You have effectively 11uF as the first cap. I usually find I need to drop below that before a significant reduction in filter OP voltage occurs, sometimes even as low as 4uF.
As Astouffer says, you can use PSUD II to find a good solution. You must make sure you give it all the relevant parameters, including the DC resistance of the transformer and any chokes.
Use good heatsinks for MOSFETs. I've seen illustrations of some finned heatsinks. e.g. this link (scroll down the page).
Hello all,
I've actually been using PSUD II for awhile now, but I must say this project has helped me learn how to really use it! Here's what I've entered for the power transformer:
This gives me a turns ratio of 3.65:1 and an estimated impedance of 17.15 ohms. So far, so good, except I'm not sure if the DC resistance is measured across the whole 800-volt winding, or is the average of the two windings measured from the center tap?
I've actually been using PSUD II for awhile now, but I must say this project has helped me learn how to really use it! Here's what I've entered for the power transformer:
- Primary voltage: 122 V
- Primary DC Resistance: 0.5 ohms
- Secondary voltage: 445V
- Secondary DC resistance: 10.5 ohms
This gives me a turns ratio of 3.65:1 and an estimated impedance of 17.15 ohms. So far, so good, except I'm not sure if the DC resistance is measured across the whole 800-volt winding, or is the average of the two windings measured from the center tap?
Apologies for coming in late; most points have been addressed with excellent replies. But I have a few general comments from long experience with tube amplifiers of just about every topology. (... and submitting them with complete respect for those above who differed!)
Regarding the Williamson and Mullard topologies in general, I think the former often fetches undeserved criticism. Both have characteristics and disadvantages of their own. Both can have stability problems if the phase design is not looked after for a particular set of component values.
Williamson used the high-L OPT transformer for the reasons explained in his design, and it is a healthy approach, though unwieldy. A different approach with a more practical OPT will certainly work, but then with appropriate phase design - it is not the topology per se that causes problems. (I use that topology with one additional L.F. time constant and a 100W high-end design with good stability.) The Mullard also has 3 L.F. time constants to contend with; same considerations.
More specific:
1. I do not consider 20dB of NFB 'insane'! Again stability depends on the design; the original Mullard 5-20 design by Ferguson used 30dB of NFB. But it is hardly likely that one can just slap components into a previous design and expect it to be stable. That is a business that needs to be sorted out for every different design, with at least a 'scope, sine and square wave signal generator available.
Not to open an off-thread discussion here, but basically NFB is only bad when really 'insane' amounts were used in designs that were mediocre to begin with. This is still the case in too many semiconductor designs.
2. The 1M in parallel with the 0,25uF coupling cap is there in the first place to give phase neutralisation of this element at very low frequencies. The initial 6dB phase shift is 'zeroed' once the loop gain has been reduced sufficiently to obviate oscillation, giving a phase step function.
3. I frown on the 220E series cathode resistors of the 6SN7 - cannot see the purpose, other than piously trying to equalise them even more than they usually are. But it does lift the rp (internal resistance) into a less linear region, apart from decreasing gain.
4. Yes, the h.t. is simply too high. Other than great expenditure for a different power transformer, I would second the voltage reduction metod given in 'MOSFET Follies'. Reducing the PSU input caps: I fear by the time they are low enough to give the necessary effect, you are going to be so 'half-way' to an L-input filter as to compromise PSU regulation.
You may well be able then to use this drop as your fixed-bias source, without the rather inelegant step-up heater transformer arragement.
Another way of doing this is to use an amplified zener effect as a cathode voltage sink (fixed bias) in series with the 6550 cathodes, much like the above method. It could be made adjustable with a simple transistor circuit. (That is what I use above.)
Regarding the Williamson and Mullard topologies in general, I think the former often fetches undeserved criticism. Both have characteristics and disadvantages of their own. Both can have stability problems if the phase design is not looked after for a particular set of component values.
Williamson used the high-L OPT transformer for the reasons explained in his design, and it is a healthy approach, though unwieldy. A different approach with a more practical OPT will certainly work, but then with appropriate phase design - it is not the topology per se that causes problems. (I use that topology with one additional L.F. time constant and a 100W high-end design with good stability.) The Mullard also has 3 L.F. time constants to contend with; same considerations.
