There's been some discussion in another thread regarding the Blumlein "Garter Belt" bias scheme that helps to correct output tube mismatching. Posted is a circuit that combines this scheme with an active cathode bias circuit, sacrificing some of the advantages of that bias circuit, but eliminating a bias balance pot - food for thought. Values are TBD, just the concept is presented for now.
Attachments
I sat down and did some thinking about what I needed for a Garter Belt" NLD bias circuit. Assuming the plate voltage for the 6CW5s is ariound 250 V, about 16V of negative bias is needed for a plate current of 40mA. I made an E-WAG and and decided that the Garter resistors (R6 and R7) would need to drop 20% of the 2.5V reference voltage of the TL431 at 40mA. This works out to a resistance of around 12 ohms. Bypass capacitors C1 and C2 need to be out of the way at audio frequencies of interest. I chose a corner frequency of 5 Hz. This gives me a requirement of ~2700uF. I have some 2200 uf/6.3V low ESR caps around, so I'll try those. The resulting schematic is shown. I'll try populating the circuit with some random 6CW5s and see how well it balances, and whether the TL431s will need additional compensation (some like to take off and oscillate at a couple of MHz, so that they give squirreley results while appearing to sort of work).
Attachments
Last edited:
Itested a breadboard of the "NLD Garter bias circuit yesterday, and caught a mistake in grounding that essentially nullified the "garter" part of the circuit. The circuit should look as shown here instead. The gounds on the TL431s are shifted so that they can "see" the voltage on the garter resistors. I've made the changes in my breadboard, and will try it again today.
Attachments
The circuit as modified worked nice, so it's full steam ahead. I need to tweak the values a little to accommodate 6CW5s that are not exactly like the average characteristics, but that's life in tube-land.
How would one change the values to work with a tube of a different type, say, EL84, 6V6, 6550? This silicon is just starting to make sense to me, but not so sure on logistics.
I'd hold on to your horses until I get this thing built, so I can test it with an actual audio signal. The 6CW5s used in the test were actually tolerably well-matched out of the box, so I want to try subbing some more tubes to get some that aren't matched as well. I have some 8CW5s of different brands, so I could try them while I'm bringing up the amp on bench supplies. I may or may not want to change the "garter" resistor values (R6 and R7) depending on the results of that testing.
But, to answer your question somewhat, you choose an operating bias current for your tubes and look up the required bias voltage from the typical curves. The divider resistors for each TL431 (R4-5, R8-9) are set for the proper voltage drop, taking into account the voltage dropped across the garter resistors. If you are unsure of how to calculate the resistor string, you may want to start with a simple cathode bias network as shown in the schematic in post 13 - easier to build and understand. The first thing you want to do, though, if you haven't done it already, is to download the spec sheet for the TL431 and familiarize yourself with how it operates. It's a variable shunt regulator, and the Darlington pass transistor is used to increase the amount of current it can handle. The cascode transistor cleans up some aspects of the transient response and increases the voltage that can be handled by the bias circuit, as the TL431 has a maximum operating voltage of 35 volts or so.
This is as much answer as I want to deliver at ~4 in the morning...
But, to answer your question somewhat, you choose an operating bias current for your tubes and look up the required bias voltage from the typical curves. The divider resistors for each TL431 (R4-5, R8-9) are set for the proper voltage drop, taking into account the voltage dropped across the garter resistors. If you are unsure of how to calculate the resistor string, you may want to start with a simple cathode bias network as shown in the schematic in post 13 - easier to build and understand. The first thing you want to do, though, if you haven't done it already, is to download the spec sheet for the TL431 and familiarize yourself with how it operates. It's a variable shunt regulator, and the Darlington pass transistor is used to increase the amount of current it can handle. The cascode transistor cleans up some aspects of the transient response and increases the voltage that can be handled by the bias circuit, as the TL431 has a maximum operating voltage of 35 volts or so.
This is as much answer as I want to deliver at ~4 in the morning...
Build is chugging along for this beast - it should be ready for BA 2010 if there aren't too many gotchas in the power-up. Case will be deep blue, with cast-iron gray side panels, and an orange neon switch in front - jazzed-up Bud box construction. Transformers and SMPS are inside, tubes on top.
