• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Class AB2 - Kink, how and why?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,

I was doing a small audio-amp with Compactrons and ended up with what you see in the schematic attached. Sorry for the rather chaotic way it's been drawn.
In order to extract maximum output I choose to drive the 6AD10's in AB2.
It turned out that this is a though little tube. Spec's say 12W combined A/G2 dissipation but I ran it with Va around 350V and dropping ~26W without red-plating. After the abuse I went to safer grounds as you see in the schematic. Rather fancy having the CF built-into the OP-Tube :D
With some heavy-duty Edcor OPT is sounds very good indeed, but seeing the waveforms on a scope I wonder about just one thing:
What is the small 'kink' you see when you feed the amp a sinewave and amplify the section just about when 6AD10 G1 starts dropping it's impedance way down to a few 100 Ohms? Scope on 6AD10/Grid1.
It occurs when G1 reaches around +1V and is very slight and barely noticable. Obviously a distortion-analyzer will notice it.
The CF in my construction will happily swing G1 to +12-15Vpeak without any further distortion. The 6AD10 give in at around +3 to 4V at G1 so this amplitude is not needed, just notiched.
Setting another CCS-current has no influense on the 'kink' as well as changing Va of the CF.
Actually the ONLY way I found to influence the 'kink' was by using another OPT..?!

Please shed some light, thanks...

rgds,

/tri-comp
 

Attachments

  • Mini_Gee-Tah_v2.6 - Hi-Fi.pdf
    25.3 KB · Views: 264
Last edited:
My first thought is that the driver may not be keeping up. If this is the case, distortion should be detectable on the grid of the output tube. If not, the screen current might be high in the G1 positive region of your load line and the screen supply might be sagging. If both of those are not the case, maybe the tube in question just doesn't have very good positive grid characteristics. That probably isn't likely, though.
 
I think I know what's happening

The 6AD10A section 2 (CF driver) unit has 2500 uMho transconductance. As a cathode follower with CCS the equivalent drive impedance is 400 ohms. This is equivalent to a 400 ohm resistance in series with an ideal voltage source.

As you drive the grid positive, it begins to draw current. The higher the positive grid voltage, the lower the effective impedance. It's like a diode and will reach a few mA of current with just a few volts positive. at 1mA you're already losing 0.4V of drive...

So what is happening is that the slope of the drive signal at the section 1 grid is changing as you drive into the grid current region of A2. This will be audible and seen as both even and odd harmonic distortion.

I'm not sure without testing how much grid current you'll need to source, but 400 ohms seems way to high driver impedance for a practical A2 amp. I would add a power drive circuit. One MOSFET, one resistor, one zener diode to protect the gate. Solder right over the tube socket ;-)

Look up my thread on the Meteor amp for some example grid current measurements and clean class A2 operation.
 
Last edited:
Michael,

CF is being run as a Triode.

I get your point, but it's ONLY the instant when G1 starts to draw current the distortion is apparent.
Immediately after the 'kink' or distortion-point the curves look very normal up to around +14V.

FET-point taken. No biggie if it helps, but then no combined CF/OP-Tube which I think is rather nice.

I'll check-out your Meteor later. At wrk now..
 
Michael,

CF is being run as a Triode.

I get your point, but it's ONLY the instant when G1 starts to draw current the distortion is apparent.
Immediately after the 'kink' or distortion-point the curves look very normal up to around +14V.

FET-point taken. No biggie if it helps, but then no combined CF/OP-Tube which I think is rather nice.

I'll check-out your Meteor later. At wrk now..

I guess I'll need to see the scope shot to be sure. In my experience there is a kink if the grid circuit has significant resistance. The top looks normal because the grid load is progressive and hardly impacts the sine wave shape. If you used a triangle or ramp you may see the slope change.

The MOSFET is about 20 ohms equivalent impedance and will pretty much eliminate the kink. Triode CF + CCS load plus MOSFET follower may be a really nice A2/AB2 driver. You load the previous stage with tube grid, you still have the triode characteristic due to the CF changing plate-grid voltage with signal swing, the MOSFET gate is well driven by the low impedance CF, and it has super low output impedance.

