Well, I did some more playing with the first stage on a B+ of a little over 200v (still no tripler). With a plate resistor of 22k and a cathode resistor of 270 ohms (1v bias), I end up with about 4mA (a drop of 90v across the 22k) and a plate voltage of 110v. Sounds good, but the gain has dropped dramatically. I haven't measured, but I can't audibly clip the amp with full volume. Before, I could clip at about 3/4 volume on the control.
I expected a gain drop with the reduced plate load, but this is far too much to be of any use, especially as there is no global feedback (which would drop the gain further).
Obviously, I am not going to get the gain and the current at the same time without a vast increase in B+.
As a comparison, I briefly tried 2v bias but it raised the plate to 135v which was stressing the Vkh of the next stage tube. But I did get some quick measurements:
Ia=3.4mA, Rk=750 ohms, B+ lifts to 209v.
I am still not quite certain which B+ I should take when doing load lines:
If I pull the input tubes I have 250v, if I only run the cathodyne stages I have about 230v, with all stages connected the B+ depends on bias but can be anywhere from 195v to 210v.
I wish I had some of the LR8N regulators, then I could lock a B+ for the pre-stages and take one variable out of the equations 😡
I'll be interested to see how your measurements compare with your theoretical calculations.
Gary
I'm not sure either.
What is the B+ ?
The voltage from the PS under load, or without load.
Could someone give a definitive answer please.
Gary, I don't think I will try for the extra current through the 6N1P until I get the MJ book which I have heard discusses the DC-coupled cathodyne at length (Bevois Valley amp). What I said in an earlier post about biasing the cathodyne isn't right, I was confused. What I have noticed so far, is that the voltage dropped across the anode and cathode resistors of the cathodyne is always similar, which makes sense as the same current flows through both under quiescent conditions. So the voltage on the grid (from the anode of the previous stage) biases the cathodyne, as you are already aware. If the voltage on the grid is too high, the anode and cathode resistors will have large voltage drops across them leaving very little voltage across the valve itself. For example, if I biased the first stage as per the loadline I drew in my previous post and it has say 164V on the anode, then on the cathodyne, there will be say 166V dropped across the cathode resistor and 166V across the anode resistor, total approx 330V, leaving only 20V across the tube (given a B+ of 350V). Don't think its going to work.
The B+ for doing loadlines needs to be the expected B+ with the amp running. In our case if we bias for more current in the driver stage then the B+ for the driver stage will drop a bit more voltage across the 22K resistor feeding the 33uF caps. I have already changed the 22K for 10K and thinking of fitting 6K8, not a huge change but every bit helps.
It seems the LED's keep upping the bias. That's one thing I miss about the LM317. I fitted a 4K7 screen resistor to the higher side which took the screen volts down to about 237V. Now all the output valves are running 42/43 and 43/44 mA. Will keep my eye on it. Thinking of ordering some LR8N3 from tech-diy.com. I have some 6CG7 tubes on the way to try as well. Won't be much happening here until book, tubes, regs arrive.
Ian.
The B+ for doing loadlines needs to be the expected B+ with the amp running. In our case if we bias for more current in the driver stage then the B+ for the driver stage will drop a bit more voltage across the 22K resistor feeding the 33uF caps. I have already changed the 22K for 10K and thinking of fitting 6K8, not a huge change but every bit helps.
It seems the LED's keep upping the bias. That's one thing I miss about the LM317. I fitted a 4K7 screen resistor to the higher side which took the screen volts down to about 237V. Now all the output valves are running 42/43 and 43/44 mA. Will keep my eye on it. Thinking of ordering some LR8N3 from tech-diy.com. I have some 6CG7 tubes on the way to try as well. Won't be much happening here until book, tubes, regs arrive.
Ian.
