 4 'LW6s will do the job where it will probably take 6 X 26HU5s per channel to make 500 WPC for extended periods of time. I have about 25 LW6s (26 or 36 don't remember which) but they are not all the same. I have almost 50 26HU5's and most are identical GE's.
All development work will be done with some small sacrificial sweep tubes that I have lots of. I know that I got a box of at least 100 6DG6s with a few 6W6s mixed in. I have maybe 30 6EZ5s. These have the same pinout as the usual audio tubes like the 6V6, 6K6 and 6L6GC. If you connect a plate cap to pin 3, then the 6BQ6GT, 6BQ6GA(bigger), 6CU6(same as the 6BQ6GA), 6DQ6(even bigger), and the 6GW6 will work in the same socket. I have a box full of the 6BQ6GA/6CU6 somewhere. I have no 21EX6's and only one old RCA 6EX6. It doesn't look like it would live long in one of my experiments. The plates in the BQ6GA/6CU6s are bigger.
Either way you look at managing a big tube amp that runs on 600+ volts, the early experiments are going to make some dead parts. Back in about 1990 I bought a big batch of guitar amp quality 6600 ohm to 0-4-8-16, "80VA" OPTs. I set out to see how much power I could squeeze through one. Somewhere north of 150 watts on about 750 volts my Radio Shack quality dummy load bank went open. This caused the OPT to instantly burst into flames leaving a really big burnt spot on my workbench that remained until I sold the house in 2014. There were also parts of a fat electrolytic cap embedded in the ceiling, but that was a different non audio Dumm Blonde story.
I never got to finding out how many of those OPT's and sweep tube pairs could be connected together to make a Marty McFly guitar amp. I still have a bunch, so maybe it's time.......
I had several steps planned for “the big ones”. First, a prototype of the “350” watter. Using the 8T800 as the OPT. Eight of those 21EX6’s ought to drive it, on 700 volts (expected to reduce to 600 at load). Once working, go to the 6HJ5 which is a more modern better tube. Got a batch big enough for a stereo pair plus enough to retube. And I just ordered some spares. The catch is that it’s 8 ohm only, unless I want to drive 700 ohms instead of 1400, and reduce the supply voltage, requiring higher current tubes. For THAT exercise, I have a batch of 21LG6, which can push more current without using up my 26LW6 stash which I need for The Big One. For that exercise, I may use a pair of interleaved 36V toroids per channel I have laying around and limit the supply to roughly 400V at load to get the target 340V peak across each primary, and get the right power at 4 ohms.
The big one will be 4 ohm only, since no single driver box can handle the power anyway. That 15T950 works out to 1k a-a. If I can get 700V at load, it should make 800 watts. 800V at load to get a kW is probably just wishful thinking, even with a 2kVA supply. Six paralleled LW6’s will stay around 43 watts each at max dissipation, and most of the time it would be less. Don’t need crazy current - I may even get away with 100-110V g2. Of course, I’m going to want TWO of them in the end. Think of it - those driving four of my 32 cubic foot 30 Hz bass horns, 350 per side driving horn loaded 12” mids (12MH32’s), and a pair of 2” D3300 compression drivers driven by my 6550 “bass” amps.
In the 26LW6 department ive managed to scrounge together 24 of the GE large bottle single plate style, 35 of the smaller bottle Sylvania, and a handful of oddballs. One good pair of the double-plate variety, and one good pair of the 36 volt large bottle GE. Those will get used for something special. Will any of it ever get built? You bet, just a question of when. I didn’t think i’d ever see the day when my new house gets built, and that’s moving closer and closer despite my recent foot injury.
The big one will be 4 ohm only, since no single driver box can handle the power anyway. That 15T950 works out to 1k a-a. If I can get 700V at load, it should make 800 watts. 800V at load to get a kW is probably just wishful thinking, even with a 2kVA supply. Six paralleled LW6’s will stay around 43 watts each at max dissipation, and most of the time it would be less. Don’t need crazy current - I may even get away with 100-110V g2. Of course, I’m going to want TWO of them in the end. Think of it - those driving four of my 32 cubic foot 30 Hz bass horns, 350 per side driving horn loaded 12” mids (12MH32’s), and a pair of 2” D3300 compression drivers driven by my 6550 “bass” amps.
In the 26LW6 department ive managed to scrounge together 24 of the GE large bottle single plate style, 35 of the smaller bottle Sylvania, and a handful of oddballs. One good pair of the double-plate variety, and one good pair of the 36 volt large bottle GE. Those will get used for something special. Will any of it ever get built? You bet, just a question of when. I didn’t think i’d ever see the day when my new house gets built, and that’s moving closer and closer despite my recent foot injury.
