sofia curved 815
I had an HF instability problem w/ a pair of Brook 10's; they had no cap with the feedback resistor. I played with values, something like a 200 pf did the job best.
At some point I will breadboard an 815 amp using either a long tailed pair or concertina splitter. But I wanted to curve some 815's on my Sofia to get more of an idea on how to work with both elements in parallel.
Attached are a few curves. Settings were: plate 400v, 150mA max; screen 225v 60 mA max; 5 curves from -25v to -35vdc. I curved appx 6; these are the best two.
To my surprise, all of them curved more like a tetrode than a beam tube. I realize 60 mA max limit on the screen is pushing it; but I just couldn't get good curves with a 30-40mA limit. Either it wouldn't draw them (current limiting) or Cg1 settings had to be made more and more negative that they were approaching cutoff. I did run a few up to 80mA on the screen!
-Kent
I had an HF instability problem w/ a pair of Brook 10's; they had no cap with the feedback resistor. I played with values, something like a 200 pf did the job best.
At some point I will breadboard an 815 amp using either a long tailed pair or concertina splitter. But I wanted to curve some 815's on my Sofia to get more of an idea on how to work with both elements in parallel.
Attached are a few curves. Settings were: plate 400v, 150mA max; screen 225v 60 mA max; 5 curves from -25v to -35vdc. I curved appx 6; these are the best two.
To my surprise, all of them curved more like a tetrode than a beam tube. I realize 60 mA max limit on the screen is pushing it; but I just couldn't get good curves with a 30-40mA limit. Either it wouldn't draw them (current limiting) or Cg1 settings had to be made more and more negative that they were approaching cutoff. I did run a few up to 80mA on the screen!
-Kent
Attachments
Haha. Trying to look at a 10khz square wave scared me because the output was a hideous mess. Then I looked at the laptop's output and discovered the amp was reproducing a hideous mess perfectly. So once i've put together a square wave generator I'll adjust the high end stuff further. I _was_ able to reduce ringing on a 1khz square wave by adding capacitance across the FB resistor.
No more LF oscillation either. It drops off sharply below 30Hz or so, but I assume that's the limit of what the output transformer is capable of (and indeed it's labeled as such)
http://mexico.limpoc.com/~eric/low-end.gif
(the blip at the very bottom is an artifact of the PC sampling, I'm pretty sure)
So... adding capacitance across the FB resistor of course causes proportionally more signal fed back as frequency increases... And adding capacitance across the plate load of the input effectively reduces the plate load resistor with increasing frequency, so that the amplitude of the signal going to the PI stage decreases, right? Or should I be thinking more about the resonant frequency of the RC combinations there? I'm trying to understand how you picked the values you recommended for those caps.
So once i've put together a square wave generator I'll adjust the high end stuff further. I _was_ able to reduce ringing on a 1khz square wave by adding capacitance across the FB resistor. No more LF oscillation either. It drops off sharply below 30Hz or so, but I assume that's the limit of what the output transformer is capable of (and indeed it's labeled as such)
http://mexico.limpoc.com/~eric/low-end.gif
(the blip at the very bottom is an artifact of the PC sampling, I'm pretty sure)
So... adding capacitance across the FB resistor of course causes proportionally more signal fed back as frequency increases... And adding capacitance across the plate load of the input effectively reduces the plate load resistor with increasing frequency, so that the amplitude of the signal going to the PI stage decreases, right? Or should I be thinking more about the resonant frequency of the RC combinations there? I'm trying to understand how you picked the values you recommended for those caps.
qq, your LF response trace looks pretty good.
BTW, if you find you're getting HF oscillation (t's a risk with beam tetrodes, especially ones designed for RFR use), you might need to use 'stoppers' on the screen and control grids. In fact, you should use them anyway. Use carbon composition resistors if you have them (certainly NOT wirewound). For the screens, use 100 ohm max., rated at 2w (because they can get hot frfom the screen electrode). For the control grids, use around 4.7k (as long as you're not using a direct-coupled cf driver, otherwise the stoppers have to go on the driver grids). Solder the stopper resistors as close as is practically possible to the tube pins concerned (screen or control grids), with a very short lead, or they won't be effective.
It is usually considered preferable to do most of the phase correction at the plate of the first stage, i.e. to get as close as possible to a flat frequency response from 20Hz to 20kHz with no peak at high frequency. You then apply a shunt capacitor across the feedback resistor, as necessary, to trim th eHF response and get rid of any remaining ringing on square waves. Crowhurst took the view that the common practice of using only a FB resistor shunt capacitor is a poor solution, because it can mask what's really going on inside the amp.
I just picked 'rule of thumb' starting values. They are not correct for any particular OP transformer but they are fairly typical of the values that are used with OP beam tetrodes/pentodes. You have to start somewhere!
