The SS Project.....AB763 with PPICF

[SILVERGUN]

Thanks for the help term and sluckey,,,,I edited my previous post to reflect the correct wording (plate=cathode)

There was one other note that I wanted to review about that last experiment that I ran...

I had one DMM connected to the grid of one side of the CF, and the other one connected to the cathode of that same triode
At idle I was reading -22vdc on the grids and -19vdc on the cath.,,,,I had to get the signal pretty loud to get those voltages to move but once they did, they stayed evenly spaced at 2 volts apart all the way up to -55 grids ...-53 cath.

Those negative supply voltages didn't move easily, and it took a very "loud" signal to get them to down to -55 and -53

Could that move be a direct reflection of the point in the volume control where the clipping starts to become audible?
My mind wants to believe that we are seeing the voltage drop, as a result of increased current...........am I thinking correctly?

From what I've read,,, the high negative supply is placed on the grids to inhibit grid current (and I had questioned what value that needs to be).......is there a chance that I put so much signal there that it overtook the negative supply, and "introduced" grid current,,,which might have caused those volt. readings to drop (more negative)?

Maybe I'm way off, but I have to ask......

[SILVERGUN]

If you read his AC-couple CF page ( http://www.freewebs.com/valvewizard/accf.html ), in those exmaples, his input signal voltage is higher, and he chooses a higher bias:

"In this case a bias of about -3.4V is chosen, giving about 50Vpk before clipping, into easy loads" (from linked doc)

regarding, plate voltage , check the datasheet for the tube you are using for this rating.   If its 300, then the difference between the plate and cathode shouldn't exceed 300.   If you have 350 @ P,  and -47 @ K,,  then 397 exceeds that number.
The SVT schematic I referenced earlier:

the 5/69 rev.B sch I've got shows plate at 220v, grid at -77v, and cathode at -47v.

has a  bias of -30, and a plate voltage of +297.  ( the 220v supply comes from it's dedicated 6146B's screen power supply that provides 220v).

Thanks term,,,I had missed that ,,,,partially due to the fact that I was looking harder at the stage being DC coupled to the output stage in stead of being focused on it being AC coupled to the PI
So I'm just studying the cathode follower in general, and trying to grasp the entire concept.....
I think that I've been so distracted, and worried about trying to match the design to the original, that I've blocked out some very helpful info. (like that SVT scem.)

[jazbo8]

I'm still mulling over the apparent bias shift from your measurements, I still would like to see some scope shots - I don't really trust the DMM readings when you have large AC sitting on top the negative bias. Also, perhaps we are spending too much effort on the CF after all it is just a follower - it simply follows whatever appears on its grid, and if we are not talking about AB2 operation, then the PI is what we should be focusing on. All the effort that we have spent on the CF mainly centers around IMO safe operation, with the reduced B+ and lower cathode voltage, I think you got that sorted already.

[tubenit]

Did you see the TAG thread about the CF on a AB763?

As far the bias you will need -135 to 140 volts on the grids of the PI.

http://ampgarage.com/forum/viewtopic.php?t=21942

[Glennjeff]

SILVERGUN.

Don't think grid current is an issue as grid remains more negative than cathode at all times. It is in the nature of a cathode biased valve stage. Once the ability of the bias network to provide negative voltage is exceeded everything is being pulled negative, the signal is being clipped on the negative side and is no longer centred around zero volts DC, but has a negative DC component to it.

jazbo8,

I don't think it's too cool having an amp that goes radically out of bias if you turn it above 7.
What do you think needs adjusting with the PI. Or do you mean testing it for adequate operation.

tubenit,
From that thread you referenced I notice about 15 V AC indicated coming out of the PI, the SS CF can't handle that in this configuration IIRC.

Note also that the output tubes have protection diodes strapped cathode to ground, think this aspect of the design needs to be addressed also.

Also, the bias adjustment pot is configured differently.

All the best

[smackoj]

Wow Gunny! I love that sound. That amp has the BB King vibe and tone gyratin' all thru it. Nice work Mad Professor.

smacko jack     

[jazbo8]

I don't think it's too cool having an amp that goes radically out of bias if you turn it above 7.
What do you think needs adjusting with the PI. Or do you mean testing it for adequate operation.

