I built that amp probably 20 years ago. It worked, but I did not like the sound. I now know it was due to an extremely high output impedance. I run the mosfet driver's drain from a higher voltage source through a resistor or CCS chip for current limiting. Bypass the drain with a capacitor large enough to supply enough current to handle a fat transient, but not large enough to melt a grid. I do the same for the screen supply on sweep tubes.
6JH5 , So similar.
Let's see: 6HZ5/6JD5 $5, 6JK5 $6, and 6JH5 $8
That settles that. 6HZ5 is the best tube. (checked so far)
Let's see: 6HZ5/6JD5 $5, 6JK5 $6, and 6JH5 $8
That settles that. 6HZ5 is the best tube. (checked so far)
Last edited:
For practical LV use, you're looking at 7500 Ohm Rp with these. Good luck finding a 23000 Ohm OT for triode loading use. I guess some N FDBK could help, but then its just like using a regular Beam Power tube. Or triple them up. At least they look nice without the plate caps and are cheap for a 35 Watt TV tube. $30 for 6 tubes gets you a 300 Watt amplifier if you can find a suitable OT.
If they had just put the usual grid 2 into these, they would be blockbusters. And probably all sold out by now.
If they had just put the usual grid 2 into these, they would be blockbusters. And probably all sold out by now.
Last edited:
Oops, correction,
6 tubes in class AB -Triode- mode would only get like 150 Watts.
The tubes are so linear, it looks like one could just draw a load line orthogonal to the triode curves. So a 4K load line for each tube.
Roughly a 16K OT for two P-P tubes, or more practically an 8K OT for 4 (2+2) tubes. Roughly, check those #s.
6 tubes in class AB -Triode- mode would only get like 150 Watts.
The tubes are so linear, it looks like one could just draw a load line orthogonal to the triode curves. So a 4K load line for each tube.
Roughly a 16K OT for two P-P tubes, or more practically an 8K OT for 4 (2+2) tubes. Roughly, check those #s.
Last edited:
I would definitely try to reduce the effective plate resistance by using feedback. If there is gain to burn, then I'll burn it.
Thanks for your explanation, George! Now I got it how regulation on the primary side was done
.Interestingly here in Europe the timeline of HV regulation was the other way round: When color TV's appeared at the market from about 1967 on, any set had it's HV regulated by a ballast triode PD500 or PD510 in parallel with the CRT. A few years later the manufacturers had learned to drive the sweep tube accordingly to keep the acceleration voltage (and image width) constant, hence decreasing the set's power consumption some bit. Dedicated tubes for this purpose like the ones in this discussion were unknown here.
Best regards!
Since the 6JK5 did not have very good curves for the medium high plate voltage (and + grid1) test, I dissected it. Appears to be the same as the other 6HZ5/6JD5 and 6JH5 tubes EXCEPT for 1 thing. It has the boxlet mods to the plate (extended inside box section attached to the crimped plate supports, just outside the beam forming sheet slot) , without the Barkhausen fins usually inside them. While the other tubes have conventional flat plates with no Barkhausen fins. Without the Barkhausen fins inside the boxlets, the plate is effectively more than twice as far away from the grid 1. No wonder they don't work well at reduced plate V. To make matters worse, I see International brand 6HZ5/6JD5 WITH the boxlets. So beware of those.
Some time down the line, I'm really going to need to sit down with a small pile of 6HV5as and plot the plate curves at ~700V plate voltage and positive grid bias via a mosfet follower. Since I don't have SA's nifty-doodle curve tracer, I'll have to do things the old, tedious way and plot the results in Excel, just like I did with the screen-driven 6CD6. I'll include a small sense resistor between the follwer and 6HV5A grid so I can also keep track of the grid current as a function of drive. I need to cruise back to Vinylsavors's site and look at his dissection of a 6HV5A to see if I can find any Barkhausen stubs on the beam forming plates.