BTW, if you find you're getting HF oscillation (t's a risk with beam tetrodes, especially ones designed for RFR use), you might need to use 'stoppers' on the screen and control grids. In fact, you should use them anyway. Use carbon composition resistors if you have them (certainly NOT wirewound). For the screens, use 100 ohm max., rated at 2w (because they can get hot frfom the screen electrode). For the control grids, use around 4.7k (as long as you're not using a direct-coupled cf driver, otherwise the stoppers have to go on the driver grids). Solder the stopper resistors as close as is practically possible to the tube pins concerned (screen or control grids), with a very short lead, or they won't be effective.
It is usually considered preferable to do most of the phase correction at the plate of the first stage, i.e. to get as close as possible to a flat frequency response from 20Hz to 20kHz with no peak at high frequency. You then apply a shunt capacitor across the feedback resistor, as necessary, to trim th eHF response and get rid of any remaining ringing on square waves. Crowhurst took the view that the common practice of using only a FB resistor shunt capacitor is a poor solution, because it can mask what's really going on inside the amp.
kstlfido,
Are you sure the screen voltage was stable? Is it possible that the tubes you were testing were defective?
From your graphs the control grid voltages are not clear, which makes it difficult to understand what they're trying to say.
Those are real, genuine plain tetrode curves - they're awful! They're behaving as though there were no screen at all. Beam tubes that behaved like that would never reach the market!
Are you sure the screen voltage was stable? Is it possible that the tubes you were testing were defective?
From your graphs the control grid voltages are not clear, which makes it difficult to understand what they're trying to say.
Hi ray_moth-
Yes it was a surprise to me as well.. I curved six 815's.... ALL had the screen-grid tetrode 'knee'. It's like there are no beam forming plates at all!
As for stability; I suppose I could monitor it. But for a check, I ran a 6L6 after the 815's and the curves came out nice and flat. I've tested EL34's, KT's... the curve tracer has always appeared to be in good working order.
I have appx 12 more used ones I can run through. If one keeps operation out of the 'knee', wouldn't it be OK?
Possibly the kink in the curve didn't pose a problem for the 815's designed purpose- a class C RF amplifier.
Yes it was a surprise to me as well.. I curved six 815's.... ALL had the screen-grid tetrode 'knee'. It's like there are no beam forming plates at all!
As for stability; I suppose I could monitor it. But for a check, I ran a 6L6 after the 815's and the curves came out nice and flat. I've tested EL34's, KT's... the curve tracer has always appeared to be in good working order.
I have appx 12 more used ones I can run through. If one keeps operation out of the 'knee', wouldn't it be OK?
Possibly the kink in the curve didn't pose a problem for the 815's designed purpose- a class C RF amplifier.
qq said:
qq- Um, yep. Something around 13.5 watts! I am pushing them. But only for a moment. I measured screen current for the highest curve plot- it spiked to 60 mA then settled down and pulled some 30 mA (6.75 w). The lesser curves pulled less, as expected. Can you measure your screen current? I am curious.
I checked the Sofia screen voltage. 224.2 volts. Rock solid throughout the test. I have curved another eight 815's; all show the tetrode 'kink'- some more than others. One was so 'kinky' it literally shut the tube down!
I am curious as to why my tests really don't match RCA's data. The screen is pulling much more current than the RCA data says. Something's amiss.
I will play around with the Sofia with lesser screen voltages. I gather this tube was not primarily designed for audio but for class C RF- kinks and such were less of an issue. Is it possible the RCA data might be wrong???I've never had the screen draw over 20mA... but note that I am not ever driving the grids positive.
I've been messing around with screen supplies so I don't have consistent data to plot, but with screen at 170V, bias at around -20v, and driving it with a 1khz sine wave to just before clipping, the screen draws about 15mA (2.5W) and I am getting about 12 watts of clean output power.
I don't currently have a stiff enough screen supply to try higher voltages, but I'll let you know when I do..
more curves
OK- second batch of curves. I think I know why they looked so kinky before. I got a little heavy handed with my screen voltages! I also measured the screen current as they were tested. The meter would jump to a peak then gradually go down as each curve was drawn. It did this 5 times for five curve traces per test.
These are with both sections tied together.
Much more sane current numbers- Since the curves start from left to right, the peak currents below occur early in the trace; (under 100 volts on the plate) and one would rarely operate the tube there.
Curve 1-
Plate 450v; screen 170v.
-15.0v curve: 36mA peak
-17.5v curve: 29mA peak
-20.0v curve: 21mA peak
-22.5v curve: 14mA peak
-25.0v curve: 09mA peak
Curve 2-
Plate 450v; screen 150v.