I put the anode and cathode resistors of the cathodyne back to 43K from 68K. This pulled the B+ from the tripler down from approx 350V to 334V due to increased current draw in the cathodyne. Changed the 10K dropper resistors to 6K8 to restore it back to 350V. Gary noted a big improvement by adding 0.0068uF caps to ground from the power transformer secondary to ground. I didn't have any and haven't gone shopping yet but found a 0.47uF 630V cap and put that across the doubler's output. So I did 2 changes at the same time, so the 0.47uF may have nothing to do with it, but this is the best I have heard this amp, I really like it.
The biggest improvements came from 1) fitting the DC-coupled cathodyne driver section, 2) fitting the voltage tripler to get 350V on the driver stage and 3) fitting the LED's to the output tubes' cathodes, which comes with additional work (construction and bias management) but it is well worth it.
If I was to try a different approach, it would be along the lines of the "El Cheapo" driver stage and using a voltage tripler to generate negative 130V and use that for the CCS on the long-tailed pair phase splitter, and also use the neg voltage for fixed bias. I won't be rebuilding this amp to find out how that sounds, but ~$200 for another new amp to try it out on is a possibility. I need to keep this amp as is to compare against.
The biggest improvements came from 1) fitting the DC-coupled cathodyne driver section, 2) fitting the voltage tripler to get 350V on the driver stage and 3) fitting the LED's to the output tubes' cathodes, which comes with additional work (construction and bias management) but it is well worth it.
If I was to try a different approach, it would be along the lines of the "El Cheapo" driver stage and using a voltage tripler to generate negative 130V and use that for the CCS on the long-tailed pair phase splitter, and also use the neg voltage for fixed bias. I won't be rebuilding this amp to find out how that sounds, but ~$200 for another new amp to try it out on is a possibility. I need to keep this amp as is to compare against.
No, they look all the same. I matched each string at 15mA to give the same voltage, also the 4.7R resistors help the strings share the load.
Ian, have you looked at the "El Cheapo" diagram? Might give you some other things to try. It uses 12AQ5 output tubes, which are a 12 volt heated analog of the 6P1 in a little 7-pin package.
The driver is a ECC81 in a LTP.
The driver is a ECC81 in a LTP.
Last edited:
Calculation questions. Not second guessing, trying to learn. I worked through basic calculations on your input stage. ( 43K on splitter version)
If I did things correctly, and the bad zerox for 6N1's,
the first stage load line is between 348V and 1.6mA. S,, the graph put the target bias at .8ma. Yo have measured about 1. I am guessing that is not atypical. So you have about 4Vp-p swing on the input before things get ugly. Guess that is OK for 1.2dBv.
The splitter line between 348 and 4mA. This limits the maximum symetrical swing, as the voltage on the grid ifs fixed by the plate of the input, an artifact of it's bias. In a perfect world, the splitter grid would be Half B+ minus the bias. . Are there other parameters I don't get yet?
In the output, do I have this correct? The higher the bias, the larger the class A overlap, the lower the bias, the less overlap. So, from the original 13V bias, they raised in in mine to 15 to lower distortion, or claim class A or something. You have lowered to 10.6 to get more higher efficiency, and maybe closer to the promised 10W.
Besides lower gain, what would be the effect of bringing back some amount of NF to a divider in the cathode of the input? Reducing 3rd order HD is the goal.
If I did things correctly, and the bad zerox for 6N1's,
the first stage load line is between 348V and 1.6mA. S,, the graph put the target bias at .8ma. Yo have measured about 1. I am guessing that is not atypical. So you have about 4Vp-p swing on the input before things get ugly. Guess that is OK for 1.2dBv.
The splitter line between 348 and 4mA. This limits the maximum symetrical swing, as the voltage on the grid ifs fixed by the plate of the input, an artifact of it's bias. In a perfect world, the splitter grid would be Half B+ minus the bias. . Are there other parameters I don't get yet?
In the output, do I have this correct? The higher the bias, the larger the class A overlap, the lower the bias, the less overlap. So, from the original 13V bias, they raised in in mine to 15 to lower distortion, or claim class A or something. You have lowered to 10.6 to get more higher efficiency, and maybe closer to the promised 10W.
Besides lower gain, what would be the effect of bringing back some amount of NF to a divider in the cathode of the input? Reducing 3rd order HD is the goal.