We are both planning to use toroidal OPT's. They are somewhat picky about DC imbalance and audible distortion on bass frequencies can happen with far less imbalance than with a conventional EI OPT.
When you say double-plate variety do you mean that it has two smaller tubes inside wired in parallel like the 6KN6 in the first picture? Early 6KN6's had two smallish sweep tubes inside wired in parallel. This gave the highest peak current capability (1.5A) of any sweep tube I have ever seen, but many were poorly made, and even a good one will red plate in the area where the plates meet. These can often be found cheap at hamfests since they will not work right in RF power amps due to high capacitance. The later 6KN6s were simply remarked 6KD6s with one set of guts.
The second picture shows a fat bottle 6LW6 with two large fins attached to the plates at the seams. These can eat a lot of power without redness, but this is one of four bad ones that I have. This tube was well made and could eat 80 watts with no hint of redness. I was using them in a triode wired SE amp which was basically a hacked up SSE board feeding a Hammond 1628SEA OPT. The first version ran about 400 volts of B+ in cathode bias and the cathode was around 75 volts. In about 6 months of daily use two of these tubes went gassy and into red plate runaway. The second version of the amp got a different power transformer and the OPT was rewired for 2.5K ohms. B+ was adjusted to put the screen to cathode voltage at 275 (spec is 280) with the total dissipation just under 40 watts. It took longer, but two more tubes died before I dismantled the amp.
I guess our definitions of "kilowatt amp" are a bit different as well. Since my OPTs are rated for "400 watts @ 20 Hz" I'm looking at a 2 channel amp with 500WPC. I have no use for such an amp and no speakers capable of eating that power, but I would do it just because I can. Its right up there with "why do you need 500 horsepower when you sit in rush hour traffic moving slower than walking speed." Yes, If I left work at 5:00 I could walk home faster than driving. It just happens to rain nearly every afternoon in the summer in South Florida.
The "proof of concept" as we called my prototypes at work is this. The final circuit will be something like this with either 2 or 3 UNSET Power Head boards per channel and one Universal Driver board. This test happened last December. The little SMPS in the upper left corner lights the 26HU5's and a big variable supply is set on 625 volts which is delivered to the AC polypropylene cap via some Radio Shack clip leads. Power output is 350 watts @ 0.817% THD and 525 watts @ 3.79% THD with visible clipping. There is a slight red glow in the tube plates in a dark room at the 350 watt level with none seen at full power (normal behavior and idle current dependent).
When you say double-plate variety do you mean that it has two smaller tubes inside wired in parallel like the 6KN6 in the first picture? Early 6KN6's had two smallish sweep tubes inside wired in parallel. This gave the highest peak current capability (1.5A) of any sweep tube I have ever seen, but many were poorly made, and even a good one will red plate in the area where the plates meet. These can often be found cheap at hamfests since they will not work right in RF power amps due to high capacitance. The later 6KN6s were simply remarked 6KD6s with one set of guts.
The second picture shows a fat bottle 6LW6 with two large fins attached to the plates at the seams. These can eat a lot of power without redness, but this is one of four bad ones that I have. This tube was well made and could eat 80 watts with no hint of redness. I was using them in a triode wired SE amp which was basically a hacked up SSE board feeding a Hammond 1628SEA OPT. The first version ran about 400 volts of B+ in cathode bias and the cathode was around 75 volts. In about 6 months of daily use two of these tubes went gassy and into red plate runaway. The second version of the amp got a different power transformer and the OPT was rewired for 2.5K ohms. B+ was adjusted to put the screen to cathode voltage at 275 (spec is 280) with the total dissipation just under 40 watts. It took longer, but two more tubes died before I dismantled the amp.
I guess our definitions of "kilowatt amp" are a bit different as well. Since my OPTs are rated for "400 watts @ 20 Hz" I'm looking at a 2 channel amp with 500WPC. I have no use for such an amp and no speakers capable of eating that power, but I would do it just because I can. Its right up there with "why do you need 500 horsepower when you sit in rush hour traffic moving slower than walking speed." Yes, If I left work at 5:00 I could walk home faster than driving. It just happens to rain nearly every afternoon in the summer in South Florida.
The "proof of concept" as we called my prototypes at work is this. The final circuit will be something like this with either 2 or 3 UNSET Power Head boards per channel and one Universal Driver board. This test happened last December. The little SMPS in the upper left corner lights the 26HU5's and a big variable supply is set on 625 volts which is delivered to the AC polypropylene cap via some Radio Shack clip leads. Power output is 350 watts @ 0.817% THD and 525 watts @ 3.79% THD with visible clipping. There is a slight red glow in the tube plates in a dark room at the 350 watt level with none seen at full power (normal behavior and idle current dependent).
