I've been thinking... I think I may modify my screen-drive amp project.
I see potential, for a VERY LOW COST PP screen-drive amp, with maybe 50w/ch or so output.
Here's my thoughts so far:
CRITICAL PART #1: Output transformers- since it's screen-drive, the bias currents in the output tubes will be VERY LOW. Hence, any DC imbalances between the tubes will also be very low, in proportion. This leads me to toroidal output transformers... I was looking at various options, and there's an Antek power transformer... 230 (2x 115V) to 9V (two parallel secondary windings), 100VA. That gives, if my math is right, about 5.2K to 8 ohms. Best part, is price is only about $18 each.
Critical part 2: PowerDrive module, courtesy of TubeLab. Yes, I'm going over to the dark side... but to do this in any sort of economical manner, seems to warrant solid-state for the current driver (source follower) driving the G2s of the output tubes...
Other than that, the amp will probably be a fairly "conventional" Williamson design... input amp, cathodyne phase splitter, diff amp. I'm thinking something like a 12AT7 or 6DJ8 as an input tube (maybe a 6N1P?), and a 6SN7GTB-class tube (something that can handle at least 450V on the plate) as the diff amp. Should be able to swing enough voltage, especially driving a MOSFET follower.
I am thinking of using plate-to-plate feedback from the output tubes to the diff amp, as a way of reducing the necessary loop feedback over the whole amp. Should improve stability, I would think.
As for output tubes- I'm looking for something that's STILL in production somewhere... which kind of limits me to tubes like the EL590S (expensive) or maybe the 6146B (in production, and MUCH cheaper).
Anyone care to look over the specs of the 6146B and let me know if it would be ostensibly suitable? I'm thinking B+ of somewhere around 475-500V or so... that's a bit low for a 6146, but I guess I could bump it up a bit if needed. Would like to keep it under 500V if possible...
So, any thoughts? I'm thinking that this MIGHT be do-able for less than $400 total in parts...
Thanks for any musings of any sort, as I descend into potential madness...
Regards,
Gordon.
I see potential, for a VERY LOW COST PP screen-drive amp, with maybe 50w/ch or so output.
Here's my thoughts so far:
CRITICAL PART #1: Output transformers- since it's screen-drive, the bias currents in the output tubes will be VERY LOW. Hence, any DC imbalances between the tubes will also be very low, in proportion. This leads me to toroidal output transformers... I was looking at various options, and there's an Antek power transformer... 230 (2x 115V) to 9V (two parallel secondary windings), 100VA. That gives, if my math is right, about 5.2K to 8 ohms. Best part, is price is only about $18 each.
Critical part 2: PowerDrive module, courtesy of TubeLab. Yes, I'm going over to the dark side... but to do this in any sort of economical manner, seems to warrant solid-state for the current driver (source follower) driving the G2s of the output tubes...
Other than that, the amp will probably be a fairly "conventional" Williamson design... input amp, cathodyne phase splitter, diff amp. I'm thinking something like a 12AT7 or 6DJ8 as an input tube (maybe a 6N1P?), and a 6SN7GTB-class tube (something that can handle at least 450V on the plate) as the diff amp. Should be able to swing enough voltage, especially driving a MOSFET follower.
I am thinking of using plate-to-plate feedback from the output tubes to the diff amp, as a way of reducing the necessary loop feedback over the whole amp. Should improve stability, I would think.
As for output tubes- I'm looking for something that's STILL in production somewhere... which kind of limits me to tubes like the EL590S (expensive) or maybe the 6146B (in production, and MUCH cheaper).
Anyone care to look over the specs of the 6146B and let me know if it would be ostensibly suitable? I'm thinking B+ of somewhere around 475-500V or so... that's a bit low for a 6146, but I guess I could bump it up a bit if needed. Would like to keep it under 500V if possible...
So, any thoughts? I'm thinking that this MIGHT be do-able for less than $400 total in parts...
Thanks for any musings of any sort, as I descend into potential madness...
Regards,
Gordon.
I'm completely with SY on the sweep tube recommendation, and for his stated reasons. Now I'll add some more of my own:
Don't know of anyone using screen drive with the 6146B, but this is quite commonly done with sweep tubes.
