Yeah, I was a little disappointed I couldn't play with ultralinear since the screens are connected together internally... Using a single 815 p-p I'm getting a little over 10 watts at <2% THD (as measured by laptop and random software, so take with grain of salt
Twice that would be a pretty loud little amp.. Gas regulator tubes would indeed be cool looking, but I think you'd need two of them - with a B+ of 400v, you need to drop at least 180v to get down to a safe screen voltage for the 815, so maybe a pair of 0B3s.
From your schematic and also from the tube data, which is not very clear, I gather that the screens (as you say) and the cathodes are internally strapped together. That means that if you use cathode bias it forces you into Class A because the 2 beam tetrodes in the same envelope cannot be given individual cathode resistors. It also means that you can't use UL.
However, for tetrode-mode class AB1, AB2, B1 or B2 it should work fine, with the cathodes grounded, neg. bias voltage on the grids and the screens fed from a low-impedance voltage source.
You will need to devise a method of measuring plate current, for balancing purposes, because you won't be able to use the popular and safe method of measuring voltage drop across cathode resistors, since the cathodes are strapped internally together. Be cautious, because both leads of your meter will be at high voltage!
Just a thought - the 815 is intended for transmitter use, which entails intermittent bursts of power. You might find it gets too hot under continuous use, especially with two tetrodes in the same envelope. This is one very good reason to avoid Class A. Other transmitter tubes, like the 6146, used to come to grief for the same reason in audio use.
However, for tetrode-mode class AB1, AB2, B1 or B2 it should work fine, with the cathodes grounded, neg. bias voltage on the grids and the screens fed from a low-impedance voltage source.
You will need to devise a method of measuring plate current, for balancing purposes, because you won't be able to use the popular and safe method of measuring voltage drop across cathode resistors, since the cathodes are strapped internally together. Be cautious, because both leads of your meter will be at high voltage!
Just a thought - the 815 is intended for transmitter use, which entails intermittent bursts of power. You might find it gets too hot under continuous use, especially with two tetrodes in the same envelope. This is one very good reason to avoid Class A. Other transmitter tubes, like the 6146, used to come to grief for the same reason in audio use.
I think it's the baem plates (suppressor) and the cathode which are tied together.....
qq- I was also thinking fixed bias. That and gas regulation is not too complicated. Albeit I haven't built mine yet!
ray_moth- if one follows the 815's CCS ratings and keeps plate wattage down from maximum, shouldn't that be OK?
qq- I was also thinking fixed bias. That and gas regulation is not too complicated. Albeit I haven't built mine yet!
ray_moth- if one follows the 815's CCS ratings and keeps plate wattage down from maximum, shouldn't that be OK?
Firstly, I realize that what I posted could be ambiguous.
What I meant to say was that, from the RCA data sheet I've seen, the screens are strapped together; in addition, the cathodes are also strapped together. (Cathode is also strapped to beam plates, as you mention, but that's always the case with beam tetrodes). I certainly did NOT mean that the screens are strapped to the cathodes!
As to the safe margin for continuous running of 815s, it's all a matter of degree. As I mentioned, the 6146, a powerful and popular tube for modulator use, did not tolerate the continuous heat when it was used in audio amps - the base eventually became loose. I don't know what level of dissipation would have been comfortable for it in audio use. The same MIGHT be true of the 815, I don't know, but the presence of two power beam tetrodes in a single envelope suggests to me that there could be problems in this regard.
For safety's sake, I would avoid Class A, which is at its hottest with no signal. Notice that, for CCS operation, the suggested quiescent plate current is only 20mA, whereas the full signal plate current is 150mA. This suggests a very cool bias setting, almost in Class B.
What I meant to say was that, from the RCA data sheet I've seen, the screens are strapped together; in addition, the cathodes are also strapped together. (Cathode is also strapped to beam plates, as you mention, but that's always the case with beam tetrodes). I certainly did NOT mean that the screens are strapped to the cathodes!
As to the safe margin for continuous running of 815s, it's all a matter of degree. As I mentioned, the 6146, a powerful and popular tube for modulator use, did not tolerate the continuous heat when it was used in audio amps - the base eventually became loose. I don't know what level of dissipation would have been comfortable for it in audio use. The same MIGHT be true of the 815, I don't know, but the presence of two power beam tetrodes in a single envelope suggests to me that there could be problems in this regard.