Attachments
The large double plate variety is the one on the far right. There are two separate LARGE plates, but the guts in the middle are the same as the others. You can see the beam plates. The ones I have quantity of are the other two, after a couple years of scrounging. A couple of other onset twosey oddballs. I thought that one on the right was an oddball till I got another, and both work so they’re a pair. I think I got the first one at Stan’s. The second came from a “mixed brands” lot. I was actually excited to get it, rather than saying WTF is this???
How many to parallel? 5 would deliver all the peak current needed with an unloaded supply of 900V, and be ok on dissipation with a full load supply of 700V. When I load test the final supply I’ll see if I want to bump it up to 6. Yes, a 12 tube amp. With 6 drivers, LTP, and front end - but those are all $3 mini 9 pin except the 6CG8 LTP which is a bit more. The smaller variety will work sustained with 4 pair of the 24-28 watt tubes. So naturally I started thinking about bias stability, especially considering te end use.
The circuit I’m planning isn’t anything “special”. Just your basic Mullard with individual k-followers driving the output tubes. A low impedance on g1 should help with the drift, and provide an easy load for the phase splitter despite driving the universe with it. Now why would anyone want or need to do this? My real hobby is PA. I have a 30kW SS PA, all of which only sees action once in a blue moon. But its a blast. I’m on that road to retirement (they don’t WANT me to quit but the end is in sight). But an all tube PA that’s big enough to do something real with has been on the bucket list. Considering this use, even if only occasional, a bias controller seems very prudent. Show stopping amp failures or shutdowns is a non-no.
How many to parallel? 5 would deliver all the peak current needed with an unloaded supply of 900V, and be ok on dissipation with a full load supply of 700V. When I load test the final supply I’ll see if I want to bump it up to 6. Yes, a 12 tube amp. With 6 drivers, LTP, and front end - but those are all $3 mini 9 pin except the 6CG8 LTP which is a bit more. The smaller variety will work sustained with 4 pair of the 24-28 watt tubes. So naturally I started thinking about bias stability, especially considering te end use.
The circuit I’m planning isn’t anything “special”. Just your basic Mullard with individual k-followers driving the output tubes. A low impedance on g1 should help with the drift, and provide an easy load for the phase splitter despite driving the universe with it. Now why would anyone want or need to do this? My real hobby is PA. I have a 30kW SS PA, all of which only sees action once in a blue moon. But its a blast. I’m on that road to retirement (they don’t WANT me to quit but the end is in sight). But an all tube PA that’s big enough to do something real with has been on the bucket list. Considering this use, even if only occasional, a bias controller seems very prudent. Show stopping amp failures or shutdowns is a non-no.
Attachments
"The large double plate variety is the one on the far right. There are two separate LARGE plates, but the guts in the middle are the same as the others."
I have some of those. I think that they are 6LW6s. I guess I never looked at them close enough to see that there were two separate plates. I moved all of the valuable tubes over to an old house trailer on the property when the Radon guys were in the basement busting concrete. The 'LW6s are still in there.
"Just your basic Mullard with individual k-followers driving the output tubes. A low impedance on g1 should help with the drift, and provide an easy load for the phase splitter despite driving the universe with it."
The engineer in me would choose a mosfet follower, but that's what we were taught in the "mission critical" two way radio division at Motorola where I spent most of my career. There it was "choose the best possible part for the task regardless of cost." I did three years in the cell phone division where it was "choose the cheapest thing that will last through the warranty period."
"I’m on that road to retirement (they don’t WANT me to quit but the end is in sight)."
Motorola was in a downsizing mode for most of the last 20 years I was there. I was in a research group that "defined the state of the art in mission critical radio technology" and I had become one of only two people who built prototypes and EVBs. The "simulating and calculating" engineers thought that building real stuff was beneath them. I actually liked making stuff and someone needed to do it. When the "big boys from Schamburg" came to town my bench was often one of the first places they came to since that's where you could actually SEE some of the new tech alive and working. After Mot was broken up and parted out, we just couldn't call up the Semiconductor Products group and have a specialized chip designed for a product with a 10K per year run rate as the Semi group was busy making chips for phones that ran at 1 MILLION units a MONTH! We had to design our own chips and have them fabbed by an outside vendor. I made EVBs and real radios with those chips. When the Chicago boys were in town, I usually had an EVB hooked up to a speaker and dialed in to the local police department.