Nope, not cool and probably isn't really the case... we'll have to wait for the scope shots...

Since it is just a follower, I believe nothing needs to be changed in Fender's classic design - after all that is part of the sound that I think SG is looking for...

[Glennjeff]

Excellent observation jasbo8. Doing the worst I can to do the best I can for SG. As I have said many times, "I am a newbie".

[SILVERGUN]

Did you see the TAG thread about the CF on a AB763?

Thanks T, no I hadn't seen that....
It kinda sounds like a pretty cool idea though   

I learned how to share information from you, openly and without hesitation, so it was only a matter of time.....

(he didn't have to make it sound like it was his idea though)   
Unless maybe it's one of our guys working as a secret agent   
Maybe I'll learn something from the thread....
The only reason we're still on this is because I don't have the technical knowhow to say:
this is what it is, this is why it is, and here's how you want to do it..........
It's why I'm here.....to learn and share
 

[SILVERGUN]

I'm still mulling over the apparent bias shift from your measurements, I still would like to see some scope shots - I don't really trust the DMM readings when you have large AC sitting on top the negative bias. Also, perhaps we are spending too much effort on the CF after all it is just a follower - it simply follows whatever appears on its grid, and if we are not talking about AB2 operation, then the PI is what we should be focusing on. All the effort that we have spent on the CF mainly centers around IMO safe operation, with the reduced B+ and lower cathode voltage, I think you got that sorted already.

Sounds good jaz,
I'll get the scope hooked up ASAP.....

Just for the record, for all involved parties, this amp sounds great, it just develops some distortion when you crank it up,,,,,sound familiar?......I was aiming for more pure clean all the way up on the dial, and we will get there........I've just slowed things down because I need everyone to dumb-down their explanations 

after all it is just a follower - it simply follows whatever appears on its grid
That's the way I need to view this.....instead of treating it like it's some super secret, highly engineered, Dumblefied stroke of genius that I could never possibly understand 

Thanks again for all of your help......I'll let you know when the scope's ready, and I can post another clip of the amp cranked so we're all on the same page

[SILVERGUN]

Did you see the TAG thread about the AB763

As far the bias you will need -135 to 140 volts on the grids of the PI.

They need sluckey over there  

[SILVERGUN]

OK, I got the scope hooked up and ran a short, narrated video clip shot.....looks like we've found the culprit.....this shows the clipping as it occurs at the PI...and it starts right around 7 on the vol. control
I looked at the CF and it was just a carbon copy of what we see coming off of the PI, and it didn't change (or clip any more) as it came off of the cathode of the CF....
What was I waiting for?  

http://youtu.be/nRiVU4Z9s5c

[jazbo8]

Was the scope's input AC or DC coupled? Was the center line at -19V when the video starts? And the y-axis was set to 50V per division? Sorry for the all questions, just want to make sure...

As far the bias you will need -135 to 140 volts on the grids of the PI.

I think that's typo, grids of the PI?!

[Glennjeff]

Excellent stuff SG.

There is a possibility that the CF could be loading the PI down. Would you be able to repeat that with CF's disconnected from PI coupling cap's and with 470k resistor's from output of coupling caps to ground? I get PI distortion with my D'Mars PI, which is constructed almost identical to yours, at high volumes but just to be thorough.

I ask because the clipping you are getting there is not symmetrical.

Also as jazbo8 mentioned, it would be helpful to know the settings on the scope.

All the best

[SILVERGUN]

Was the scope's input AC or DC coupled? Was the center line at -19V when the video starts? And the y-axis was set to 50V per division? Sorry for the all questions, just want to make sure...

AC coupled....not sure where the center line was set...I honestly just tried to dial in a clear wave in the center of the screen
What does -19v correspond to?....(I hope that isn't a super dumb question )  
I'm pretty sure the Y axis V/div was set to .5V (probe x10) when the video started and then I switched it to 2V to look at the signal coming out of the PI,,,,just to get it on the screen
I basically cut to the chase and just zero'd in on where I found the clipping.......so the first wave represents the signal as it goes into the grid of the PI.....and the second wave is where I saw it clipping heavily coming out of the PI (not the CF)

There is a possibility that the CF could be loading the PI down. Would you be able to repeat that with CF's disconnected from PI coupling cap's and with 470k resistor's from output of coupling caps to ground? I get PI distortion with my D'Mars PI, which is constructed almost identical to yours, at high volumes but just to be thorough.
I ask because the clipping you are getting there is not symmetrical.