Another benefit of CF + MOSFET for class AB/AB2 drive is the output tube grid need not be driven negative beyond cutoff at the highest plate voltage excursion. If you're using a MOSFET drive you need to provide for the full negative grid excursion anyway because of the -Vgs limit. With a CF plus MOSFET you can get away with a low voltage (-) driver supply rail. I'm planning to use this in my next G1+G2 dual grid drive amp, which needs only positive grid voltage.
 
Last edited:
Sorry for the late posting.
I had 500+ vacation-pic's from The Nile that needed to go from the camera before taking these :D
Anyway, here we go.
Yellow = G1
Blue = Speakerterminals with dummyload, 8R, Full-OP = 12.25W
First 3 pictures show 1KHz Sine with the 'kink' developing when turning up volume
Next 3 shows 1KHz Ramp, same development.

You'll clearly see the distortion in the 'Exaggerated' pictures with volume beyond reasonable.
G1-curves are 'bending' just around +2V and just before the output starts flat-topping.

Next posting shows relations with G2.
 

Attachments

  • AB2-Kink 06_Ramp_Exaggerated.jpg
    AB2-Kink 06_Ramp_Exaggerated.jpg
    399.8 KB · Views: 92
  • AB2-Kink 05_Ramp_Full-OP.jpg
    AB2-Kink 05_Ramp_Full-OP.jpg
    377.7 KB · Views: 345
  • AB2-Kink 04_Ramp_G1-at-0.jpg
    AB2-Kink 04_Ramp_G1-at-0.jpg
    389.4 KB · Views: 359
  • AB2-Kink 03_Sine_Exaggerated.jpg
    AB2-Kink 03_Sine_Exaggerated.jpg
    384.6 KB · Views: 364
  • AB2-Kink 02_Sine_Flat-top.jpg
    AB2-Kink 02_Sine_Flat-top.jpg
    387.3 KB · Views: 387
  • AB2-Kink 01_Sine_Full-OP.jpg
    AB2-Kink 01_Sine_Full-OP.jpg
    380.1 KB · Views: 391
Now, these pictures are historical.
I'll never be able to repeat them, because throwing a 22uF cap at G2 and subjecting the tube to full AB2++ just won't set well with the tube.
Shurely enough it died! So, I'm $1 down ;)

Here are the results.
Yellow = G1
Blue = G2
The 'kink' clearly shows up in G2.
(G2 curves are off-set voltage readings, depicting a max. voltage-drop of 50V minus the B+ sag over 1K = around 40-50mApeak WITHOUT the cap. No telling what beating it took WITH the cap...)

Any ideas?

rgds,

/tri-comp
 

Attachments

  • AB2-Kink_ 01_G1+G2_No-OP.jpg
    AB2-Kink_ 01_G1+G2_No-OP.jpg
    373.2 KB · Views: 242
  • AB2-Kink_ 02_G1+G2_G1-at-0.jpg
    AB2-Kink_ 02_G1+G2_G1-at-0.jpg
    369.7 KB · Views: 228
  • AB2-Kink_ 03_G1+G2_Full-OP.jpg
    AB2-Kink_ 03_G1+G2_Full-OP.jpg
    395.5 KB · Views: 229
  • AB2-Kink_ 04_G1+G2_Exaggerated.jpg
    AB2-Kink_ 04_G1+G2_Exaggerated.jpg
    384.5 KB · Views: 226
  • AB2-Kink_ 05_G1+G2_G2-22uF.jpg
    AB2-Kink_ 05_G1+G2_G2-22uF.jpg
    377 KB · Views: 221
Are you sure your loadline even gets you to useful AB2 output? If the loadline goes under the Vg=0 knee you are just going to really abuse the screen and control grids and get no useful output (tubes will saturate). If that is the case, you may as well put a grid stopper on to control the runaway currents. Just something to check on.

I once triode-connected some KT88s and drove them hard in AB2 with a mosfet until I killed them. What I didn't realize at the time was that once you saturate, you're done. Pushing harder will just destroy the tubes.

Edit: I guess if you have more compression on the amplifier's output than on the driver output, there are probably some saturation issues going on. Hard to tell with the scales on the pictures attached above.
 