-15.0v curve: 24mA peak
-17.5v curve: 17mA peak
-20.0v curve: 11mA peak
-22.5v curve: 06mA peak
-25.0v curve: 03mA peak
Curve 3-
Plate 450v; screen 125v.
For this test I changed the bias points to -08v to -18v. I wanted to see what I could get closer to 0v grid. I did not take current readings for this test. However, this has the same screen current limiting as the first two. So I approximate the max current on the screen to be ~35 to 40mA.
First thing I noticed- tetrode kink is less- much less! So keeping screen voltages on this tube on the 'lighter' side is a good thing!
OK- second batch of curves. I think I know why they looked so kinky before. I got a little heavy handed with my screen voltages! I also measured the screen current as they were tested. The meter would jump to a peak then gradually go down as each curve was drawn. It did this 5 times for five curve traces per test.
These are with both sections tied together.
Much more sane current numbers- Since the curves start from left to right, the peak currents below occur early in the trace; (under 100 volts on the plate) and one would rarely operate the tube there.
Curve 1-
Plate 450v; screen 170v.
-15.0v curve: 36mA peak
-17.5v curve: 29mA peak
-20.0v curve: 21mA peak
-22.5v curve: 14mA peak
-25.0v curve: 09mA peak
Curve 2-
Plate 450v; screen 150v.
-15.0v curve: 24mA peak
-17.5v curve: 17mA peak
-20.0v curve: 11mA peak
-22.5v curve: 06mA peak
-25.0v curve: 03mA peak
Curve 3-
Plate 450v; screen 125v.
For this test I changed the bias points to -08v to -18v. I wanted to see what I could get closer to 0v grid. I did not take current readings for this test. However, this has the same screen current limiting as the first two. So I approximate the max current on the screen to be ~35 to 40mA.
First thing I noticed- tetrode kink is less- much less! So keeping screen voltages on this tube on the 'lighter' side is a good thing!
Attachments
Those seem much more reasonable, though the old RCA data sheets don't show any kink even down below 100v. They give graphs for 125 and 200v screen voltages.
The screen currents you're measuring are close to what I'm getting in practice.
I'm getting pretty good results using the williamson PI/driver setup and 815's (not paralleled) - after some adjusting with folks' help on here, frequency response is very flat from around 30Hz to as high as I can generate with my laptop... the amplitude on the scope doesn't vary visibly across the whole sweep, 1khz at full power gives a second harmonic down 45-50dB and third close to that, and fourth lost in the noise.
I don't have a plate transformer big enough to experiment with going to higher power by driving the grids positive, but I've ordered some FETs for source followers and will scrounge the surplus shops to try to find a bigger transformer.
I'm listening to records on it now. There's a whole lot of bass, and I'm wondering if I want to filter the input a tiny bit to keep stuff like turntable rumble under control, as it's quite audible. Sounds good to me, but I don't necessarily have the speakers (or the ears) of a True Audiophile.
The screen currents you're measuring are close to what I'm getting in practice.
I'm getting pretty good results using the williamson PI/driver setup and 815's (not paralleled) - after some adjusting with folks' help on here, frequency response is very flat from around 30Hz to as high as I can generate with my laptop... the amplitude on the scope doesn't vary visibly across the whole sweep, 1khz at full power gives a second harmonic down 45-50dB and third close to that, and fourth lost in the noise.
I don't have a plate transformer big enough to experiment with going to higher power by driving the grids positive, but I've ordered some FETs for source followers and will scrounge the surplus shops to try to find a bigger transformer.
I'm listening to records on it now. There's a whole lot of bass, and I'm wondering if I want to filter the input a tiny bit to keep stuff like turntable rumble under control, as it's quite audible. Sounds good to me, but I don't necessarily have the speakers (or the ears) of a True Audiophile.
Now using a LR8N3 based regulator to get screen supply from B+, and also an independent regulated bias supply. The screen seems to stay rock solid, and draws little enough power that the regulator's pass transistor just a little warm, even dropping 200+ volts.
Can't hear any hum at all now, even with my ear up against the speaker cone.
There's a tiny amount of high frequency noise... where does that come from? Semiconductor noise from the grid and screen regulators? The PSU rectifier diodes? thermal noise in resistors? It's definitely not enough to care about at all, but I'm just curious where it originates.
I'll post a full schematic in its final state..
Next is trying it with a friend's fancy speakers. All I have is my pair of bookshelf Infinity 2000.1s.
eric
Can't hear any hum at all now, even with my ear up against the speaker cone.
There's a tiny amount of high frequency noise... where does that come from? Semiconductor noise from the grid and screen regulators? The PSU rectifier diodes? thermal noise in resistors? It's definitely not enough to care about at all, but I'm just curious where it originates.
I'll post a full schematic in its final state..