Yes, I can perform the test you requested.....If the CF is loading down the PI, it's doing the exact opposite of what it's intended to do...
I found that my PI was horribly out of balance,,,and spent some time trying to get it into balance,,,,and managed to balance all of the gain out of it  
But I'm learning, and it's fun!!
It's probably better this way,,,,that someone didn't just come along and give me all of the answers.........

[jazbo8]

A few things to verify: -19V, because your DMM measurements showed the bias shifted from -19V to -52V, we would like to know if that is indeed happening. So you should set the scope input to DC couple and the vertical scale to 20V/div. First set the zero line where you want it (the top line on the scope is usually a good place), then if the bias is -19V, without any signal applied, you should see a horizontal line about 1 division down from the reference line. Now take measurements with the volume settings at 1, 3, 5, 7, etc. and make a note of where the center line of the sine-wave signal is, for each case - that is your bias voltage, in theory that center line should not shift, you may need to let it settle down a bit after you turn the volume pot.

[Glennjeff]

Here is what I'm getting out of my PI.

The first one is just under saturation.

The second one is driving it into very audible saturation.

The third one is totally hammering the PI.

Notice there is no hard clipping. It's managing 10V RMS (28V Pk to Pk) clean as a whistle and gets pretty grungy by 13 V RMS. These levels are in good agreement with those written on the circuit from TAG forum mentioned by Tubenit in reply #153

You're doing a great job SG, learning is a struggle, even for a spring chicken like yourself, borderline impossible at my age, but I'm not going down without a fight.

Now I'm going to build your CF section and convert my PA to fixed bias.

[jazbo8]

Here is what I'm getting out of my PI.

So these are the signals at the grid of the CF (AC coupled), right? Does the D'Mar have the same PI arrangement like the Fender, sorry too lazy to look it up ;-)

[Glennjeff]

jasbo8,

These are at the plates of the PI with no CF, AC coupled into cathode biased PA.

The PI is almost identical, one resistor is 820 ohms instead of 1 k. (I'll change that in the rebuild).

[Glennjeff]

OOOPS!

Forget that I had put a long plate 12AX7 in the PI as experiment.

Results are the same for lower drive but the 12AT7 can take a bigger beating.

[SILVERGUN]

A few things to verify: -19V, because your DMM measurements showed the bias shifted from -19V to -52V, we would like to know if that is indeed happening. So you should set the scope input to DC couple and the vertical scale to 20V/div. First set the zero line where you want it (the top line on the scope is usually a good place), then if the bias is -19V, without any signal applied, you should see a horizontal line about 1 division down from the reference line. Now take measurements with the volume settings at 1, 3, 5, 7, etc. and make a note of where the center line of the sine-wave signal is, for each case - that is your bias voltage, in theory that center line should not shift, you may need to let it settle down a bit after you turn the volume pot.

I will do exactly that tonight.....thanks for keeping me focused jaz
I kicked right into troubleshooting mode and just tried to nail down the clipping.......
I wound up replacing the NFB resistor with a 5K pot, and tried to dial in a "duplicate" waveform into the grid if the bottom half of the phase inverter,,,,and although it did clean up the signal, it took away all of the gain of the PI......so I'll get that set back to 820R (where we were), and go from there
*So now I'm studying the AC LTP 

These are at the plates of the PI with no CF, AC coupled into cathode biased PA.

Thanks Gj, that's a great reference point for me to shoot for....you can see how squared off my signal gets off the plates of the PI 

 

[Glennjeff]

SG,

With regards to jazbo8's test procedure. Could you do that for a couple of different values of bias supply voltage, eg -85V and -130V, and see how that effects the outcome, both of onset of clipping and effect on bias drift (if any).