Using sweep tubes will also allow you to stay comfortably under your 500V desired supply limit.
I would think tubelab's powerdrive topology would be pretty ideally suited to screen drive applications as well. The mosfet driver can do all the heavy lifting required to drive the screens leaving you to choose any tube you like for the voltage amplification chores.
Quiescent current can be quite low easing concerns about dc imbalance induced core saturation. (power toroid opts) You can always servo the bias voltage to the control grid later if this proves to be an issue, just make sure that the servo is really good at ignoring signal induced fluctuations in cathode voltage.
Matching the output tubes for transconductance at several different operating currents would probably help the ac balance, further improving linearity. (Probably doing something like fixing the control grid bias and varying the screen grid voltage and comparing those results tube to tube.)
I'm assuming that you will use at least some limited overall loop feedback to lower output impedance and further improve linearity.
If I were still a pushpull dude I would be pretty excited at the opportunity this project provides for learning..
I am sure George (tubelab) will find this thread shortly and weigh in with some really good recommendations on readily obtainable and inexpensive sweep tubes for your amplifier. The real issue is not whether or not the tube you want to use is still in production, but whether or not it is currently affordable and likely to be available in the future - just because it is currently available does not mean the preceding criteria are met..
Don't know of anyone using screen drive with the 6146B, but this is quite commonly done with sweep tubes.
Using sweep tubes will also allow you to stay comfortably under your 500V desired supply limit.
I would think tubelab's powerdrive topology would be pretty ideally suited to screen drive applications as well. The mosfet driver can do all the heavy lifting required to drive the screens leaving you to choose any tube you like for the voltage amplification chores.
Quiescent current can be quite low easing concerns about dc imbalance induced core saturation. (power toroid opts) You can always servo the bias voltage to the control grid later if this proves to be an issue, just make sure that the servo is really good at ignoring signal induced fluctuations in cathode voltage.
Matching the output tubes for transconductance at several different operating currents would probably help the ac balance, further improving linearity. (Probably doing something like fixing the control grid bias and varying the screen grid voltage and comparing those results tube to tube.)
I'm assuming that you will use at least some limited overall loop feedback to lower output impedance and further improve linearity.
If I were still a pushpull dude I would be pretty excited at the opportunity this project provides for learning..
I am sure George (tubelab) will find this thread shortly and weigh in with some really good recommendations on readily obtainable and inexpensive sweep tubes for your amplifier. The real issue is not whether or not the tube you want to use is still in production, but whether or not it is currently affordable and likely to be available in the future - just because it is currently available does not mean the preceding criteria are met..
Look at this thread for the low down on cheap tubes. I think the 6BQ6 for 98 cents qualifies as cheap, and 50 watts per pair in screen drive is possible.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=128533
I have a pair of genuine Plitron toroidal OPT's. These OPT's are specified for use at 400 watts and 20 Hz! They are huge.
Yes the DC offset is very low at idle because the current is low. That is only half of the problem. As the output power and hence the tube current increases the average current through the output tubes is unequal because no two tubes are ever perfectly matched. I have observed some low frequency saturation effects at reasonable power levels and frequencies. (30 watts at 30 Hz).
I will eventually figure out how to tame these offsets, and I will post the results. For now I am still figuring out some unique ways to connect the sweep tubes that I got on sale.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=128533
I have a pair of genuine Plitron toroidal OPT's. These OPT's are specified for use at 400 watts and 20 Hz! They are huge.
Yes the DC offset is very low at idle because the current is low. That is only half of the problem. As the output power and hence the tube current increases the average current through the output tubes is unequal because no two tubes are ever perfectly matched. I have observed some low frequency saturation effects at reasonable power levels and frequencies. (30 watts at 30 Hz).
I will eventually figure out how to tame these offsets, and I will post the results. For now I am still figuring out some unique ways to connect the sweep tubes that I got on sale.
"CRITICAL PART #1: Output transformers......100VA"
You better check the primary inductance on those xfmrs. The larger the VA rating, the lower the inductance:
2X core area -> 1/2 the turns for a given voltage
2X core area and 1/2 the turns -> L = k* Area* turns^2 -> 1/2 the inductance. 10X core area -> 1/10 inductance
Nice 100VA tweeter amp.