Next is trying it with a friend's fancy speakers. All I have is my pair of bookshelf Infinity 2000.1s.
eric
Well, I have still been thinking about my 815 amp project. I have discovered it is very close to 2 x 2E26's in one bottle.
The big stumbling block is the screen voltage. I broke apart a dead 815 to examine.
Pics-
http://kentleech.com/815-chop/
Sure enough, the screen is appx. the same distance to the control grid as the control grid is to the cathode. At least it is aligned with the grid.
I guess this rings true with what my curves earlier show- the screen needs very little voltage to effect operation. I suppose this was for efficiency in RF applications.
The only way I see operating it in ultra-linear is to find an OPT with separate screen windings. Acro TO-350's were designed to do this and are rare as hen's teeth. Sowter makes a TO-350 copy for £195!! Plitron makes UL separate screen torroids for about $400. All of my cool vintage OPT's are tapped screens.
Then I stumbled across this link-
http://locofonic.alphalink.com.au/valves.htm
scroll down 1/3 of the page.
Hmm, putting a pot between B+ and the plate (parallel to the OPT pimary) and the pot tap to the screen. Oh, and putting a cap between pot and plate.
Would it work? Who knows. Worth trying out.
The big stumbling block is the screen voltage. I broke apart a dead 815 to examine.
Pics-
http://kentleech.com/815-chop/
Sure enough, the screen is appx. the same distance to the control grid as the control grid is to the cathode. At least it is aligned with the grid.
I guess this rings true with what my curves earlier show- the screen needs very little voltage to effect operation. I suppose this was for efficiency in RF applications.
The only way I see operating it in ultra-linear is to find an OPT with separate screen windings. Acro TO-350's were designed to do this and are rare as hen's teeth. Sowter makes a TO-350 copy for £195!!
Plitron makes UL separate screen torroids for about $400. All of my cool vintage OPT's are tapped screens.Then I stumbled across this link-
http://locofonic.alphalink.com.au/valves.htm
scroll down 1/3 of the page.
Hmm, putting a pot between B+ and the plate (parallel to the OPT pimary) and the pot tap to the screen. Oh, and putting a cap between pot and plate.
Would it work? Who knows. Worth trying out.
You can't try UL with a single 815 because the screen grids are tied together. You can't even try cathode feedback (the other kind of distributed load) because the cathodes are also tied together. However, if you use a pair of 815s, the tetrodes in each 815 being connected in parallel, then these restrictions no longer apply.
I think you may be able to find an OPT with separate screen winding from Lundahl, but I can't remember. I know they have a cathode feedback OPT.
I don't believe it could work, it's just wishful thinking. If it were as simple as that, it would already be common practice. You can achieve UL operation of small signal pentodes with resistor networks but not power tubes because they need a very low impedance to feed the screens.
I think you may be able to find an OPT with separate screen winding from Lundahl, but I can't remember. I know they have a cathode feedback OPT.
I don't believe it could work, it's just wishful thinking. If it were as simple as that, it would already be common practice. You can achieve UL operation of small signal pentodes with resistor networks but not power tubes because they need a very low impedance to feed the screens.
schematic
I feel bad, it's been a year and a half since my last update on this - I ended up with a job that's kept me pretty busy. But better late than never, I figured I'd post an actual schematic of the most recent state of the 815 amp.
In looking at the actual amp to draw the schematic I am noticing various stuff where I'm not sure why I did it, such as the lack of a bypass cap on the cathode resistor of the input stage, and the only 5:1 ratio of coupling cap values between the stages.
It is currently using solid-state regulated supplies for screens and independently adjustable output tube grid bias (cathodes grounded, ~ -30ish volts on grids) I've adjusted thescreen supplies and grid bias for an operating point with idle plate currents of about 15mA.
Frequency response is fairly flat and it doesn't oscillate. 10 khz square wave input shows minor overshoot but no ringing.
http://xymoc.com/~eric/815.gif
(apologies for any errors in the schematic, I've probably missed some)
eric
I feel bad, it's been a year and a half since my last update on this - I ended up with a job that's kept me pretty busy. But better late than never, I figured I'd post an actual schematic of the most recent state of the 815 amp.
In looking at the actual amp to draw the schematic
I am noticing various stuff where I'm not sure why I did it, such as the lack of a bypass cap on the cathode resistor of the input stage, and the only 5:1 ratio of coupling cap values between the stages. It is currently using solid-state regulated supplies for screens and independently adjustable output tube grid bias (cathodes grounded, ~ -30ish volts on grids) I've adjusted thescreen supplies and grid bias for an operating point with idle plate currents of about 15mA.
Frequency response is fairly flat and it doesn't oscillate. 10 khz square wave input shows minor overshoot but no ringing.
http://xymoc.com/~eric/815.gif
(apologies for any errors in the schematic, I've probably missed some)
eric
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.