[SILVERGUN]

Gj,,,Yeah...I kinda did that, but not to that extreme with the voltages....what I'm noticing is that the AX7 is more succeptable to the drift than the AT7, and it is more likely to drift when you start at a lower negative supply voltage.......which I hope explains my question ("why the high negative supply?")
At this point I have also raised my B+ so that I now have 380vdc on the plates of the 6L6s, because of the higher neg. supply, I needed to increase the plate voltage to get some current moving (at least until I figure out a better way to balance it all)
  V/div is set to 2V (with x10 probe), DC coupled

Here's the first version that jaz asked for, and I'm definitely watching it drop from -19vdc down to -53 very predictably
http://youtu.be/gYCPr7WM79Q

Then I put in a 12AT7 at -25vdc at idle, and it dropped to -32vdc, with the volume maxxed out:
http://youtu.be/Q9JH0-FnBSU

I left the AT7 in there and increased the negative supply to -50vdc, and it ONLY dropped to -53vdc:
http://youtu.be/sRdLm52xKJY

[SILVERGUN]

Gj, here's the result of the test you asked me to do where I disconnected the coupling caps from the PI and CF and connect thru 470K res. to ground.
Like before, I show the signal going into the PI and then coming off the plate clipped 
Looks like we've got some Pi work to do....
There is obviously no NFB at this point because the signal never made it to the OT

http://youtu.be/lisieu6ASB4

[jazbo8]

Ok, thanks for posting the new clips, it is pretty clear from the AT shots that the bias did NOT shift all that much (the center line of the sine-wave stayed pretty much on the line at idle), but for the AX, the bias appeared to be pulled down - the waveform became asymmetrical about the bias line, the question is why, my quick guess is the tube failed to deliver the current

I will be traveling for the next few days, when I get back, I will try and re-run the simulations with the output tubes, since my current sims only have the CF stage and that does not tell the whole story.

[Glennjeff]

SG,

Thanks for your patience. These results are not too bad.

In the original AB763, if our PI's  are pretty close to original manufacture, we had an amp that was designed to see 10V AC going to the grids of the PA for a clean sound, and about 13V AC for a saturated sound. If the CF's are replicating those voltage levels OK then we don't have too much to worry about in terms of distortion versus volume.

If that is 20 V per division (2 v/div X 10) then you are getting about 14 V RMS from AX7 and 20V RMS at moderate saturation which is fine.

Where I'm a little confused is that, in earlier listening tests, you said that the AT sounded more distorted and at lower volumes than the AX

The average DC level is drifting more negative at extreme high settings which actually shuts the 6L6's down a bit so it is not a "she's going to blow if we give here any more captain" situation, an unexpected protective measure in a way.

Also the saturated output from the CF's appears more symettrical than the output from the PI, who says CF's don't make a difference eh

Could you post the current circuit diagram of PI CF PA and (-)ve bias supply with no signal DC voltage readings. If you're still running 250V on CF plates and -85 V on (-)ve bias supply then the tubes should not have a problem with switch on voltage, 10% is nothing when we know those spec ratings are conservative.

All the best.

[Glennjeff]

I have no idea what I've done, but;

[SILVERGUN]

Where I'm a little confused is that, in earlier listening tests, you said that the AT sounded more distorted and at lower volumes than the AX

Could you post the current circuit diagram of PI CF PA and (-)ve bias supply with no signal DC voltage readings. If you're still running 250V on CF plates and -85 V on (-)ve bias supply then the tubes should not have a problem with switch on voltage, 10% is nothing when we know those spec ratings are conservative.

Yeah, I'm a little confused about that too,,,but that's OK....if it turns out that the AT7 is the better choice, then that's what it'll be.....looks that way now

Yes, I can post the adjusted voltage chart later,,,, I'm still using the higher negative supply of approx. -220vdc

Here's a great article on "tuning" the PI: (I'll be adjusting the breadboard to accomodate trimpots in critical locations to simplify PI tweaking)
http://www.aikenamps.com/LongTailPair.htm

[rzenc]

Very very nice work guys!!! Congrats to you all for pushing the boundaries.

SG, AFAICT, you seem to be getting LTP PI overdriven. I have a suggestion which might help??

I believe you are still using the AB763 preamp posted on first page. If so, I will suggest removing the cathode bypass capacitor on the last stage prior to PI. It will chop down signal and might help with LTP PI overdrive??

Whether you perform this test, let us know how it turns out.