For safety's sake, I would avoid Class A, which is at its hottest with no signal. Notice that, for CCS operation, the suggested quiescent plate current is only 20mA, whereas the full signal plate current is 150mA. This suggests a very cool bias setting, almost in Class B.
ray_moth said:
A very simple (and somewhat surprising) method of measuring plate current is to attach your (well insulated !) hand meter across the respective plate winding of the OPT.
The internal resistance of the meter being much lower than the one of the winding, almost all the current flows through the meter, not trough the winding and the result is precise.
Not valid with signal applied because the tube(s) has/have no longer a correct load but sees a short circuit.
Yves.
Yes, the cathodes are strapped together, the screens are strapped together, and the cathodes are strapped to the beam plates. The only separate parts are the grids and plates. Given that, I don't actually see how you could do any balancing of the tubes, other than just making sure the outputs of the phase splitter/drivers are equal in amplitude.
Also, Iunderstand that one doesn't use cathode bias if one is going to go outside of class A, but I don't intuitively understand why. Is it because if the tubes are allowed to go into cutoff, the balance between the plate currents on each tube isn't maintained, and the cathode resistor's voltage drop changes?
I.e. in class A, if tube 1 draws more current, tube 2 will be drawing proportionally less current, causing the overall draw on the shared cathode resistor to be the same.. but once one tube goes into cutoff, it can't keep decreasing as the other tube increases, so the overall current goes down, thus lowering the bias voltage?
I'm guessing that leads to crossover distortion, but that's pure speculation.
thanks for all the help.
qq
Also, Iunderstand that one doesn't use cathode bias if one is going to go outside of class A, but I don't intuitively understand why. Is it because if the tubes are allowed to go into cutoff, the balance between the plate currents on each tube isn't maintained, and the cathode resistor's voltage drop changes?
I.e. in class A, if tube 1 draws more current, tube 2 will be drawing proportionally less current, causing the overall draw on the shared cathode resistor to be the same.. but once one tube goes into cutoff, it can't keep decreasing as the other tube increases, so the overall current goes down, thus lowering the bias voltage?
I'm guessing that leads to crossover distortion, but that's pure speculation.
thanks for all the help.
That's probably because you're thinking in terms of cathode bias. If you think of fixed bias, then you can adjust the DC balance by adjusting the individual negative voltages on the grids.
The cathodes are connected to ground and a negative voltage is applied to each grid via the 'grid leak' resistor (or via the driver, if using direct-coupled cf drivers). An arrangement of potentiometers is used, so that the bias is adjustable. You can 'dial in' the amount of negative grid bias at each grid, to obtain the required current through the respective plate and to balance them so that the current is the same for each.
Yes, that's right. With 'mild' class AB1 (i.e. close to class A) you can get away with cathode bias, if you use separate cathode resistors and each is bypassed by a large value capacitor (this is impossible with an 815 tube, because the cathodes are strapped together). The cap holds its charge long enough for the differing current at cutoff to have negligible effect on the bias voltage. However, it can't work for deep class AB or class B, because the excursion of the tube into cutoff lasts too long for the cap to be able to 'cover up' for it.
Not crossover distortion, but non-linearity: as the bias changes, so does the internal resistance of the tube. Crossover distortion is caused by a discontinuiity at zero volts (the crossover point), if one tube doesn't start to conduct in time before the other cuts off. It is a danger with Class B and is avoided by raising the quiescent currents of the tubes a little. Crossover distortion is particularly objectionable because it is not dependent on the signal magnitude, therefore it becomes highly noticeable with low level signals.
If you wnat to read more about this stuff (and more), I can recommend the technical articles at Aiken's site
I'm playing with fixed bias at about 20mA per unit. Initial results are very bad - awful distortion as soon as I go past pure class A. This is with about -30v fixed bias, and feeding the grids with very clean and well matched 20v p-p drive signals.
I'm pretty sure the problem is with my screen supply. Nominally it's at 220v, but it's just a voltage divider off B+, and as soon there's any substantial signal, it sags down to like 190 or less and the signal at the plates looks like this:
http://mexico.limpoc.com/~eric/bothplates.jpg
I'm thinking of using zener diodes for the screen - DigiKey has 47v ones rated for 30mA, so a string of four would get it down to ~212v at well over the rated max screen dissipation of 4.5W. Gas tubes would look cool but I have too many tubes already.