Here's Vinylsavor's 6HV5A dissection - from the size of the plate it looks like he's doing up a GE or Westinghouse tube. It also looks like there are no Barkhausen fins inside the beam forming plates.
http://vinylsavor.blogspot.com/2013/06/tube-of-month-6hv5a.html
Here's Vinylsavor's 6HV5A dissection - from the size of the plate it looks like he's doing up a GE or Westinghouse tube. It also looks like there are no Barkhausen fins inside the beam forming plates.
http://vinylsavor.blogspot.com/2013/06/tube-of-month-6hv5a.html
Last edited:
Here is a pic of the 6JK5 Beam Triode plate with the "boxlets" (or cavities) but missing the Barkhausen fins normally inside them. (had poor reduced V curves)
Then a pic of some TV Sweep tube plates, with common "boxlets" and typical Barkhausen fins inside in the right side column. Bottom right plate what would normally be expected here.
Top left plate is similar to the plain 6HZ5/6JD5, 6JH5 Beam Triode plates. (which had good reduced V curves)
So looks to me that one has to be quite selective to get Beam Triodes that work well at reduced B+ and positive grid 1 excursion. Most of the Beam Triodes I see on Ebay have the dumb "boxlets".
Then a pic of some TV Sweep tube plates, with common "boxlets" and typical Barkhausen fins inside in the right side column. Bottom right plate what would normally be expected here.
Top left plate is similar to the plain 6HZ5/6JD5, 6JH5 Beam Triode plates. (which had good reduced V curves)
So looks to me that one has to be quite selective to get Beam Triodes that work well at reduced B+ and positive grid 1 excursion. Most of the Beam Triodes I see on Ebay have the dumb "boxlets".
Last edited:
Oh, I also tried overheating (over voltage) the 6JK5 HTR (since they were spec'd at less current than the other tubes), but that didn't help the curves.
A couple of those could probably replace the 211 in the SE amp I've got on the bench quite nicely. They won't look nearly as sexy, though.
I've tried hard to get less than .01% @ 1W in an SE amp (and never quite made it) and these might just be the way to do it.
I've tried hard to get less than .01% @ 1W in an SE amp (and never quite made it) and these might just be the way to do it.
They might do it. The OT might be the dist. problem though at that level. Might need the positive (current derived) Fdbk scheme to eliminate the magnetizing current effect (via neg. resistance against winding resistance). Audio Precision patent 4614914
Current sense resistor in output cathode, Fdbk resistor to driver tube cathode over auto bias resistor.
Current sense resistor in output cathode, Fdbk resistor to driver tube cathode over auto bias resistor.
Last edited:
I have gotten down to .013% at 1W, but getting any lower than that seems really hard. I have compared distortion on the primary to distortion on the secondary and there isn't much difference. There is always more distortion on the secondary but always only a small fraction of the total distortion. Of course, this is a mid-band distortion measurement where the output transformer is on its best behavior.
As you point out, combining other techniques would probably be fruitful. I still haven't added any feedback around the output transformer. I've just been playing with different tubes and operating points to get distortion cancellation. It's hard to find time to work on this stuff much with teen kids.
As you point out, combining other techniques would probably be fruitful. I still haven't added any feedback around the output transformer. I've just been playing with different tubes and operating points to get distortion cancellation. It's hard to find time to work on this stuff much with teen kids.
Design it to push the distortion down even more, using a non inductive 8 Ohm load resistor.
Or leave it the way it is; in either case try the following:
Then connect a real world loudspeaker instead of the load resistor, and see if the distortion stays that low.
Most loudspeakers can take 1 Watt at 1kHz, but your ears will not, so put the loudspeaker in a big box, and then cover it with a large comforter bed cover.
Please report the results.
Thanks!
Or leave it the way it is; in either case try the following:
Then connect a real world loudspeaker instead of the load resistor, and see if the distortion stays that low.
Most loudspeakers can take 1 Watt at 1kHz, but your ears will not, so put the loudspeaker in a big box, and then cover it with a large comforter bed cover.
Please report the results.
Thanks!
Two speakers of a typically low-ish small box sensitivity of 88dB SPL/1W/1M sum with music uncorrelated like noise to +3dB, so 91dB SPL/1W(each channel)/1M. Subtract from this room loss if not near field (1M): let's call it -10dB or whatever number you've measured in your room. This number is now our 0dBW sensitivity rating, one Watt average per channel (0dBW) makes this much sound. For our example 0dBW equals 81dB SPL at listeners' ears.
