I did it for a friend who let me down very badly. Thanks for letting me show it. As the grid of EL34 No 2 is hard to ground one must use a scope if adding feedback. In transistor long tail pairs there is folklaw about extra stability problems of the cathode ( emitter coupled ) Blumlein pair. To my mind this is folklaw. Usually the real advantage of the single transistor input stage was the high Cbc of the typical 1962 VAS transistor as in JLH 10 watt design.
One thing dislikers of transistor amps do not know is that the amplifier of the D Self/ HC Lin type is completely linear except in one horrible section. That is if enough trouble taken in small things. The VAS as it is called is in fact a transimpedance stage like a DAC integrator. Now to a simple idea. A pentode if like the triode ECC88 that could work at sensible voltages the real advantages of transistors could be realised and no transimpedance needed in audio range and way above ( no transmission line ). Feedback could be loop or local and likely both. Good AB stages only need 20 dB feedback, so AB is not the problem if 30 MHz fast. I could see an RH84 with suitable output valves as being a possible stage 1 and 2 of an H C Line type idea ( it was the best and still is as it can use input shunt feedback ). If this idea sounds as good as the Croft OTL I would think it interesting. Fast, clean and deep. Not a 300 B sound. Realistic final HT of 125V and AC coupled via a capacitor. Very cheap electolytics out measure Sowter output transformers. No Folklaw please on caps are bad.
One thing dislikers of transistor amps do not know is that the amplifier of the D Self/ HC Lin type is completely linear except in one horrible section. That is if enough trouble taken in small things. The VAS as it is called is in fact a transimpedance stage like a DAC integrator. Now to a simple idea. A pentode if like the triode ECC88 that could work at sensible voltages the real advantages of transistors could be realised and no transimpedance needed in audio range and way above ( no transmission line ). Feedback could be loop or local and likely both. Good AB stages only need 20 dB feedback, so AB is not the problem if 30 MHz fast. I could see an RH84 with suitable output valves as being a possible stage 1 and 2 of an H C Line type idea ( it was the best and still is as it can use input shunt feedback ). If this idea sounds as good as the Croft OTL I would think it interesting. Fast, clean and deep. Not a 300 B sound. Realistic final HT of 125V and AC coupled via a capacitor. Very cheap electolytics out measure Sowter output transformers. No Folklaw please on caps are bad.
This is a concept and might be poor on detail. It will need more gain. An ECC83 would be ideal. Or op amp to prove concept. The FET's unlike bipolars will run close to V max. 250 V types exist. Don't use IRF/IRFP types as they need specail biasing. The FET's shown I think have the gate zener and back EMF diode built in. The harmonic distortion without feedback of the FET's is 0.8% from memory. 0.05% if a complimentary feedback set up of NPN PNP + PNP NPN. If the amp was reduced to gain of 2 I suspect it would do fine. The a gain of 30 plus to stage one not shown.
I built something like this recently to power a turntable albeit transformer driven and no loop feedback. It is rather good. FET's need about 1.4 V DC gate to gate to give the standing current. Not the 5 V people say, that's industrial FET's. .
I was in a harry when I wrote that. The bootstrapped constant current source will allow a pentode on a limited HT to think it sees a much higher HT or higher impedance. The reduction in distortion is dramatic. One can use a MJE 350 CCS. It has many limitations that the 1960's way doesn't. The voltage gain of a MOSFET output is about 0.78. That is ideal. <0.93 is not a problem. The Cmfbp output is 0.98 which often needs a serries resistor to avoid ringing. One could run the pentode ( made triode for simplicity ) at MOS FET HT x 2. Even so a CCS is worth having. Yes is is positive feedback to the anode. Should be negative to the cathode as the output is inphase with the anode as a source follower ( current amp ).