Hi tvr, I am trying to learn too, so be careful of any explanations I give, and the schems I post are for discussion only and could be exemplary of bad design practice or anything like that.
Seems you are missing one thing with the datasheet's static anode characteristics, the graph of Ia vs Va with grid voltages drawn across it. The graph is for voltages across the actual anode of the valve, relative to the cathode. No external parts are taken into account. So the voltage across the anode in the first stage of my schem is 104 at the anode - 2.2 on the cathode = 102V across the valve (10Jan2010 schem) or 134.8 - 2.83 = 132V (23Jan2010 schem). The valve doesn't care if it has 300V above it or below it, all it cares about is the actual voltage from anode to cathode, and the grid voltage which will control the current flow.
Again there is only approx 130V across the valve with 43K resistors and 350V B+. Cathode current is about 2mA. I have had 3 various cathodyne (split-load) splitter setups in this amp now and the grid sits at the anode voltage of the preceding stage and the cathode of the splitter moves to a voltage to create the right bias for that given setup. There are near symmetrical voltages across the cathode and anode resistors of the splitter because the same current flows through both of them. Then there is the voltage from the anode to the cathode of the valve, so there are 3 voltage drops at work here. So no, the grid does not sit at half B+. With an ac signal the anode of the first stage varies its voltage say 6Vp-p so you can imagine the cathode of the splitter stage varying a similar amount trying to track the grid voltage. If the grid voltage starts rising at any instant, valve current will increase, which will increase the current through the cathode resistor, which will raise the cathode voltage, thus the cathode "tracks" the grid to an extent. And due to increased current the voltage drop across the anode resistor will also increase (because its all in series), thus creating an identical output signal at both the cathode and anode, but out of phase.
Or you could visualize it this way - consider the anode resistor, the valve, and the cathode resistor, to be 3 rubber bands attached together in series and stretched a bit. Consider the length of each rubber band to represent the voltage drop across each component. When the middle rubber band contracts (when the valve current increases) the 2 outside rubber bands stretch more (more voltage across them). When the middle rubber band relaxes (when valve current decreases) the other 2 rubber bands take up the slack and get shorter and have less voltage across them.
It is a wierd setup, been thinking about it for days and days, its hard to visualize exactly what is going on...it's the best I can come up with so far.
No, I am far from the "promised 10 watts" hahaha. Very far away. As far as I understand, Class AB1 happens when one valve goes into cut-off (class AB1 means each valve is fully conducting for more than half each cycle, or in cut-off for less than half a cycle). This cannot happen if the loadline is in a position where the valve cannot go into cut-off, i.e. biased for high current like mine is. I am out of my depth here. Valvewizard has some good stuff on it. Does the Morgan Jones book shed any more light on this? I would think with a 2 channel scope one could monitor the input signal and the voltage across the output tube cathode resistor and see if the output tube goes into cut-off and at what power that starts happening, would that be the best way to see?
MJ books are very good, but not as complete as I was hoping. Both of his, and ValveWizard have very good supplemental on execution, not just the valves.
From Morgan, you draw the line based on the full B+, as if the valve was off, then no current so the full voltage would show up on the plate. You put the other end on the current it would draw if the tube was a short and the current limited by the total of the external parts. That is for an existing design. For new, it would be more of setting the voltage, then moving a clear ruler around that point to find a current that crosses the lines where it is reasonablly symmetrical and you have enough swing on the grid. If I got this wrong, someone please pipe up!
I will do some more benchmarks to measure the A-B range. It looked like my splitter was clipping before my output. Not too worried about that as I intend to start with your splitter and move from there. Thinking about it, if each cathode was fed from a ccs, I don't see why the paraphase should be so bad. I want to measure what swing is needed to clip the output. Still thinking about a ss input. Several of the diff line drivers can swing more than 30V. If that is not enough, then use it as the input splitter, and use the now free tubes for a diff amp driver stage. Extra power supplies, and on and on.