I really did it to myself this time, starting with the NLD concept, which is not all that complicated assuming reasonably well-matched output tubes. Then I had to add the Blumlein "Garter" matching concept, which makes this project about twice as complicated. There is a word for this kind of stuff, and I think it's called "hubris", usually awarded a couple of extra lightning bolts up the wazoo by the Gods. I'm feeling a little scorched this evening, as Murphy joined the proceedings. Anyway, I think the "NLD plus Blumlein" concept will work out in the end, just in time for me to junk it in favor of a simpler approach involving source follower-fed output grids. Anyway, I'm building this NLD prject so I can show it to Stu at this year's Burning Amp (assuming he actually shows), and then it's on to another concept that's loads simpler, addresses the issues Stu handled with the original "RLD" amp and adds Blumlein "Garter" style matching. I hate it when I obsolete myself.
Anyway, I nailed myself big time this evening by constructing my NLD/garter bias boards using TLV431s instead of TL431s. For those not in the know, the TLV431 has a 1.25V reference voltage vs the TL431's 2.5V reference. This explains why my amp was drawing 1/2A of bias. Fixed that... Next up, the EI/Yugoslav 6CW5s I'm using are a little friskier than average, and need some more bias to calm them down. I'll fix that tomorrow, then debug the input stage.
Just as a heads up for those who are starting out - don't plan a project that uses not one, not two, but three new concepts all at once and not expect to spend a lot of time in debug, especially as the dumb-*ss mistakes need to get worked out before you can even begin to address any lurking conceptual flaws. Experienced hackers know that the Gods love to have a grin or two at the expense of the ambitious...
Anyway, I nailed myself big time this evening by constructing my NLD/garter bias boards using TLV431s instead of TL431s. For those not in the know, the TLV431 has a 1.25V reference voltage vs the TL431's 2.5V reference. This explains why my amp was drawing 1/2A of bias. Fixed that... Next up, the EI/Yugoslav 6CW5s I'm using are a little friskier than average, and need some more bias to calm them down. I'll fix that tomorrow, then debug the input stage.
Just as a heads up for those who are starting out - don't plan a project that uses not one, not two, but three new concepts all at once and not expect to spend a lot of time in debug, especially as the dumb-*ss mistakes need to get worked out before you can even begin to address any lurking conceptual flaws. Experienced hackers know that the Gods love to have a grin or two at the expense of the ambitious...
That's because I'm getting somewhere, albeit a bit more slowly that I planned. If I were still cornered, I might be a bit more snarly... I just successfully got my new preamp front panel drilled and primed, though I managed to drop the fresh-primed panel on the grimy basement floor first time around. Mineral spirits cleaned the mess off (still wet), and it's now out of harm's way, painted again, hanging up to dry overnight. I'll bake it tomorrow to really set up the primer, then do the finish coat, a very dark blue-green metallic paint.
Last edited:
In my case the EI 7189A's that I have are fakes. They are relabled Russian 6P14. (I may have guessed that Russian number wrong). They are OK at normal EL34 voltages but Murphy smiles from inside the tubes if 7189A voltages are applied to them. The give away is the sharp pins that seem to be found only on older Russian tubes.
In my case the gods seem to have an affinity for things like electrolytic caps.
I think the 6CW5s I have may be from EI. They have long plates and a couple of small radiators tacked to the screens. They also draw about the right amount of filament power. They are marked "6CW5" and "Yugoslavia". They have also stood up to a fair amount of abuse without complaining. I don't intend to abuse them to much, as I have B+ set at 275V, and the screen set at around 200.
This afternoon I spent 3 hours in the lab at work (a clean, well-lighted place, unlike my basement), and got a lot of debugging done on the NLD. Some resistor value changes increased the bias voltage for my 6CW5s, and put the output bias currents close to where I want them. Next up was a look at the input stages. These were actually more or less ok. I changed the cathode resistors on the 6AM4s to 82 ohms from 100 ohms so I could get a little less drain voltage for my specified drain current of 5ma. One of the tubes is weak, so I'll bring a bunch back to work tomorrow to find one that makes the grade - then it's applying a signal, closing the loop, and, and bringing up the amp with the SMPS. The big disadvantage of P-P amps compared to single-ended is that there's twice as many parts to debug...