It worked much better when using shared cathode bias, but I was barely straying out of class A then. I guess going further toward B starts demanding screen current?
Anyway, I imagine with a stiff screen supply there will be a lot more power output available.
I've been reading the articles at aikenamps.com and they've been very helpful.
I'm pretty sure the problem is with my screen supply. Nominally it's at 220v, but it's just a voltage divider off B+, and as soon there's any substantial signal, it sags down to like 190 or less and the signal at the plates looks like this:
http://mexico.limpoc.com/~eric/bothplates.jpg
I'm thinking of using zener diodes for the screen - DigiKey has 47v ones rated for 30mA, so a string of four would get it down to ~212v at well over the rated max screen dissipation of 4.5W. Gas tubes would look cool but I have too many tubes already.
It worked much better when using shared cathode bias, but I was barely straying out of class A then. I guess going further toward B starts demanding screen current?
Anyway, I imagine with a stiff screen supply there will be a lot more power output available.
I've been reading the articles at aikenamps.com and they've been very helpful.
The screen current in Class AB will vary by a similar factor to the plate current, as you increase the signal. The screen voltage must be kept constant, or the results will be disappointing (or worse!). A voltage divider is unsuitable for feeding the screens, even if you do bypass it with a cap, because it doesn't have a low enough impedance.
The 'purist' approach would be to provide a separate low impedance power supply specially for the screens, using a choke input filter. However, that takes up more space and it's heavy and costs money.
Nowadays, regulation of the screen supply is an effective, convenient and inexpensive solution, thanks to the availability of solid state devices that can operate safely at the high voltages involved. I think it's best to do as SY suggests (or choose some other suitable solid state regulator design). There have been a number of threads on this recently, for example . this one or this one.
The 'purist' approach would be to provide a separate low impedance power supply specially for the screens, using a choke input filter. However, that takes up more space and it's heavy and costs money.
Nowadays, regulation of the screen supply is an effective, convenient and inexpensive solution, thanks to the availability of solid state devices that can operate safely at the high voltages involved. I think it's best to do as SY suggests (or choose some other suitable solid state regulator design). There have been a number of threads on this recently, for example . this one or this one.
I am assuming that if one is not planning on driving grid current, one wants the screen voltage as high as the tube allows, so that you can have the bias voltage as large as possible for your chosen idle current, so a bigger drive signal will fit in between the negative bias voltage and ground.
Are there other things to consider about screen voltage?
Playing with a screen voltage of ~160v, I find I need to bias the output tubes at something like -20v to get a reasonable idle current; feeding a ~35v p-p drive signal works ok but produces very little output power and more distortion than with pure class A.
Possibly I should just quit fooling around with this and build a nice easy project someone else has already designed...
Are there other things to consider about screen voltage?
Playing with a screen voltage of ~160v, I find I need to bias the output tubes at something like -20v to get a reasonable idle current; feeding a ~35v p-p drive signal works ok but produces very little output power and more distortion than with pure class A.
Possibly I should just quit fooling around with this and build a nice easy project someone else has already designed...
Your assumption makes sense to me. Without trying to drive grid current, the maximum amplitude of the grid signal you can have is from whatever the grid bias point is to 0v. The screen and grid are basically acting in opposition. The higher the screen voltage, the more negative the grid bias must be in order to attain the required quiescent current. Thus a higher screen voltage allows a bigger signal to be fed to the grid.
The way you have set up the amp seems reasonable. You should be OK with 150v on the screens. -20v bias also seems reasonable for that screen voltage, but you don't mention what the quiescent current is.
There are a number of possible reasons why the results you're getting might not be satisfactory. The linearity (or otherwise) of the 815 is one of them. It's a bit of an unknown for audio use, AFAIK, so we don't know what to expect. Another possible problem is if the screen voltage is insufficently stable (it must be VERY stable).
One question is whether your OPT gives the required plate-plate load impedance (6.2k) and whether it is good enough to handle the unequal primary currents of deep Class AB. Some OPTs sound perfectly OK in Class A but start to saturate in AB. You could try allowing quiescent current to be a little higher, to see if that helps.