Mastering room levels are classically calibrated to 85dB SPL at listeners' ears at 0VU and allow for a professional headroom of 20dB above that. This is a very, very loud level, used by folk who need to hear small errors and are willing to risk their hearing to do so. Live concerts with SR at this SPL are common enough, but you pay for it for some days. I left a Lucinda Williams concert for being only a little louder than this (estimated), and I love Lucinda Williams.
So, where should we be measuring amplifiers' performance, assuming a single number to be most telling? First, we should all campaign to stop talking about WATTS, a linear measure of something that has a log importance. WATTS have been used to sell amplifiers since the dawn of man, and his toys, and the $cost$ varies linearly with WATTS, so no surprise there.
But I'd suggest that we're wrong to continue thinking that way. WATTS, a linear scale, tries to convince us that 50 WATTS is twice as better as 25 WATTS. Of course, it's actually only a 3dB higher peak output capability. Better peak overload handling design is more important. We should be thinking in log terms, the way we hear, and talking not in WATTS, but in dBW.
How does this rant relate to measurements at 1 WATT (0dBW)? It's a really good start, and we could all hope that it becomes part of a new standard. 81dB SPL is still a pretty loud musical level in most folks' homes, approaching peak levels for many of us most times. Noise floors in residences run in the 20 to 35 dB SPL range for folk likely reading this, so your music listening range starts from there, although uncorrelated sound can be heard way below the noise floor. This 60dB-ish range is where the music lies, so that's where we need to be measuring performance.
And how does this 60dB range translate into good old American WATTS? If 0dBW is 81dB SPL at listening position, and we call that a very loud 0VU, then our professional peak amplifier requirement is +20VU = +20dbW. And our music listening range of -60dB below 0VU starts at -60dBW = 1 milliWATT.
It's classically difficult to measure distortion at -20dBW, -40dBW, and -60dBW levels, so we don't. But don't mean they don't matter.
Sorry about the rant,
Chris
Mastering room levels are classically calibrated to 85dB SPL at listeners' ears at 0VU and allow for a professional headroom of 20dB above that. This is a very, very loud level, used by folk who need to hear small errors and are willing to risk their hearing to do so. Live concerts with SR at this SPL are common enough, but you pay for it for some days. I left a Lucinda Williams concert for being only a little louder than this (estimated), and I love Lucinda Williams.
So, where should we be measuring amplifiers' performance, assuming a single number to be most telling? First, we should all campaign to stop talking about WATTS, a linear measure of something that has a log importance. WATTS have been used to sell amplifiers since the dawn of man, and his toys, and the $cost$ varies linearly with WATTS, so no surprise there.
But I'd suggest that we're wrong to continue thinking that way. WATTS, a linear scale, tries to convince us that 50 WATTS is twice as better as 25 WATTS. Of course, it's actually only a 3dB higher peak output capability. Better peak overload handling design is more important. We should be thinking in log terms, the way we hear, and talking not in WATTS, but in dBW.
How does this rant relate to measurements at 1 WATT (0dBW)? It's a really good start, and we could all hope that it becomes part of a new standard. 81dB SPL is still a pretty loud musical level in most folks' homes, approaching peak levels for many of us most times. Noise floors in residences run in the 20 to 35 dB SPL range for folk likely reading this, so your music listening range starts from there, although uncorrelated sound can be heard way below the noise floor. This 60dB-ish range is where the music lies, so that's where we need to be measuring performance.
And how does this 60dB range translate into good old American WATTS? If 0dBW is 81dB SPL at listening position, and we call that a very loud 0VU, then our professional peak amplifier requirement is +20VU = +20dbW. And our music listening range of -60dB below 0VU starts at -60dBW = 1 milliWATT.
It's classically difficult to measure distortion at -20dBW, -40dBW, and -60dBW levels, so we don't. But don't mean they don't matter.
Sorry about the rant,
Chris
I spent about two years off and on playing with Pete Millett's Distortion Cancelling Push Pull amp and multiple variations thereof. It resembled a little breadboard that I had created 20 some years ago with a pair of 6AU6's and a pair of 6AQ5's, so I hacked one up and made all sorts of stuff with it. The distortion cancelling thing does seem to apply over a frequency range that's less than the entire 20 Hz to 20 KHz range and is highly OPT related even without feedback from the secondary side. It is also quite dependent on the signal level used in the test and the idle current in the output tubes.