Last edited:
Being not absolutely satisfied with this proto-type it became time to take it apart ...or to find out what's killing it!
I decided on the 2nd. option.
The sound quality of this amp is OK, not VERY OK or even brillant, but OK. Parts are not expensive (Except the Edcor OPT) so I set my perspective accordingly.
What irritates me the most is the harshness of the vocals. Bass is really strong, potent, firm. No problems there, but vocals, pianos are ..ahem, NOT triode-sound!
After listening for a while, maybe @5W or thereabouts, you start getting annoyed with the sound. Maybe it's 3rd. hamonics because of the pentode OP, I don't know. On a scope I can't really say what's eating the sound; all seems well way past 20KHz even with ramps and squares. Output drops a little after 22KHz but that's fine.
Altering input-circuits, phase-inv, operating-points of CF/CCS/OP-Tubes makes no difference, there's no way to make this amp sound like my 2C34 triode-PP amp that always has a sweet tone.
As a desperate measure, just before canning the amp I thought why not make it into a triode-PP? I expected output-wattage to drop substantially but hoped for enough to suit my B&W speakers.
After triode-strapping the 6AD10 OP-pentode I had to redesign for separate bias-settings. The tubes aren't matched close enough to get by with a common preset and that shines through with triode-strapping apparently. Obviously that's no problem!
Power-on and ...... :cloud9:
I couldn't believe what had happened! The amp sounds almost as sweet as my 2C34-PP! Gone is the harshness, just sweet ear-pleasing melodic sound. I ran Eagles, Hell Freezes Over at high volume (probably ~ 10W and well into AB2) and enjoyed all of it. Of course the individual recordings are not of the same standard, but I already knew that from my other amp. What SHOULD sound perfect DID!
Something else strange. The output, with the Edcor rewired for 3,8K a-to-a, kept at 18W/4R and 16W/8R. The CF now swings g1 to +12V at full output and the 'kink' is almost gone both in the CF-output and checked at the speaker-terminals. You have to be zoomed-in extensively on my scope to notice it. Frequency-response at full output flat from 18Hz and up to 22KHz whereafter it drops slowly to about 2/3 output at 45KHz. After that I don't care.
That is not half bad for a 6AD10 class-AB2 PP-amp.
Now I'm left with a problem: To build or not to build :D

rgds,

/tri-comp
 

Attachments

  • Mini_Gee-Tah_v3.0 - Hi-Fi.pdf
    30.1 KB · Views: 81
Being not absolutely satisfied with this proto-type it became time to take it apart ...or to find out what's killing it!
I decided on the 2nd. option.
The sound quality of this amp is OK, not VERY OK or even brillant, but OK. Parts are not expensive (Except the Edcor OPT) so I set my perspective accordingly.
What irritates me the most is the harshness of the vocals. Bass is really strong, potent, firm. No problems there, but vocals, pianos are ..ahem, NOT triode-sound!
After listening for a while, maybe @5W or thereabouts, you start getting annoyed with the sound. Maybe it's 3rd. hamonics because of the pentode OP, I don't know. On a scope I can't really say what's eating the sound; all seems well way past 20KHz even with ramps and squares. Output drops a little after 22KHz but that's fine.
Altering input-circuits, phase-inv, operating-points of CF/CCS/OP-Tubes makes no difference, there's no way to make this amp sound like my 2C34 triode-PP amp that always has a sweet tone.
As a desperate measure, just before canning the amp I thought why not make it into a triode-PP? I expected output-wattage to drop substantially but hoped for enough to suit my B&W speakers.
After triode-strapping the 6AD10 OP-pentode I had to redesign for separate bias-settings. The tubes aren't matched close enough to get by with a common preset and that shines through with triode-strapping apparently. Obviously that's no problem!
Power-on and ...... :cloud9:
I couldn't believe what had happened! The amp sounds almost as sweet as my 2C34-PP! Gone is the harshness, just sweet ear-pleasing melodic sound. I ran Eagles, Hell Freezes Over at high volume (probably ~ 10W and well into AB2) and enjoyed all of it. Of course the individual recordings are not of the same standard, but I already knew that from my other amp. What SHOULD sound perfect DID!
Something else strange. The output, with the Edcor rewired for 3,8K a-to-a, kept at 18W/4R and 16W/8R. The CF now swings g1 to +12V at full output and the 'kink' is almost gone both in the CF-output and checked at the speaker-terminals. You have to be zoomed-in extensively on my scope to notice it. Frequency-response at full output flat from 18Hz and up to 22KHz whereafter it drops slowly to about 2/3 output at 45KHz. After that I don't care.
That is not half bad for a 6AD10 class-AB2 PP-amp.
Now I'm left with a problem: To build or not to build :D

rgds,

/tri-comp

Outstanding! A great deminstration of the difference between AB2 on a pentode load line vs triode load line.