I'm waiting to see the Antek tube output xfmr. pricing.
By the way, the 6146B has g2/g1 Mu of 4.5, surprisingly similar to the horiz. output tubes (usually 3.0 to 4.2 range). So g2 drive should be workable for this tube, but at $15 to $20 a tube it appears.
Don
You better check the primary inductance on those xfmrs. The larger the VA rating, the lower the inductance:
2X core area -> 1/2 the turns for a given voltage
2X core area and 1/2 the turns -> L = k* Area* turns^2 -> 1/2 the inductance. 10X core area -> 1/10 inductance
Nice 100VA tweeter amp.
I'm waiting to see the Antek tube output xfmr. pricing.
By the way, the 6146B has g2/g1 Mu of 4.5, surprisingly similar to the horiz. output tubes (usually 3.0 to 4.2 range). So g2 drive should be workable for this tube, but at $15 to $20 a tube it appears.
Don
The Antek 100VA power toroids are maybe 64 Hy, the 50 VA possibly twice that according to their data sheets. But since they'll saturate at about 300V at 60 Hz, or 100V at 20 Hz, they won't support much low-frequency power.
I've been thinking about Crowhurst's twin-coupled circuit with toroids...
I've been thinking about Crowhurst's twin-coupled circuit with toroids...
tubelab.com said:
I'm thinking that plate-to-plate (output to diff amp) feedback might help this... but I don't know if it would reduce the effects you mention to a low enough level for it not to be an issue. Seemed to help on both ends of the spectrum on a Heathkit W4 amp (which also seems to have marginal output transformer capacity, in some respects), according to what I've seen in a few places...
Apparently, Bob Carver is working on this type of thing, if I correctly read what he's writing recently. He's talked about a "DC restorer" circuit... sounds suspiciously like what we're talking about here...
Regards,
Gordon.
smoking-amp said:
That's what I was looking at. The 200V max screen voltage rating led me to think that the gain of G2 might be in the right ballpark...
And, $15-20 per tube is OK, for a "production" amp of 50w/ch, I'd think. Not much different than some EL34s or 6L6s, and less than most KT88s...
Now, to get this transformer thing licked...
Regards,
Gordon.
hey-Hey!!!,
I'd suggest building a smaller one first. George's 6BQ6 is a good place to start. Getting the topology together and getting a handle on its workings is probably a good idea. There are lots of good sized sweeps in the size you'd need for 50 Watts. I am thinking a 17W plate would do just fine and those are really easy to get in adequate qty.
cheers,
Douglas
I'd suggest building a smaller one first. George's 6BQ6 is a good place to start. Getting the topology together and getting a handle on its workings is probably a good idea. There are lots of good sized sweeps in the size you'd need for 50 Watts. I am thinking a 17W plate would do just fine and those are really easy to get in adequate qty.
cheers,
Douglas
The pinout of the 6146B is close to a lot of the octal horiz. output tubes. Just two pins need swapping. Could put a jumper pair on the board to select.
The plate size of the 6146B is exactly the same as the 17 Watt HO tubes anyway. Apparently different Watt rating systems are used for HO duty and other services. An example of this is the 6KV6A rated at 28 Watts for shunt regulator duty, and the virtually identical 6KM6 rated at 20 Watts for HO duty.
"I've been thinking about Crowhurst's twin-coupled circuit with toroids..."
Me too, just the thing for getting twice the inductance and voltage rating at LF. But screen grid drive will be a NASA project that way.
OT: got gasoline? none left in Hickory tonight.
Don
The plate size of the 6146B is exactly the same as the 17 Watt HO tubes anyway. Apparently different Watt rating systems are used for HO duty and other services. An example of this is the 6KV6A rated at 28 Watts for shunt regulator duty, and the virtually identical 6KM6 rated at 20 Watts for HO duty.
"I've been thinking about Crowhurst's twin-coupled circuit with toroids..."
Me too, just the thing for getting twice the inductance and voltage rating at LF. But screen grid drive will be a NASA project that way.