I also spent considerable time playing with similar tests on a two stage SE amp (an original TSE board with a 300B output tube) incorporating no feedback at all where I discovered that there is an ideal idle current where the third harmonic disappears. I spent several weeks chasing this before I discovered that the "ideal" current varies with the test frequency, power level and OPT being used. Some kind of cancellation was happening, but I never dug much deeper.
I have been saying this for years. Look at it from a different perspective. Say I have an arbitrary amp where clipping starts at 10 watts. We don't want to listen to clipping, so our music peaks must not reach 10 watts. Much of the "stuff" called popular music today is uber-compressed so that the entire dynamic range fist into 20 to 40 dB. Even though most delivery mediums support a much larger dynamic range, compressed music SOUNDS louder, and LOUD sells. So if we crank up the volume, and the bass to just below the clipping point on our 10 watt amp, the AVERAGE power level is 100 milliwatts or less, often far less.
I have bad hearing due to Meniere's Disease and it's getting worse. I often use my Yamaha NS-10M Studio speakers for testing because I have owned them for about 25 years and know their sound. They are 87 or 88 Db sensitivity depending on which spec you read. They used to be in a 10 X 10 foot room when I lived In Florida, but now they are in a 1800 square foot basement with the air handler though my listening position is still 1 to 2 meters from the speakers. Measurements show that my average listening level is well below 1 watt, often closer to 100 mW depending on how the music was processed before delivery.
Very low level distortion measurements are quite difficult with all the noise generated today. Pressing the 400 Hz high pass filter and the 30 KHz low pass filter buttons on the 8903A often drop the THD reading on a 1KHz tone by 20 dB on an unshielded test breadboard test amp. The major HF culprits are the LED shop lights, LED light bulbs and the computer system. Even the TV upstairs makes some noticeable HF noise.
It is still necessary to know "how many watts" an amp makes before it clips, but that number is often inflated or measured under "ideal" conditions. It is more important to know how the amp behaves under all of the usual listening conditions. This is best conveyed with a THD VS power @ several frequencies, and a frequency response VS power at a fairly high power level on a tube amp to expose under sized or poor quality OPT's.
I've traced the 6HZ5 (flat plate) and 6JK5 (boxlet/cavity plate) Beam Triodes again ( I have a 2nd 6JK5 after busting open the other one) with the Beam plate connected to the plate this time for each, instead of to the cathode. This makes for some small improvement in curves for the 6JK5, and a smaller improvement for the 6HZ5. The non boxlet 6HZ5 still has the best curves, but the boxlet 6JK5 is looking a little better than before.
I then measured the current drawn by the beam plate for each tube type when beam plate is connected to the plate, and was surprised to find rather lower than expected beam plate current draw for both tubes. So the scheme looks quite practical (beam plate will not be getting red hot). Pics and beam plate current data below:
Pics (4th curve up is 0V on grid 1) 0.5V grid steps, Vertical 20mA/div., Horiz 100V/div
a) 6JK5 normal connection to cath.
b) 6JK5 beam plate to plate conn.
c) 6HZ5 normal connection to cath.
d) 6HZ5 beam plate to plate conn.
Beam plate to plate conn. seems to increase plate current for all curves and moves the "knees" slightly left.
------------------------------------------------------------------------------------------------------------
The 6HZ5/6JD5/6JH5 curves look good enough to make an amplifier, so I've ordered an 8K OT from Edcor to try these out. (2+2 tubes for class AB P-P with around 750V B+
and 1+1 tubes for class A P-P , probably around 450V B+)
6JK5 Beam Plate Current data: (avg. over plate V sweep)
g3 mA, +g1V plate mA
5.3 mA, 5V, 100 mA
3.35 mA, 2.5V, 69 mA
2.55 mA, 1.25V, 50 mA
6HZ5 Beam Plate Current data: (avg. over plate V sweep)
7.5 mA, 5V, 118 mA
4.5 mA, 2.5V, 79 mA
3.36 mA, 1.25V, 56.7 mA
1.96 mA, 0V, 30.4 mA
I also tried a 2.7K resistor in series with the Beam plate to plate conn. but didn't see much difference. That was just in case the beam plate would need it's V dropped a little below the plate to stay cool, but isn't needed.