Last edited:
The FET's don't invert phase and the cathode is in anti phase from the anode, surely it should work? I am just doing what looks out of phase, maybe it won't be working as a standard phase spitter when so aranged? It seems to me if we could get it this simple we might just arrive at a good class AB amp with virtually no nasty harmonics. The FET's have a curve that will respond well to feedback. The RH84 might have a 6K transformer, that's 27 : 1 ratio into 8R. 5 watts seems what it could give. That is a conservative 2.5 vrms from 2.5 vrms in. Allowing for the 27 : 1 ratio we have a very nice 100 watts from 1 Vrms more or less with the higher HT. We have to be realistic and add at least 20 dB feedback. That means 10 V rms in. As 30 Vrms is a sensible maximum output from a gain stage we might go as far as 36 db of feedback. 20 db might yeild a 0.1 % from the buffer stage and 36 dB 0.04%. It hovers on the possible. The sound might be better than imagined as the FET's like feedback and seem not to ask for complex solutions.
If I screwed up the feedback sorry. To my lets find out eyes it seems to have the correct relationship. I drew the AC waves to show it and inverted where it is, the FET's have zero phase shift at audio as shown. I saw the pentode as like a Dynaco phase splitter where when we add the two halves we would get zero. With a transformer it is easier as we can change phase by which end we ground. We would not easilly be able to set up the FET's with voltage gain. It is possitive feedback to the Anode.
If I screwed up the feedback sorry. To my lets find out eyes it seems to have the correct relationship. I drew the AC waves to show it and inverted where it is, the FET's have zero phase shift at audio as shown. I saw the pentode as like a Dynaco phase splitter where when we add the two halves we would get zero. With a transformer it is easier as we can change phase by which end we ground. We would not easilly be able to set up the FET's with voltage gain. It is possitive feedback to the Anode.
Hey guys I'm back to working on this amp. I've used the Edcor WSM 10k-10k as an input transformer twice now. Both Pete Millet's two stage 6SN7 uniamp front end. One driving 2A3 and the other driving Ad1's. Both sounded very good to me and those who heard them. Plus I think they are about $10 each.
Any way, I've come up with a new schematic using 12AU7 and 7591. I did the calculations and 7591S needs minuscule current to drive it and only about 34 volts pp so 12AU7 can be run conservatively. So I've attached that schematic here.
By the way its running fixed bias at about -17 volts which is not on this schematic.
Any way, I've come up with a new schematic using 12AU7 and 7591. I did the calculations and 7591S needs minuscule current to drive it and only about 34 volts pp so 12AU7 can be run conservatively. So I've attached that schematic here.
By the way its running fixed bias at about -17 volts which is not on this schematic.
Attachments
And a question? SY turned me on to this schematic in a earlier post (see attached). I've built a couple RH84's and liked the sound. All of the RH amps use plate to plate feedback. The attached uses Schade feedback at the plate and the cathode. It also uses global feedback. My question is what is the effect of feedback at both the plate and the cathode? In this case R21 is 330k and R22 is 150k. Does Schade feedback conform to a formula? Or is it trial and error?
Attachments
Reading Kitic he didn't use a cathode capacitor on his input ECC81 to keep Rp as high as possible. Doing my own tests there seems to be an exchange of where distortion in lowered. When using an ECC82 still without the cathode capacitor distortion was very low and perhaps a nicer curve. The greater difference seemed to be gain which slightly favoured Kitic. Kitic also argued a very low output Rp for the contrived triode from pentode ( EL84,34,807,KT88 ). His suggestion being about 30% lower than the supposed triode with g2 to anode. If so it is important to do it his way as it enhances the low damping factor a little. That is allow as much shunt feedback as possible into the output pentode ( EL 34 my prefered ). I found that using an input pentode in place of ECC81 took me from 2.5 V sensetivity to 620 mV ( 4 Vrms ECC82 or 12HB7A ). EF184 if I get the number right can almost mimic an ECC81 in triode. That means it can be anything you want within this recipe, in theory the Rp of a pentode is what we want ( or a FET drain ). I used a spectrum analyser at all stages of my design. I even used input pentodes with low emission ( from Chipping Norton Studios junk bucket, Remember Fairground Attraction wonderful LP? )and got near indentical curves except with slightly more distortion. I must have tried 20 valves to be sure including different makes. Some the white box types ( Z&I Aero Services were some ). I had been told I would not be able to duplicate results. Not true. I call this front to back Ultra linear. The mix of near triode with pure pentode input looks like an ideal triode if SE ( in PP , SE driven perhaps all triode is best ). I used 82% triode end UL at the output as a spare tap allowed that. It seemed just like triode except more sensetive. If 43% available I feel convinced it would have been a good idea especially if PP. I am reasonably sure UL and loop feedback a difficult combination. GEC suggest ways to make it work in the 1972 designs book.