OT: got gasoline? none left in Hickory tonight.
Don
I have a few more experiments to try besides screen drive. That is why I got a good stash of 98 cent and $2 sweep tubes. I don't worry about accidentally blowing them up. I would hesitate to try something new with a $15 tube.
My first experiment in "dual drive" resulted in two blown mosfets, one severed in half. I am not exactly sure what caused the exploded fets, but they cost more than the tubes, so I will have more meters and scope probes hooked up next time. The 100 uF motor run cap stores about 30 Joules of energy, more than enough to blow mosfets in half!
I plan to revisit screen drive, "dual drive" and the cathode follower before choosing a topology for a big amp. All three have advantages and drawbacks. More experiments as time permits.
Media induced panic has made supplies scarce and prices up over $4 for regular. Sherri is up north again and reports similar prices.
OK, "dual drive" part two, the sequel. Same setup, 6BQ6GTA's (the 98 cent variety). This time I put seperate stopper resistors on each grid. I used 1/2 watt 120 ohm carbon comp resistors hoping that they would act as fuses maybe saving the mosfets. I fired it up, and slowly increased the plate voltage with one hand while controlling the grid (G1 + G2) voltage with the other.
Fortunately the screen resistors were indeed sacrificed to save my mosfets and the culprit is extreme oscillation. This thing will wipe out the TV and make my computer speakers go crazy for about two seconds before the resistors evaporate.
I get the same results with the fat 6BQ6's and 6DQ6's. Different value screen resistors, same result. Different OPT, same result. Disconnecting the driver board from the mosfets, same result.
For now, no more "dual drive" experiments. I have a good size pile of toasted resistors, most blown in half. I have made "dual drive" work before with a 6L6. The "dual drive" and screen drive circuits both have pentode like output impedance, requiring a sizeable dose of NFB to lower.
I decided to experiment with a P-P cathode follower output stage. The obvious drawback here in the extreme drive voltage requirement. I am running the 98 cent 6BQ6GTB's into a 625 ohm CT OPT. Power output is 15 watts and is limited by available drive voltage. The tubes are wired in triode mode. Idle current is 10 mA per tube, with a plate voltage of 250 volts. The distortion remains below 2% until clipping. The output impedance is low. The output voltage rises about 50% when the load is removed. Don't do this on a conventional tube amp. It may blow up.
Tomorrow I will dig up some cheap toroids and try some of them.
Tom B.
"The Antek 100VA power toroids are maybe 64 Hy, the 50 VA possibly twice that according to their data sheets. But since they'll saturate at about 300V at 60 Hz, or 100V at 20 Hz, they won't support much low-frequency power."
I was looking at the Antek datasheet and noticed your 64 Hy figure apparently comes from the idle current measurement at line voltage. This is a measure of near (actually beyond the peak) peak inductance at max signal. But for small signal conditions the L will be much smaller.
I have a 130 VA Avel- Lindberg toroid here, so I measured it with line voltage excitation (actually at 50% line to get nearer to the peak) and again with an L meter. Results for 2X115V windings are: 185.6Hy at line excitation and 6.6 Hy with the L meter at 120 Hz. Thats a factor of 28 increase.
Out of curiosity I then went and measured the full primary on a Hammond 1650T OT (1900 Ohm P-P):
284 Hy at 70% of max voltage rating (60 Watt level, close to the Mu peak) and 8.5 Hy with the L meter. Thats a factor of 33.5 increase.
The Antek probably similarly has around 64/28 = 2.3 Hy at small signal level. For a 5.2K Ohm primary at 20 Hz, one would like at least 2 pi f L = 5200 or Lmin. = 41.3 Hy.
Checking the results for the Hammond 1650T, 2 pi f L = 1900 or L min. = 15 Hy. So 8.5 Hy is a little low, but then the L meter signal is very small. Using the same "fudge" factor for the Antek => 41.3x(8.5/15) => 23.4 Hy. So the Antek 100 VA is still looking too low for pri. L by 10x. A 10 or 20 VA xfmr should work about right.