I then measured the current drawn by the beam plate for each tube type when beam plate is connected to the plate, and was surprised to find rather lower than expected beam plate current draw for both tubes. So the scheme looks quite practical (beam plate will not be getting red hot). Pics and beam plate current data below:
Pics (4th curve up is 0V on grid 1) 0.5V grid steps, Vertical 20mA/div., Horiz 100V/div
a) 6JK5 normal connection to cath.
b) 6JK5 beam plate to plate conn.
c) 6HZ5 normal connection to cath.
d) 6HZ5 beam plate to plate conn.
Beam plate to plate conn. seems to increase plate current for all curves and moves the "knees" slightly left.
------------------------------------------------------------------------------------------------------------
The 6HZ5/6JD5/6JH5 curves look good enough to make an amplifier, so I've ordered an 8K OT from Edcor to try these out. (2+2 tubes for class AB P-P with around 750V B+
and 1+1 tubes for class A P-P , probably around 450V B+)
6JK5 Beam Plate Current data: (avg. over plate V sweep)
g3 mA, +g1V plate mA
5.3 mA, 5V, 100 mA
3.35 mA, 2.5V, 69 mA
2.55 mA, 1.25V, 50 mA
6HZ5 Beam Plate Current data: (avg. over plate V sweep)
7.5 mA, 5V, 118 mA
4.5 mA, 2.5V, 79 mA
3.36 mA, 1.25V, 56.7 mA
1.96 mA, 0V, 30.4 mA
I also tried a 2.7K resistor in series with the Beam plate to plate conn. but didn't see much difference. That was just in case the beam plate would need it's V dropped a little below the plate to stay cool, but isn't needed.
Last edited:
Next up, I'm going to try summing the grid 1 current back into the plate current, using a 1000V driver Mosfet. (only around +6V grid drive, but the plate is going up to +750 V on the curve tracer. Would need a 2500 to 3000V Mosfet for a real Amp I think (or series chain). I'm curious to see what happens to the triode plate curves then.
Then if you have an effective 3000V Mosfet, you could also double the cathode (and plate) current with a Current Mirror under the cathode. Would halve the Rp of these tubes, making them more attractive. Could also use a Mirror/Cascode Sweep tube slave for this job, and more current multiplication too.
The 6HZ5/6JD5/6JH5 only have a 25V "knee" at lower currents (or 60V knee at 120 mA), so B+ could be dropped considerably with more current multiplication. This will never compete with a plain old Sweep tube though. But maybe some "cachet" here if the triode curves are better than a 300B. Plastic surgery for Beam Triodes.
Then if you have an effective 3000V Mosfet, you could also double the cathode (and plate) current with a Current Mirror under the cathode. Would halve the Rp of these tubes, making them more attractive. Could also use a Mirror/Cascode Sweep tube slave for this job, and more current multiplication too.
The 6HZ5/6JD5/6JH5 only have a 25V "knee" at lower currents (or 60V knee at 120 mA), so B+ could be dropped considerably with more current multiplication. This will never compete with a plain old Sweep tube though. But maybe some "cachet" here if the triode curves are better than a 300B. Plastic surgery for Beam Triodes.
Last edited:
Back in 2006, I ran some experiments with mosfet based current mirrors to make a super 45 triode with one 45 tube and a 4 X current mirror made from 5 mosfets. It had 45 like curves when hand traced but blew up whenever I ran it into an OPT with a speaker connected. This was long before it was known that mosfets also suffer from SOA failures and it used mosfets from the 1990's. We have better fets today, and we know how to stick them in the cathode of a big sweep tube.
My BJT version that ran a 6Y6GA on 140 volts into a 600 ohm OPT did survive and sounded quite nice.
I have a few each of several flavors of these beam triodes. When I get a chance I'll stick some into the unset board and see what happens. It should be possible to tweak the feedback resistors to trade off some Mu for Rp.
My BJT version that ran a 6Y6GA on 140 volts into a 600 ohm OPT did survive and sounded quite nice.
I have a few each of several flavors of these beam triodes. When I get a chance I'll stick some into the unset board and see what happens. It should be possible to tweak the feedback resistors to trade off some Mu for Rp.