Then one day in late 2013 I got THE letter from the CEO. Short version......I was too old and made too much money for a guy who "just built stuff." I could take the money and leave, or I could take the boot print on my butt and leave. My boss assured me that it was a bluff just like the previous bluffs. I didn't feel good about it, so in early 2014 I took the money and left. They did cut about 300 people at the end of 2013, but I would have been safe. The whole working environment has changed so that most of the people I worked with either hate their job or have left.
I don't know what I will do with a 1KW tube amp but I will likely design it so that it can be run on two tubes per channel and a reduced B+ voltage for lower power use.
I have some of those. I think that they are 6LW6s. I guess I never looked at them close enough to see that there were two separate plates. I moved all of the valuable tubes over to an old house trailer on the property when the Radon guys were in the basement busting concrete. The 'LW6s are still in there.
"Just your basic Mullard with individual k-followers driving the output tubes. A low impedance on g1 should help with the drift, and provide an easy load for the phase splitter despite driving the universe with it."
The engineer in me would choose a mosfet follower, but that's what we were taught in the "mission critical" two way radio division at Motorola where I spent most of my career. There it was "choose the best possible part for the task regardless of cost." I did three years in the cell phone division where it was "choose the cheapest thing that will last through the warranty period."
"I’m on that road to retirement (they don’t WANT me to quit but the end is in sight)."
Motorola was in a downsizing mode for most of the last 20 years I was there. I was in a research group that "defined the state of the art in mission critical radio technology" and I had become one of only two people who built prototypes and EVBs. The "simulating and calculating" engineers thought that building real stuff was beneath them. I actually liked making stuff and someone needed to do it. When the "big boys from Schamburg" came to town my bench was often one of the first places they came to since that's where you could actually SEE some of the new tech alive and working. After Mot was broken up and parted out, we just couldn't call up the Semiconductor Products group and have a specialized chip designed for a product with a 10K per year run rate as the Semi group was busy making chips for phones that ran at 1 MILLION units a MONTH! We had to design our own chips and have them fabbed by an outside vendor. I made EVBs and real radios with those chips. When the Chicago boys were in town, I usually had an EVB hooked up to a speaker and dialed in to the local police department.
Then one day in late 2013 I got THE letter from the CEO. Short version......I was too old and made too much money for a guy who "just built stuff." I could take the money and leave, or I could take the boot print on my butt and leave. My boss assured me that it was a bluff just like the previous bluffs. I didn't feel good about it, so in early 2014 I took the money and left. They did cut about 300 people at the end of 2013, but I would have been safe. The whole working environment has changed so that most of the people I worked with either hate their job or have left.
I don't know what I will do with a 1KW tube amp but I will likely design it so that it can be run on two tubes per channel and a reduced B+ voltage for lower power use.
“The engineer in me would choose a mosfet follower”….
I bought a tube (ha ha pun intended) of FPQF5N40 specifically for that purpose. Can’t decide whether No Sand In The Signal Path will be a directive. If so it will be 6BQ7 drivers. Got loads of them, use them everywhere. And I get a better match between triodes than ANY new production AU7/AX7. Mu is smack between the two, with a low Rp.
Sand in the power supply and supervisor circuits will be a necessity. If I’m using a bias servo a semi-regulated g2 becomes possible since g1/g2 gain doesn’t have to track perfectly anymore. So I can have g2 track B+ as it falls under load. Mosfet driving an MJL4281, off a 210V raw supply tap. Voltage divider from B+ with additional ripple filter.
I bought a tube (ha ha pun intended) of FPQF5N40 specifically for that purpose. Can’t decide whether No Sand In The Signal Path will be a directive. If so it will be 6BQ7 drivers. Got loads of them, use them everywhere. And I get a better match between triodes than ANY new production AU7/AX7. Mu is smack between the two, with a low Rp.
Sand in the power supply and supervisor circuits will be a necessity. If I’m using a bias servo a semi-regulated g2 becomes possible since g1/g2 gain doesn’t have to track perfectly anymore. So I can have g2 track B+ as it falls under load. Mosfet driving an MJL4281, off a 210V raw supply tap. Voltage divider from B+ with additional ripple filter.
Off topic: the people who designed the 350 kW valve amplifier for the AM radio transmitter mentioned in this post: https://www.diyaudio.com/community/threads/350-kw-valve-amplifier-from-1934.399411/ just paralleled two complete 175 kW amplifiers, so one of them would keep working when the other failed.