I should have said this. I spent 6 months playing with that design. It sat next to my computer. I must have tried 10 ideas a day. Never a day went by when some small change for the better was made. Much included PP although I had no need to do that. I happened to have some PP transformers in a box of exactly the same size. What I found is SE almost impossible to get wrong within reason. Most internet circuits seem to claim much lower distortion than I ever got. PP was really difficult until I tried SE driven. I only learnt that as someone wanted me to complete a 211 amp. The SE to PP driver was to drive the 211. 2 x EL34 almost need as much drive when a long tail pair. To my mind SE driven PP with 43% UL would be OK. What my six months did was get me to a very nice PP in two days. I was very surprised to find the PSU only slightly less important when PP. Perhaps the SE driven the reason. I must have spent 4 of the 6 months getting the PSU right. I guess 200 hours spent.
Thanks Nigel,
there are some limitations on this amp. Years ago I started collecting vintage iron thinking it was better than the available output transformers that were reasonable priced. I have since found Heyboer in Michigan, very close to home, that can produce anything I can specify. These output transformers are pentode only. And the power transformer filament wingdings are of limited amperage. The filament windings won't support EL34. Which is my favorite output tube.
Is a high Rp neccessary to use Schade feedback? If so ECC82 may not be the right tube. Kitic seems fond of ECC81 for many of his designs. I thought low Rp ws a good thing? Unbypassed ECC82 would lose a lot of gain, probably too much. Also low frequencies would suffer. I have 5751, 12AT7, 6CG7, 6GK5 and 6N6P among others.
there are some limitations on this amp. Years ago I started collecting vintage iron thinking it was better than the available output transformers that were reasonable priced. I have since found Heyboer in Michigan, very close to home, that can produce anything I can specify. These output transformers are pentode only. And the power transformer filament wingdings are of limited amperage. The filament windings won't support EL34. Which is my favorite output tube.
Is a high Rp neccessary to use Schade feedback? If so ECC82 may not be the right tube. Kitic seems fond of ECC81 for many of his designs. I thought low Rp ws a good thing? Unbypassed ECC82 would lose a lot of gain, probably too much. Also low frequencies would suffer. I have 5751, 12AT7, 6CG7, 6GK5 and 6N6P among others.
What people call "Schade" feedback is parallel applied voltage feedback. O.H. Schade actually used series applied voltage feedback in his paper, so it is technically incorrect to call plate-to-grid or plate-to-driver plate feedback "Schade feedback."
Parallel applied voltage feedback lowers the input impedance of a stage and makes it very difficult to drive. You can take a couple of approaches to solve the problem. You can drive it with a very low output impedance and set the feedback ratio by placing a resistor in series with the low driving impedance. That is the approach I took here. It gives very consistent feedback and very low distortion in an output stage.
Another approach is to choose a driver tube that has a plate resistance equal to the desired input resistance to the feedback network that you desire to create. The resistor doesn't actually exist, but you are using the plate resistance of the driver as that resistor. It is a very elegant design trick.
The driver tube operates into an almost vertical load line with very little voltage swing, so you need to select a tube that has evenly-spaced curves into that nearly-vertical load line. This is why pentodes are generally recommended. This is also why the Kitic designs receive so much criticism; Kitic chose tubes that don't have evenly-spaced curves into a vertical load line. Also, rp of the driver tube in the RH designs will vary wildly (compared with a pentode) throughout the signal swing, which will give varying feedback with signal swing, which will give varying damping factor with signal swing. All of these are things that we normally try to avoid.
So it is certainly possible to choose driver tubes that will give much less distortion than the RH designs and much more consistent feedback ratio, but a lot of people seem to like them anyway and they have quite a following.