Hope this helps

Best Regards

R.

[Glennjeff]

SG,

Don't know if there is any clear choice as to what valve may be best in this situation. I certainly have no ideas of what may be best for you.

Just configured my PA as fixed bias in ultralinear mode and connected my 12AT7 PPICF to it. Bias came up real easy and rock solid. 55.6 mA on one valve and 55.2 mA on the other. Got that sound like you hear on the Singers, like on your amp. Real Nice. Still gets overdrivey above 7 though. Haven't done any diagnostics yet.

Only problem, can't get negative feedback to work, it sounds hideous and distorted, tried swapping phase of OT, that was no go at all.
Ultralinear provides a degree of negative feedback anyway. (Negative feedback would not work before the mods either )

I've got my AT7 wired up as in reply #120 except I backed off a little on the positive supply - 170V. I'm running the tube a bit over half a watt, I would not recommend trying this with a 12AX7, 5751 or 12AY.

The tube will never experience greater than 320V and is generally operating at about 200V. That has eliminated the question mark about switch on stress to my satisfaction. The negative rail at this level is also within tolerance for Cathode / Heater max voltages also.

My poor amp has been such an experimental lab rat that I'm going to have to rebuild it from scratch I think, every component has been chopped and resoldered and I never planned on the extra circuitry needed to create the specialised supply rails for the CF's.

All the best.

[DummyLoad]

http://www.ampbooks.com/home/amplifier-calculators/long-tailed-pair/

i use tube-cad - it's fast, easy to use and reasonably accurate, + plus you can add other less popular tubes.

--pete

[SILVERGUN]

Just configured my PA as fixed bias in ultralinear mode and connected my 12AT7 PPICF to it. Bias came up real easy and rock solid. 55.6 mA on one valve and 55.2 mA on the other. Got that sound like you hear on the Singers, like on your amp. Real Nice. Still gets overdrivey above 7 though. Haven't done any diagnostics yet.
Only problem, can't get negative feedback to work, it sounds hideous and distorted, tried swapping phase of OT, that was no go at all.
Ultralinear provides a degree of negative feedback anyway. (Negative feedback would not work before the mods either )

I spent at least 3 hours last night just trying to balance the PI to a point where I got decent gain, but without the clipping,,,and that is quite a challenge.
To get that great tone at 3 on the vol. you have to settle for a little clipping above 7........I was able to dial it back a little, but the PI is very critical in the overall tone of the AB763 with or without the CF
If you try the calculator that DL supplied above (thanks!) and enter common values, you'll notice a pretty big difference in gain between the inverted and non-inverted signal......and it doesn't take into account NFB
I tried a 12AX7 and 12AU7 in the PI and attempted to balance those as well but got too much gain with the AX7 and not enough with the AU7...

The biggest thing I learned was that if you try to balance both phases perfectly, you have to give up a lot of gain,,,,,and NFB had a huge impact on gain thru the PI
I had a lot of fun and did a lot of wave watching last night, while taking every value to the extreme......
Tonight I'll go back and try it the way I left it, and then try it with the SSS values in the PI, which I haven't done yet

SG, AFAICT, you seem to be getting LTP PI overdriven. I have a suggestion which might help??
I believe you are still using the AB763 preamp posted on first page. If so, I will suggest removing the cathode bypass capacitor on the last stage prior to PI. It will chop down signal and might help with LTP PI overdrive?

Tried it and just lost too much of the front end feel (and vol.) of the amp.....lifeless, would be the way to describe it,,,,so while I had that cap off, I decided to try to make up for that lost gain in the PI and see how that sounded, and it just wasn't real good....not sure how else to say it...we tried  

[SILVERGUN]

Sometimes I feel like I'm just rambling off into cyberspace,,,and then I see how many times this thread has been read and it makes me feel like this work has a bigger scope than what I am actually capable of realizing,,, because I'm right in the middle of it.....
I do realize that I should know this stuff, and I'm sure that there are guys who could've figured this out in way less time,,,,,but I have grown to truly appreciate the learning process.....