Back on topic: a bit over five years ago, an eldery Belgian gentleman called Anne published an idea for a class-B or -AB bias servo for valve amplifiers on a Dutch audio forum and several forum members helped refining the circuit. Its concept is quite simple: a current sensing resistor in the cathode of each valve and a slow loop (with an integrator + extra pole) for each individual valve that controls the voltage across its current sensing resistor to one silicon diode forward voltage drop. To keep the loop from misbehaving when the valve is driven hard, the signal going into the integrator is clipped to two silicon diode forward voltage drops. If you do your best, you can come up with weird signals that would cause the loop to misbehave anyway if the ratio of the peak current to the quiescent current were very large for a long time, but it works fine with practical music signals. Many forum members use it.
This links to a schematic of an early version:
https://zelfbouwaudio.nl/forum/download/file.php?id=135953&mode=view
This opening post summarizes the variants that were built and tested:
https://zelfbouwaudio.nl/forum/viewtopic.php?t=9507&hilit=biasautomaat
Back on topic: a bit over five years ago, an eldery Belgian gentleman called Anne published an idea for a class-B or -AB bias servo for valve amplifiers on a Dutch audio forum and several forum members helped refining the circuit. Its concept is quite simple: a current sensing resistor in the cathode of each valve and a slow loop (with an integrator + extra pole) for each individual valve that controls the voltage across its current sensing resistor to one silicon diode forward voltage drop. To keep the loop from misbehaving when the valve is driven hard, the signal going into the integrator is clipped to two silicon diode forward voltage drops. If you do your best, you can come up with weird signals that would cause the loop to misbehave anyway if the ratio of the peak current to the quiescent current were very large for a long time, but it works fine with practical music signals. Many forum members use it.
This links to a schematic of an early version:
https://zelfbouwaudio.nl/forum/download/file.php?id=135953&mode=view
This opening post summarizes the variants that were built and tested:
https://zelfbouwaudio.nl/forum/viewtopic.php?t=9507&hilit=biasautomaat
I wouldn't use silicon in the signal path of a "tube amp" in any configuration other than a follower. The most important parameter in any follower to achieve low distortion is Gm, and even a crummy mosfet beats all vacuum tubes.
I have used 6BQ7s in LTPs where the 2nd harmonic distortion caused by the somewhat variable Mu cancels out. I never tried them as a follower. My usual tube for a follower is a 5687 due to its high Gm.
You don’t need a crazy amount of gm as a follower driving a power tube. It’s input impedance is pretty high and mostly you’re driving the tail resistor (20-30k, which is all you want in the g1 circuit). Look at all those 12BH7’s out there. Variable mu or no variable mu that 6BQ7 is more linear than a 12AU7. When you want to drive a low Z load follower-strap a video amp pentode. 50-ish ohm Z-outs are possible. Complete overkill here, but not in the “drive rack”… which is also a work in progress.
12AU7s are common, somewhat cheap, can handle 60 mA peaks, and their nonlinearity produces mostly second harmonic which some people like, and 2H in the driver will partially cancel 2H in the output stage possibly lowering the overall THD of the amp......That's what being commonly said. My testing with the SSE shows that cancellation is partial and is not constant VS frequency and output level. The 12AT7 that I use in the SSE is a VHF TV tuner tube not unlike the 6BQ7. It works in the SSE for two reasons, the partial cancellation previously mentioned, and the CCS load which helps the linearity a lot. Most of my work with my Universal Driver board has used 6CG7's, though the board will accept any 9A and 9AJ tubes. The 6CG7 is a shrunken 6SN7.
Yeah, 5687s do get hot, especially if you run them near the maximum plate dissipation (7 watts total) with nearly 5 watts burned in the heater. Whatever driver I use will be looking into the P channel mosfet follower under the cathode in the UNSET output stage.
Having to move virtually everything in the basement at least twice has come with a lot of headaches and lots of misplaced stuff, but I did uncover a misplaced unfinished project that will soon become useful if I can find all of the parts. I had started on a 100 to 150 WPC amp with Edcor 100 watt OPT's and a big toroid for power stuffed into a custom chassis. Looks like this has been dormant for about 5 years.
Yeah, 5687s do get hot, especially if you run them near the maximum plate dissipation (7 watts total) with nearly 5 watts burned in the heater. Whatever driver I use will be looking into the P channel mosfet follower under the cathode in the UNSET output stage.
Having to move virtually everything in the basement at least twice has come with a lot of headaches and lots of misplaced stuff, but I did uncover a misplaced unfinished project that will soon become useful if I can find all of the parts. I had started on a 100 to 150 WPC amp with Edcor 100 watt OPT's and a big toroid for power stuffed into a custom chassis. Looks like this has been dormant for about 5 years.