Keeping the triode cathode unbypassed linearizes the spacing of the curves, so it is definitely advisable, but you could apply the same trick to a pentode and come out ahead.
There is a simple way of looking at this. Although never suggested in text books a triode has feedback between anode and grid as an electrostatic chain of electrons. This was mostly seen as a defect which in 1927 gave birth to the pentode. Faced with g3 the electrons no longer had an easy path to g1. If we restore that path we have a triode of sorts. One slightly complicated way to do it is to return the anode to g1( signal in ) via a capacitor if wanting to keep it simple. Usually this is a worse way. A terode was up to a point OK except had a kink in the amplifying curve. The pentode made that compromise go away ( discuss ). A kinkless tetrode was mostly a patent bust ( KT66 or the Harries valve co version of 1935 leading to Blumleins distributed load amplifer ). That's not to say for moving electrons a Beam tetrode is not a remarkable device. Mr Schade found the 807 Beam device a bit more nasty than he could live with. His 1937 amp showing the way and not using the Blumlein fix later to become UL by Hafler ( 1947, looks different ). I saw a 1960 Schade type design of RCA. It was basically a better Hafler. I am told RCA spent money to improve a TV valve to become 7199. The famous Hafler Dynaco could be a RCA idea for marketing with Hafler giving it quality parts. I will see if I still have it.
Where the Kitic idea works is if the input valve if allowed to work in an ideal way it will join in the process. The maths are very simple. The anode signal will be stopped by the ECC81 internal resistance Rp ( or whatever ) the anode load in paralell. If Rp is very high then we should focus on the anode load as where we can make a difference ( keep it as high as we can, ears can be used for this one ). Kitic found the anode of ECC81 can be directly connected to anode of the power pentode with one resistor ( later version, see RH84 also ). That works up to a point. What it demands is a first class power supply to avoid hum. I used a MOS FET capacitance multiplier to get OK hum levels. The purists might say it is introducing some of itself into the sound. It is true and can be seen. It is a minute additon of second harmonic distortion. Mostly a FET used this way is like a resistor. I can live with that.
Alex Kitic got banned from DIY Audio. People instead of understanding his concept told him he was very wrong and a plagarist. As far as I can tell Alex is penniless and only asking people to understand why the feedback trade off should be output pentode biased. It took me ages to understand this (due to what others say mostly ). Within reason what we win in one area we loose in another. If we can reduce the Rp of the output pentode we might win a free lunch of optimum damping factor. We do this by keeping the Rp of the input valve high. The capacitor to the ECC81/2 cathode influences this. On the calculator example I chose ratio was 5.25 to 1.
mh-audio.nl - Home
When ECC82 is used in place of 81 it shows no real gain change shunt feedback or not. In some Kitic circuits the gain change is 4 to 1. That suggests that something really is happening when ECC81 and with ECC82 it is not. What you will see is the ECC82 is a lower distortion device. This seems directly ( matamatically ) proportional to gain. ECC81 circa 10 ( 2.5 vrms in for 25 Vrms out ) and ECC82 about 6. The unmolested ECC81 I seem to remember was about 40 and easilly higher if in conventional set ups. If we take that to it's logical conclusion we can get 10 or more from an ECC82 using a cathode cap bypass. If we got 17 we might even beat the ECC81 on all counts ( you will ). What you loose is some damping factor.
In 1955 Wireless World asked what damping factor we need. Plenty of speakers still meet the same criteria ( Eminance 12 Lta ). Wireless World found a damping factor of 3 was enough. This was done by measuring over shoot on a bass unit. The test was simple and without any error I could see. It was repeated many times to get good data. The Kitic design lives in this area so is worth believing in. A simple test of speaker is can it live with LP warps without drastic bass filtering. You would be suprised how well the older designs cope and modern not.
On the spectrum analyser you will find the numbers favour the tranditional approach. What the numbers fail to show is damping factor. As most speakers are intended for transistor amplifers this might matter. What you might get is a frequncy responce far away from usual. It might suit and it might not. I hate the use of loop feedback on a valve design so gaining it that way is a rule I won't break. If someone paid me that would be different. Doubtless the Dynaco proves it can be done. It's not bad.