That ol' hubris has got me - this amp may not make it to Burning Amp. Too many new concepts all at once:
1) the 6AM4 triode I used for input duty is nice on paper, but the tube in the flesh (at least the GE units I've tried so far) doesn't act too close to its published characteristics. Being designed for VHF/UHF grounded grid service, it's a slippery bastard to boot, and needs some extra effort to quiet it down, including selective use of ferrite beads.
2) Using a mosfet for phase splitter appears to be a good idea for applications with limited voltage headroom, but again, precautions need to be taken to avoid parasitic oscillation. This may mean nothing more than increasing the value for the gate stopper resistor.
3) The Baldwin transformers I used appear on paper to be a nice match for the 6CW5 output tubes, but I suspect they're not all that hot. If that's the case, I may do another version of this amp with 6P14P-EV outputs and transformers I know will make the grade, as I have some 6BQ5-type output transformer sets, one from a Fisher (or is it Scott?) amp (hefty), and one set from a Knight amp (a little lighter).
4) Adding the Blumlein "garter" style auto-matching to the NLD bias network doubled the complexity of this amp, as you need a bias circuit for each tube, instead of one handing both. Again, OK in concept, but debugging is a bitch, especially when your 6CW5s need ~20V grid bias instead of 16V to quiet them down. I may look up Stan (but not right away) to get some Real 6CW5s.
5) The right channel of this unit doesn't look to bad once the transformer phase is properly sorted out. However, I'm only getting about 15V out before clipping, which means the the outputs are only swinging ~145V with a 275V supply - this doesn't sound right. Needs more investigating.
Anyway, I'll keep on plugging on this amp in the background, but shift my emphasis to lower-hanging fruit, as BA looms.
1) the 6AM4 triode I used for input duty is nice on paper, but the tube in the flesh (at least the GE units I've tried so far) doesn't act too close to its published characteristics. Being designed for VHF/UHF grounded grid service, it's a slippery bastard to boot, and needs some extra effort to quiet it down, including selective use of ferrite beads.
2) Using a mosfet for phase splitter appears to be a good idea for applications with limited voltage headroom, but again, precautions need to be taken to avoid parasitic oscillation. This may mean nothing more than increasing the value for the gate stopper resistor.
3) The Baldwin transformers I used appear on paper to be a nice match for the 6CW5 output tubes, but I suspect they're not all that hot. If that's the case, I may do another version of this amp with 6P14P-EV outputs and transformers I know will make the grade, as I have some 6BQ5-type output transformer sets, one from a Fisher (or is it Scott?) amp (hefty), and one set from a Knight amp (a little lighter).
4) Adding the Blumlein "garter" style auto-matching to the NLD bias network doubled the complexity of this amp, as you need a bias circuit for each tube, instead of one handing both. Again, OK in concept, but debugging is a bitch, especially when your 6CW5s need ~20V grid bias instead of 16V to quiet them down. I may look up Stan (but not right away) to get some Real 6CW5s.
5) The right channel of this unit doesn't look to bad once the transformer phase is properly sorted out. However, I'm only getting about 15V out before clipping, which means the the outputs are only swinging ~145V with a 275V supply - this doesn't sound right. Needs more investigating.
Anyway, I'll keep on plugging on this amp in the background, but shift my emphasis to lower-hanging fruit, as BA looms.
I thinks one of my main problems is excessive plate voltage on the 6AM4 for the 5mA plate current I'm trying to get. This puts a crimp in the voltage headroom for the mosfet concertina splitter, causing it to limit early. All this is not helped at all by overly frisky 6CW5s that require at least 20V rather than 16V for proper bias at 250V plate. All this stuff adds up quick when your B+ supply is only 275V. I've given the matter some thought and have decided to go back to my original concept of differential input/phase splitter, and have already laid out and cut the boards for the same. These will have no headroom problems even driving the abnormal 6CW5s I have on hand. I will not seriously look at the new circuit, though, until after BA, as I have a couple of other projects I need to finish before then.