From Morgan, you draw the line based on the full B+, as if the valve was off, then no current so the full voltage would show up on the plate. You put the other end on the current it would draw if the tube was a short and the current limited by the total of the external parts. That is for an existing design. For new, it would be more of setting the voltage, then moving a clear ruler around that point to find a current that crosses the lines where it is reasonablly symmetrical and you have enough swing on the grid. If I got this wrong, someone please pipe up!
I will do some more benchmarks to measure the A-B range. It looked like my splitter was clipping before my output. Not too worried about that as I intend to start with your splitter and move from there. Thinking about it, if each cathode was fed from a ccs, I don't see why the paraphase should be so bad. I want to measure what swing is needed to clip the output. Still thinking about a ss input. Several of the diff line drivers can swing more than 30V. If that is not enough, then use it as the input splitter, and use the now free tubes for a diff amp driver stage. Extra power supplies, and on and on.
Well i decided to leave mine unmodded for the time being. But i will borrow the output TXs for another EL84 amp that's soon to be built. At least i'll have something to compare with the Mengyue amp.
Hope the new amp sounds great Trebla. Good idea to have 2 amps to compare. I hope the new one sounds heaps better than the standard Mengyue.
I'm thinking of trying mosfet source followers after the splitter outputs with individual bias for each output tube (and keep the LED's as cathode loads). Another idea I have is to swap in the OPT's from a 1956 6GW8 amp I have.
I'm thinking of trying mosfet source followers after the splitter outputs with individual bias for each output tube (and keep the LED's as cathode loads). Another idea I have is to swap in the OPT's from a 1956 6GW8 amp I have.
Ian.
Well it's a good few years since i had an EL84 amp (Leak Stereo 20), and that was in a totally different system. So can't make any fair comparisons.
I imagine the new amp will sound different but we'll have to wait and see if there's any improvement.
I noticed you tried some Obligatto caps earlier on, and preferred the X class ones. I used Obbs quite a bit, and can say that they are good for taming anything that sounds bright or forward. So i can see why you came to the conclusion that there was more life in the originals.
I think Obbs are good caps in the right place though and may be worth another try one day.
If you want a fairly transparent cap that doesn't cost the earth, try Clarity Caps or black Soniqs ( more or less the same thing).
Good luck with your mods, and i'll report back when i've built the thing. Though it may be a while because i have to give up my workshop (spare bedroom), and resort to using the kitchen when it's free..
Well it's a good few years since i had an EL84 amp (Leak Stereo 20), and that was in a totally different system. So can't make any fair comparisons.
I imagine the new amp will sound different but we'll have to wait and see if there's any improvement.
I noticed you tried some Obligatto caps earlier on, and preferred the X class ones. I used Obbs quite a bit, and can say that they are good for taming anything that sounds bright or forward. So i can see why you came to the conclusion that there was more life in the originals.
I think Obbs are good caps in the right place though and may be worth another try one day.
If you want a fairly transparent cap that doesn't cost the earth, try Clarity Caps or black Soniqs ( more or less the same thing).
Good luck with your mods, and i'll report back when i've built the thing. Though it may be a while because i have to give up my workshop (spare bedroom), and resort to using the kitchen when it's free..
Rotaspec
Still not decideMeng El84 or other. What do you think about this amp 6L6 6P3 Class A single ended stereo tube amplifier - eBay (item 270493821138 end time Jan-29-10 18:39:09 PST)
Still not decideMeng El84 or other. What do you think about this amp 6L6 6P3 Class A single ended stereo tube amplifier - eBay (item 270493821138 end time Jan-29-10 18:39:09 PST)
Well, I solved the underdamped bass problem. Nothing to do with the amp.
I replaced the 35 year old caps in the crossovers of my Rogers TLs and bingo!
I was hoping for an improvement, but this is a whole new level.
4 hours and $50 well spent.
I can get back to hearing improvements in the amp now 🙂
Gary
I replaced the 35 year old caps in the crossovers of my Rogers TLs and bingo!