More specific:
1. I do not consider 20dB of NFB 'insane'! Again stability depends on the design; the original Mullard 5-20 design by Ferguson used 30dB of NFB. But it is hardly likely that one can just slap components into a previous design and expect it to be stable. That is a business that needs to be sorted out for every different design, with at least a 'scope, sine and square wave signal generator available.
Not to open an off-thread discussion here, but basically NFB is only bad when really 'insane' amounts were used in designs that were mediocre to begin with. This is still the case in too many semiconductor designs.
2. The 1M in parallel with the 0,25uF coupling cap is there in the first place to give phase neutralisation of this element at very low frequencies. The initial 6dB phase shift is 'zeroed' once the loop gain has been reduced sufficiently to obviate oscillation, giving a phase step function.
3. I frown on the 220E series cathode resistors of the 6SN7 - cannot see the purpose, other than piously trying to equalise them even more than they usually are. But it does lift the rp (internal resistance) into a less linear region, apart from decreasing gain.
4. Yes, the h.t. is simply too high. Other than great expenditure for a different power transformer, I would second the voltage reduction metod given in 'MOSFET Follies'. Reducing the PSU input caps: I fear by the time they are low enough to give the necessary effect, you are going to be so 'half-way' to an L-input filter as to compromise PSU regulation.
You may well be able then to use this drop as your fixed-bias source, without the rather inelegant step-up heater transformer arragement.
Another way of doing this is to use an amplified zener effect as a cathode voltage sink (fixed bias) in series with the 6550 cathodes, much like the above method. It could be made adjustable with a simple transistor circuit. (That is what I use above.)
I am in complete agreement! NFB is a tool, not a crutch. Good results are obtained, when the circuitry is reasonably linear open loop. IMO, the lowering of O/P impedance may be more important than the reduced distortion.
The H/K Cit. 2 employs something like 34 dB. of NFB and it's unconditionally stable. The high gm of the 12BY7 pentodes Hegeman employed provides the requisite resistance against slewing trouble.
Thanks Johan, Tom and Eli.
Johan, would you happen to have a schematic of the amplified zener bias scheme? I'll have to agree the bias arrangement on this amp is a bit of a kludge. Indeed, many power transformers of the 50's and 60's had a bias tap on the plate winding rather than a separate transformer.
I also agree that PSUDII is a tool, not a way of life. For instance, my plate voltage actually went UP(!) when I decreased the input capacitance as stated in the previous posts above.
I have read the articles by dhtrob in DIYMag, but I'll have to say that through experience I've found that this software can sometimes get one into a "vicious cycle". When you have a CLCRCRC power supply, the plate voltages and currents for the preamp section are entirely dependent on the current drawn by the output section. At the same time, the preamp section also exerts its smaller influence on the output tubes. It's tough to get all the parameters correct, especially when companies like Hammond don't publish the percent regulation or DC resistances, etc. of their transformers. It's a very delicate balancing act, all a part of the appeal of DIY I suppose.
Again, I beg your indulgence as this is the very first time I've fiddled about with such a design using only a schematic as a rough guide. All of my previous projects were proven designs built from magazine articles. The authors had worked out all the "kinks" of their designs, and they worked without a problem once built. This is much different, and I am learning quite a bit thanks to fora such as this. I am in your debt.
Johan, would you happen to have a schematic of the amplified zener bias scheme? I'll have to agree the bias arrangement on this amp is a bit of a kludge. Indeed, many power transformers of the 50's and 60's had a bias tap on the plate winding rather than a separate transformer.
I also agree that PSUDII is a tool, not a way of life. For instance, my plate voltage actually went UP(!) when I decreased the input capacitance as stated in the previous posts above.
I have read the articles by dhtrob in DIYMag, but I'll have to say that through experience I've found that this software can sometimes get one into a "vicious cycle". When you have a CLCRCRC power supply, the plate voltages and currents for the preamp section are entirely dependent on the current drawn by the output section. At the same time, the preamp section also exerts its smaller influence on the output tubes. It's tough to get all the parameters correct, especially when companies like Hammond don't publish the percent regulation or DC resistances, etc. of their transformers. It's a very delicate balancing act, all a part of the appeal of DIY I suppose.
Again, I beg your indulgence as this is the very first time I've fiddled about with such a design using only a schematic as a rough guide. All of my previous projects were proven designs built from magazine articles. The authors had worked out all the "kinks" of their designs, and they worked without a problem once built. This is much different, and I am learning quite a bit thanks to fora such as this. I am in your debt.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.