I found this by analysing the g1 input of at the time KT88. I can say what Kitic suggests is true and the things " he " wants do happen. " I " choose to not want this. My reason being I would design a speaker to suit. As it happens from 30 to 100 Hz I need about 100 watts +. From 100 Hz to 20 kHz ( +0- 1db ) I need 5 Watts. I don't use a sub woofer. I use a 15 inch driver Open Baffle. It really needs power and EQ to work. And it works very well. Flat to 40 Hz and good output at 30 Hz. My favourite amp for the big wattage is the Hitachi circuit of 1979 using MOS FET's. Double FET's to get more grunt.
Where the Kitic idea works is if the input valve if allowed to work in an ideal way it will join in the process. The maths are very simple. The anode signal will be stopped by the ECC81 internal resistance Rp ( or whatever ) the anode load in paralell. If Rp is very high then we should focus on the anode load as where we can make a difference ( keep it as high as we can, ears can be used for this one ). Kitic found the anode of ECC81 can be directly connected to anode of the power pentode with one resistor ( later version, see RH84 also ). That works up to a point. What it demands is a first class power supply to avoid hum. I used a MOS FET capacitance multiplier to get OK hum levels. The purists might say it is introducing some of itself into the sound. It is true and can be seen. It is a minute additon of second harmonic distortion. Mostly a FET used this way is like a resistor. I can live with that.
Alex Kitic got banned from DIY Audio. People instead of understanding his concept told him he was very wrong and a plagarist. As far as I can tell Alex is penniless and only asking people to understand why the feedback trade off should be output pentode biased. It took me ages to understand this (due to what others say mostly ). Within reason what we win in one area we loose in another. If we can reduce the Rp of the output pentode we might win a free lunch of optimum damping factor. We do this by keeping the Rp of the input valve high. The capacitor to the ECC81/2 cathode influences this. On the calculator example I chose ratio was 5.25 to 1.
mh-audio.nl - Home
When ECC82 is used in place of 81 it shows no real gain change shunt feedback or not. In some Kitic circuits the gain change is 4 to 1. That suggests that something really is happening when ECC81 and with ECC82 it is not. What you will see is the ECC82 is a lower distortion device. This seems directly ( matamatically ) proportional to gain. ECC81 circa 10 ( 2.5 vrms in for 25 Vrms out ) and ECC82 about 6. The unmolested ECC81 I seem to remember was about 40 and easilly higher if in conventional set ups. If we take that to it's logical conclusion we can get 10 or more from an ECC82 using a cathode cap bypass. If we got 17 we might even beat the ECC81 on all counts ( you will ). What you loose is some damping factor.
In 1955 Wireless World asked what damping factor we need. Plenty of speakers still meet the same criteria ( Eminance 12 Lta ). Wireless World found a damping factor of 3 was enough. This was done by measuring over shoot on a bass unit. The test was simple and without any error I could see. It was repeated many times to get good data. The Kitic design lives in this area so is worth believing in. A simple test of speaker is can it live with LP warps without drastic bass filtering. You would be suprised how well the older designs cope and modern not.
On the spectrum analyser you will find the numbers favour the tranditional approach. What the numbers fail to show is damping factor. As most speakers are intended for transistor amplifers this might matter. What you might get is a frequncy responce far away from usual. It might suit and it might not. I hate the use of loop feedback on a valve design so gaining it that way is a rule I won't break. If someone paid me that would be different. Doubtless the Dynaco proves it can be done. It's not bad.
I found this by analysing the g1 input of at the time KT88. I can say what Kitic suggests is true and the things " he " wants do happen. " I " choose to not want this. My reason being I would design a speaker to suit. As it happens from 30 to 100 Hz I need about 100 watts +. From 100 Hz to 20 kHz ( +0- 1db ) I need 5 Watts. I don't use a sub woofer. I use a 15 inch driver Open Baffle. It really needs power and EQ to work. And it works very well. Flat to 40 Hz and good output at 30 Hz. My favourite amp for the big wattage is the Hitachi circuit of 1979 using MOS FET's. Double FET's to get more grunt.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.