I was hoping for an improvement, but this is a whole new level.
4 hours and $50 well spent.
I can get back to hearing improvements in the amp now 🙂
Gary
I think the next thing I will try is to turn the Meng input stage into a cathode follower AC coupled to the phase splitter.
I'm going to build an Aikido line stage to drive it with. However, both ideas may have to wait for a while while I amass enough parts for the Aikido (mainly PSU and chassis, I have tons of 6Nx series tubes to try)
I'm going to build an Aikido line stage to drive it with. However, both ideas may have to wait for a while while I amass enough parts for the Aikido (mainly PSU and chassis, I have tons of 6Nx series tubes to try)
I guess this is the point where it ceases to be a Meng amp, and becomes something else entirely..🙂
Well, to be fair, it's already way beyond the original design (thankfully better).
The Aikido will be an external box to serve several amps, so it is not altering the Meng per se.
The idea of altering the first stage to CF is to drop the gain. I haven't got any nfb in circuit, so I may be able to do it with that. It's just a case of when the input stage will overload.
Gary, glad to hear about your fix for your speakers, I bet you were happy with that! Interesting plan you have with the Aikido stage. Now what's tvr going to come up with, I wonder? He has hinted at another very different direction too.
I've been reading about the use of mosfets as source followers by Eli Duttman, and gingertubes in his Baby Huey amp, Tubelab's Powerdrive scheme and R.G. Keen's Mosfet Follies. It will also give independent bias adjustment for each tube so I don't need matched pairs. I ordered the parts today, here's a tentative schematic. Comments good or bad welcome, all ears here.
I found an interesting note about possible arcing between grid and cathode with DC-coupled cathode followers at the valvewizard site, scroll down to the last paragraph. I guess the same principle applies to a DC-coupled cathodyne. For the sake of one diode, proly worth it.
Ian.
I've been reading about the use of mosfets as source followers by Eli Duttman, and gingertubes in his Baby Huey amp, Tubelab's Powerdrive scheme and R.G. Keen's Mosfet Follies. It will also give independent bias adjustment for each tube so I don't need matched pairs. I ordered the parts today, here's a tentative schematic. Comments good or bad welcome, all ears here.
I found an interesting note about possible arcing between grid and cathode with DC-coupled cathode followers at the valvewizard site, scroll down to the last paragraph. I guess the same principle applies to a DC-coupled cathodyne. For the sake of one diode, proly worth it.
Ian.
The 6P1P drive requirement is low enough that I believe it is unnecessary to use a FET drive source followers for grid drive (PowerDrive). I've seen no indication that the direct coupled phase splitter can not properly drive them. The 6N1P should be quite nice for the job.
On the other hand, if you are just interested in experimenting with PowerDrive, go for it and let us know how it works out.
The LED stack looks good, but be sure to measure your LEDs as they can vary a good bit and it is best to keep them pretty closely matched. The 4R7 resistors can only compensate for a slight missmatch. I ended up with a stack of 5 LEDs for class A operation, and the tubes were running at max power (12W IIRC). A stack of six with the 4R7s will probably bias you around 8W max.
I would also change the single 1K5 Screen Resistor to two 1K screen resistors at the tube socket. I've seen no tendency for the 6P1P to oscillate due to a lack of screen stoppers, but the resistors are cheap.
Good luck and let us know how it sounds.
On the other hand, if you are just interested in experimenting with PowerDrive, go for it and let us know how it works out.
The LED stack looks good, but be sure to measure your LEDs as they can vary a good bit and it is best to keep them pretty closely matched. The 4R7 resistors can only compensate for a slight missmatch. I ended up with a stack of 5 LEDs for class A operation, and the tubes were running at max power (12W IIRC). A stack of six with the 4R7s will probably bias you around 8W max.
I would also change the single 1K5 Screen Resistor to two 1K screen resistors at the tube socket. I've seen no tendency for the 6P1P to oscillate due to a lack of screen stoppers, but the resistors are cheap.
Good luck and let us know how it sounds.
- Status
- Not open for further replies.