I'll save the 6AM4 for a situation (like a 6BQ5 amp) where I have more voltage headroom. As far as I can see, it'll have no problem delivering the goods in such a situation.
I'll save the 6AM4 for a situation (like a 6BQ5 amp) where I have more voltage headroom. As far as I can see, it'll have no problem delivering the goods in such a situation.
I was inspired by BA today to match up the JFETs for the new front end for the No Light, so the new input boards are almost all the way stuffed. The input circuit will look very much like that in the first circuit I posted for the NLD. I'm continuing the efort so that I can finish off the amp and hear how it sounds. The real prize that came out of this effort is a much simpler concept that combines source follower drive and Blumlein "Garter" tube bias compensation/matching. This will be reserved for another thread and a amp that will use 6P14P-EVs and some nice Scott 6BQ5 output transformers I have on hand.
I brought up the new circuit this evening today, and after fixing an obvious boo-boo, the quiescent currents in the input stage settled down to where they need to be. However, the thing really likes to oscillate. It was probably a mistake to use 2SK170s in the input diff amp - way too much gain. I have the choice of adding degeneration or whisking out the 2SK170s and substituting something less jumpy, like 2SK117s (I just got a bunch in) or my long time favorite, the PN4393.
Normally, I would pay a lot of attention to your suggestions, but that one makes me think of Sting, and therefore makes me think of ultraviolence, especially on his person. I could take him back then, but like Bono these days, he's too much in people's faces for no good reason...
Last edited:
Last weekend, I spent some frustrating hours with the No-Light, crowned by the decision of one of the Yugo 6CW5s to go red-plate above ~250V (my first red-plate event). The Blumlein garter circuit in this version of the NLD has a lot of gain, and if one tube goes bad and draws a lot of current, it pushes up the other tube quite a bit, and depresses the bias voltage - positive feedback in the worst way. One solution is to reduce the value of the current sense resistors in series with each tube to reduce the sensitivity and bias adjustment range, but this then requires a huge value for the bypass capacitors (try calculating the cap value required for a 5Hz roll-off with ~3.3 ohms resistance). The amp has been a little hard to stabilize as well.
After the red-plating started to occur, I got disgusted and decided to back off and cool my jets for a bit. Three decisions came from that quiet time - one, ditch the Yugo 6CW5s and get some real ones from Stan - they're on their way. Two, get rid of the Garter bias compensation for this edition of the NLD, as it doubles the circuit complexity and makes troubleshooting a real exercise in frustration. I've come up with some simpler bias boards based on this decision. They are all populated and waiting to be patched in once I get some real 6CW5s. The bias point for the real tubes will probably be a more reasonable 16V or so. Three, add local feedback resistors from output plates to input plates to provide a short feedback path for a good portion of the excess gain. Hopefully, this will make the loop easier to stabilize.
The 6BQ5/7189/6P14P-EV version of the NLD (quite a different beast) will include a far less touchy Blumlein Garter bias circuit. I'll post more about that when the project starts to come together. It'll use some output transformers that were reportedly scavenged from a Fisher X-100 amp.
After the red-plating started to occur, I got disgusted and decided to back off and cool my jets for a bit. Three decisions came from that quiet time - one, ditch the Yugo 6CW5s and get some real ones from Stan - they're on their way. Two, get rid of the Garter bias compensation for this edition of the NLD, as it doubles the circuit complexity and makes troubleshooting a real exercise in frustration. I've come up with some simpler bias boards based on this decision. They are all populated and waiting to be patched in once I get some real 6CW5s. The bias point for the real tubes will probably be a more reasonable 16V or so. Three, add local feedback resistors from output plates to input plates to provide a short feedback path for a good portion of the excess gain. Hopefully, this will make the loop easier to stabilize.
The 6BQ5/7189/6P14P-EV version of the NLD (quite a different beast) will include a far less touchy Blumlein Garter bias circuit. I'll post more about that when the project starts to come together. It'll use some output transformers that were reportedly scavenged from a Fisher X-100 amp.
- Status
- Not open for further replies.