The way you have set up the amp seems reasonable. You should be OK with 150v on the screens. -20v bias also seems reasonable for that screen voltage, but you don't mention what the quiescent current is.
There are a number of possible reasons why the results you're getting might not be satisfactory. The linearity (or otherwise) of the 815 is one of them. It's a bit of an unknown for audio use, AFAIK, so we don't know what to expect. Another possible problem is if the screen voltage is insufficently stable (it must be VERY stable).
One question is whether your OPT gives the required plate-plate load impedance (6.2k) and whether it is good enough to handle the unequal primary currents of deep Class AB. Some OPTs sound perfectly OK in Class A but start to saturate in AB. You could try allowing quiescent current to be a little higher, to see if that helps.
It's a Hammond 1656, 6000 ohm transformer, which was the closest I could find.
Btw, I'm a full believer regarding stiff screen supplies now - I can watch the output go from ugly to perfect sine wave just by removing a 100 ohm resistor in the screen supply output.
Idle current at 20mA per section, screen is at 170V (draws 1.2mA idle and 7mA at full drive) Full power gives a 28v swing into 8 ohms, which if my math is correct is about 12 watts. Does this all sound reasonable?
Depending on how I microscopically tweak the grids while watching a spectrum display, the harmonic output shifts around, but odd ones predominate. As a typical example, at full power, just before clipping: second down 44dB, 3rd -34, 4th -62, 5th -52, 6th -57, and 7th -56. http://mexico.limpoc.com/~eric/spectrum.gif
eric
p.s. I'll also mention that the 815 tubes I'm using are very used.
I already put one aside because the two sections were so imbalanced that to get the same idle current, the bias voltages had to differ by like 50%. Since they're really cheap, I ordered a NOS pair just to see...Actually, I don't know if odd harmonics are predominating. I'm trying to use some PC software to look at the spectrum output and the results are not very repeatable. I think the PC is feeding the input back into the output, etc, causing havoc. When i fix that, I see second harmonics down around 40dB, and third around 44dB. I have no idea if this is good or bad, or even if it's accurate.
p.s. also, if i connect global nfb it picks up radio stations. very quietly, though.
p.s. also, if i connect global nfb it picks up radio stations.
very quietly, though.
That should be OK, I would think.
The screen currents seem a bit lower than I would expect but I'm not familiar with this tube. By 28v swing, what do you mean? Do you mean peak-to-peak, amplitude or RMS? (Respectively, these would mean 12w, 49w and 98w.) Since you're not venturing into grid current (AB2) I would think 12w is reasonable.
They should be. Even order harmonics created in well balanced push-pull stages get cancelled, but odd order harmonics don't. Pentodes/beam terodes produce not only 3rd order but also 5th, 7th, 9th etc. order harmonics. The higher the order, the worse it sounds and that is one reason pentodes/beam tetrode amps should treated to a generous amount of NFB. (The other main reason is to lower the OP impedance, to get decent damping for the speakers.)
Aha, the dreaded RF! Picked up by the speaker leads acting as antennas and fed directly to the IP tube cathode (or grid), courtesy of the speed-up capacitor across the global NFB resistor. Many don't believe in this problem but Norman Koren does.
How are your results shaping up now? Are your screen voltages stable yet?
Are you ready to take the AB2 plunge? You'll get much more power if you do, but you'll need decent drivers to do it, to avoid distortion, as already said. SY has had success using IRF820 MOSFETS as source follwer drivers, which are a much more convenient solution than cathode followers. He might be able to give you some tips on how to go about it, if you're interested.
Yeah, it does - in fact one of the typical operation conditions lists on the datasheet is for class AB2 audio amplification: 20mA idle plate current, 150mA max signal plate current (at 400V), -15v fixed bias, for 42 watts out, requiring .36 watts grid drive power. I would assume this is meant as a transmitter modulator.
How elaborate is the MOSFET source follower? I know even less about that stuff than I do about tubes, but I'd certainly be willing to try. I'll need a larger power transformer, though, since the one I have is only good for ~150mA.
Results look pretty good now as long as I keep the idle plate current toward the high end at ~20mA. I'm currently using a separate screen supply with choke input, perfectly quiet and solid, but it's a lot of iron.
I'd rather regulate down from B+, but 400V down to 170V (or lower, for AB2) might be too big a input/output difference.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.