Don
"The Antek 100VA power toroids are maybe 64 Hy, the 50 VA possibly twice that according to their data sheets. But since they'll saturate at about 300V at 60 Hz, or 100V at 20 Hz, they won't support much low-frequency power."
I was looking at the Antek datasheet and noticed your 64 Hy figure apparently comes from the idle current measurement at line voltage. This is a measure of near (actually beyond the peak) peak inductance at max signal. But for small signal conditions the L will be much smaller.
I have a 130 VA Avel- Lindberg toroid here, so I measured it with line voltage excitation (actually at 50% line to get nearer to the peak) and again with an L meter. Results for 2X115V windings are: 185.6Hy at line excitation and 6.6 Hy with the L meter at 120 Hz. Thats a factor of 28 increase.
Out of curiosity I then went and measured the full primary on a Hammond 1650T OT (1900 Ohm P-P):
284 Hy at 70% of max voltage rating (60 Watt level, close to the Mu peak) and 8.5 Hy with the L meter. Thats a factor of 33.5 increase.
The Antek probably similarly has around 64/28 = 2.3 Hy at small signal level. For a 5.2K Ohm primary at 20 Hz, one would like at least 2 pi f L = 5200 or Lmin. = 41.3 Hy.
Checking the results for the Hammond 1650T, 2 pi f L = 1900 or L min. = 15 Hy. So 8.5 Hy is a little low, but then the L meter signal is very small. Using the same "fudge" factor for the Antek => 41.3x(8.5/15) => 23.4 Hy. So the Antek 100 VA is still looking too low for pri. L by 10x. A 10 or 20 VA xfmr should work about right.
Don
Toroid thoughts
Any of these line toroids will saturate at about 100-120V at 20 Hz.
The DCR of 15VA units is about 150 ohms (series of course) and
25VA units about 80 ohms.
Since the OPT will be voltage-limited that suggests that the increased
power capability of the larger core is wasted. The DCR of the smaller
toroids is plenty low, and as Don points out the inductance is higher
with the smaller toroids.
It mainly depends on the anode impedance you're working at. I
think for a 6080 amp or CF output stage the higher power toroids
might make sense, given 200 ohms Za into say 800 ohm load you're
looking at 18 watts and might want a 50-100 VA toroid.
With a 2A3 you probably want 2K-3K load so that's 4-5 watts
at the 110VRMS/20Hz saturation; a 25VA toroid might be great.
I'm having no problem driving a 25VA toroid with Za = 800 ohms
down to 10Hz.
I agree here you probably want to use the smallest core you can
get away with to minimize the activation energy of the core.
Michael
Any of these line toroids will saturate at about 100-120V at 20 Hz.
The DCR of 15VA units is about 150 ohms (series of course) and
25VA units about 80 ohms.
Since the OPT will be voltage-limited that suggests that the increased
power capability of the larger core is wasted. The DCR of the smaller
toroids is plenty low, and as Don points out the inductance is higher
with the smaller toroids.
It mainly depends on the anode impedance you're working at. I
think for a 6080 amp or CF output stage the higher power toroids
might make sense, given 200 ohms Za into say 800 ohm load you're
looking at 18 watts and might want a 50-100 VA toroid.
With a 2A3 you probably want 2K-3K load so that's 4-5 watts
at the 110VRMS/20Hz saturation; a 25VA toroid might be great.
I'm having no problem driving a 25VA toroid with Za = 800 ohms
down to 10Hz.
I agree here you probably want to use the smallest core you can
get away with to minimize the activation energy of the core.
Michael
Going to a SEPP setup, or a Circlotron setup, could gain 4X on the inductance issue and 2X on the tube voltage swing limit. The Crowhurst twin-coupled circuit gains 2X on the inductance factor I think, and doubles the voltage limit too.
I've been looking at an interesting circuit that avoids the doubled B+ of SEPP and the floating B+'s of the Circlotron. Using the two complementary drive outputs from an asymmetric pentode/triode LTP: a Mosfet operates in follower mode, driving the low Z secondary in one direction (gate driven by bootstrapped pentode plate load) and the high Z primary side is driven in the opposite direction by a triode output tube's plate (it's grid driven by the triode plate half of the LTP).