The use of AB2 on a pentode is to get more plate current. The plate voltage can be swung low at any plate current with a pentode, but the peak plate current is limited by the transconductance. So at some point you need positive grid voltage to get more plate current.

With a triode,the problem is swinging the plate voltage low to get better efficiency. Positive grid voltage allows you to swing the plate voltage lower.

Your pentode amp probably had enough current drive capability and could swing the plate voltage as low as possible already. Driving into AB2 still gave you more power but mostly clipping.

Your triode amp was able to reach the same output power as the pentode because AB2 allowed the plate to swing to the same low voltage at signal peaks. Triodes aren't inherently limited to lower power, it's just the typical operating conditions that make it common.

Your amp sounds better now because the triode plate resistance is much lower and you have a better damping factor.

It would be very interesting to see your scope photos now to see if anything else changed.

Great results!
 
Hint: see what happens when plate voltage goes lower than G2 voltage.

eeh..? It never does since they're strapped or you're talking pentode-mode?.
If so why doesn't UL help with sound-quality when G2 swings with A, at least part of the way.
I did strap the Edcor for UL to test and I didn't notice much difference.
What am I missing?

I'll go with better damping. My 2C34-PP was first done w/o NFB and didn't sound especially good. After applying NFB the damping-factor/output-Z must have improved substantially and completely changed the sound towards Nirvana.
I know it's recommended to have the amplifier sound well before applying NFB, but I think that's easier said that done.

rgds,

/tri-comp

P.S.
Scope pic's are comming. Amp's hooked-up to my Stereo/TV set at the moment and I want to enjoy at least over the week-end to really notice how it performs with different sources.
 
OK, I admit I can't wait the whole week-end to put it back on the bench :)

First I did more critical & precise measuring of output power in triode-mode.
Results into a dummy-load:

Class AB1 / 8R / 1Khz = 3,5W (Edcor wired for 8R secondary)
Class AB1 / 4R / 1KHz = 3W (Edcor wired for 4R secondary)

That's not very outstanding and not worth the effort imo.

Enter AB2, Edcor wired as above:

Class AB2 / 8R / 1KHz = 12W
Class AB2 / 4R / 1KHz = 12,25W

Output of the CF swings g1 to +11V for full output and beyond that distortion is noticed at the dummy-load.

B+ sags to 300V/Full load and because it was available I added another 20V to that just to se what influence is has on output. It has ...some (i.e. around 0,8W increase only)!
The anodes swing down to +125V and then hits a brick-wall. With B+ at 300V anodes swings a fraction lower and then hits the same wall.
So, the 6AD10 must be already pushed close to max Ik; measuring 60mA, 8R/AB2/12W.
I guess the tubes are saturating and there will be no way of bashing them into giving more output.
Since the tube is already dissipating 18W at this point (B+ = 300V) there's really no sence in pursuing this problem!
Changing primary loads doesn't influence a lot. Running the 8R load on the 4R secondary winding and vice-versa just slightly changes output.

Conclusion:
AB2 works with scrap-tubes @$1,-/ea.; gives power and even sounds sweet!
What you put in front of the CF-drive is what you favour but the CF-drive is pretty universal imo. It's simple and should work with almost any odd OP-tube you'd like to try-out. When not using combo-tubes like the 6AD10 I've been using 5687 as the CF. Most likely an overkill but it works! You CAN get away with substituting the CCS with a resistor (@47K) if you're willing to scarifice frequency-response. I wouldn't recommend it, you'll notice the difference and the cost of 4xCCS' isn't anything worth mentioning.
YES, I know it complicates things that you need a separate negative supply between -65V and -100V to set-up an AB2 drive. In any case you need about half or 1/3 of that for the fixed-bias, so you're already in the right direction :)
Let your next amp be an AB2 amp ...and let us know what your findings are!


rgds,

/tri-comp
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.