Local feedback to the LTP pentode's screen gives the Mosfet a matched effective Mu for tracking the triode output's Mu. Bootstrapping or a CCS can be used for loading the LTP's pentode plate.
Version #2 uses a pentode tube in "apparent follower" mode for the high Z output side, and it's "Mu" gets set similarly to the Mosfet by local feedback to the second screen grid in a now double pentode LTP this time. (Can't find my scanner at the moment)
Don
I've been looking at an interesting circuit that avoids the doubled B+ of SEPP and the floating B+'s of the Circlotron. Using the two complementary drive outputs from an asymmetric pentode/triode LTP: a Mosfet operates in follower mode, driving the low Z secondary in one direction (gate driven by bootstrapped pentode plate load) and the high Z primary side is driven in the opposite direction by a triode output tube's plate (it's grid driven by the triode plate half of the LTP).
Local feedback to the LTP pentode's screen gives the Mosfet a matched effective Mu for tracking the triode output's Mu. Bootstrapping or a CCS can be used for loading the LTP's pentode plate.
Version #2 uses a pentode tube in "apparent follower" mode for the high Z output side, and it's "Mu" gets set similarly to the Mosfet by local feedback to the second screen grid in a now double pentode LTP this time. (Can't find my scanner at the moment)
Don
OK, I am taking a time out while the smoke clears, but I am "testing" this right now. I have a cathode follower output stage using the 98 cent 6BQ6GTB's. Last night I was getting about 15 watts with very low distortion using a huge Plitron toroidal OPT that is rated for 400 watts at 20 Hz, 1250 ohms.
Today I collected some surplus power toroids and connected one up. It has 18 and 24 volt secondaries and a 230 volt CT primary. My guess is that it is about 300 VA. I was getting 30 to 35 watts at really low distortion. Under 1% at 20 watts climbing to 2.4% at 30 watts. This thing had flat power bandwidth. The distortion was 3.3% at 30 watts and 20 Hz! No sign of the saturation monster anywhere.
While marvelling at the high power output and low distortion I didn't see that the current meter was burried. I started to sense that something was hot, really hot. It was the output tubes, glowing brightly. I am currently trying some different tubes.
This shows that even a largish power toroid can work if it is fed from a low enough impedance. This is why I am temporarilly avoiding screen drive.
Top secret power toroid tip. Find a 230 to 230 volt isolation transformer. Wire the primary and secondary in series to form a 460 volt center tapped winding (more inductance) and wind a new secondary over this.
It just occured to me that when operating in Circlotron mode, the center of the primary is at ground potential. So one could connect two of these toroid (or any) xfmrs in series (total of 4 x 120V windings) with the center tap grounded. This would give 8x the inductance and 4x the primary voltage range over the usual simple center tapped P-P scheme. It still requires two floating B+ supplies.
One really needs to know how the xfmrs are wound though. Some (probably most), will have the two 120V windings co-wound (loosely bifilar), causing very large capacitance issues. One should do a capacitance test between the two primaries (each winding shorted).
The 230 V to 230 V isolation xfmr George suggested should give low distributed capacitance. But may be somewhat high on leakage L.
Seems to me that the way to go about this would be to use a bunch of small split bobbin xfmrs with a single 120V winding on one bobbin and a LV winding on the other. Then connect the 120V windings in series and the LV windings in parallel. Have to use SEPP or Circlotron to drive the single primary then. Or use the Asymmetric P-P drive I mentioned in post #17 (no floating supplies).
Don
One really needs to know how the xfmrs are wound though. Some (probably most), will have the two 120V windings co-wound (loosely bifilar), causing very large capacitance issues. One should do a capacitance test between the two primaries (each winding shorted).
The 230 V to 230 V isolation xfmr George suggested should give low distributed capacitance. But may be somewhat high on leakage L.
Seems to me that the way to go about this would be to use a bunch of small split bobbin xfmrs with a single 120V winding on one bobbin and a LV winding on the other. Then connect the 120V windings in series and the LV windings in parallel. Have to use SEPP or Circlotron to drive the single primary then. Or use the Asymmetric P-P drive I mentioned in post #17 (no floating supplies).
Don
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.