After many hours in front of the scope, I have a new found respect for the phase inverter,,,,as it is the heart of this design, and has much more to do with the sound and drive of an amp than I ever truly recognized before.......probably because the values always look fairly close, and the architecture of the LTP stays constant throughout,,,, it lull's you to sleep with it's simplicity
In reading some more of the available info on Dumble, it now stands out to me that the PI was a large focus of his.....

So,,,in my previous post I had mentioned that I had tried the 12AU7 in the PI, and was unimpressed with the lack of gain,,,,BUT, when I went back at it last night, I used the sss values in combination with th AU7 and it was a pure winner......so I'm questioning what I thought I knew about the tube compliment for the sss.....( someone else had mentioned 12AU7-PI and 12AT7-CF ),,,,and that's where we are now

I was unable to get the AT7 to not clip while keeping an acceptable amount of gain,,,but with the AU7 the clipping was immmediately gone and I was able to increase the gain and keep a fair amount of balance by manipulating the bias and tail resistors in conjunction with NFB........an "ah-ha" experience for sure....
Here's the updated values and voltage chart for now: (with changes highlighted in red)

[Glennjeff]

Spent the morning doing some "test measuring" on the beastie.

Now I know this is not meant to be hi-fi but,

Conventional PI's seem most inadequate to drive a pair of 6L6's for an ULTRA CLEAN amp design, managing only about 12 volts RMS before getting audibly grungy, and noticeably squared up on the signal analyser.

PPICF's are much more suited to the task managing to generate about 48 V RMS before getting grungy.

Whilst I was doing this "measuring the beastie" thing, I measured the input impedance of the 6L6's grid. Whilst on the fuzzy low end of my measuring equipment it came in at a surprisingly low 33 k ohm @ 220 Hz. This may account for the PI's poor performance when used without CF's. High impedance source being loaded down.

At these levels I measured clean, as in not noticeably grungy, as 20 Watts RMS with PPICF making the power for standard PI's about 5 Watts RMS before grunge.

Those who thought that putting CF's after the PI amounted to a "simple buffer stage" that was "a waste of parts" and "not at all an interesting problem to solve" were plain wrong. Sorry to be brutal about it.

As they said on the TAG forum "Like the difference between night and day".
When I do crank the amp up real loud, the overdriven sound is also greatly improved. (Although my ears are still ringing 3 hours later)

Now need to think and research how to get the PPICF's generating about 100 V RMS, or even if that is necessary.

We do like that distortion after all.
I must conclude that most of that distortion, in the past, has possibly been from PI saturation rather than output tube saturation.

Kudos for you're pioneering and imaginative spirit SILVERGUN.

All the best.

[SILVERGUN]

I must conclude that most of that distortion, in the past, has possibly been from PI saturation rather than output tube saturation

I agree,,, and that really surprises me....not what I expected at all
Everyone always talks about pre-amp distortion or power tube breakup but you never hear anyone say "that is some great PI clipping"
I'll never take the PI for granted again, and I'm really thankful that we pushed ourselves right into this conclusion....it definitely changes the way I see the whole picture now 

I think I can safely say that the PPICF mystery is solved, and I feel comfortable moving on to tweaking the pre-amp, and seeing where I can go from here.....as it is , I think this amp is a funk players dream, because of it's bouncy response, and mix-cutting punchiness...
It's got me thinking of doing some James Brown covers   

[jazbo8]

Finally had a chance to run some sims on the CF/PA, unfortunately, I was not able to duplicate the bias shift as shown in SG's video, still wonder if that's due to the scope settling time... Anyway, here is the output waveform - stepping the input from 10Vpk-60Vpk, clearly the grid of the PA clamps the input waveform at ~0V when the input voltage exceeded 40V, also there was very little grid current due to the 10k grid stopper used in the simulation, so what was causing the bias shift in the video?



While we are on the subject of the PI, here are some useful info:

An essay by HPB well worth a read.
Basic LTP design tutorial by Aiken.
Balancing the LTP by Ampbooks.

Happy reading...  

[SILVERGUN]

so what was causing the bias shift in the video?

It still does it,,,only to a much lesser degree now with the AT7 in the CF
With -24vdc on the cathodes of the CF at idle, it wil drop to -27 with the volume control on 10........and it doesn't start to pull down right away,,,it waits until we hit about 6 on the vol. control to start dropping
 
It's gotta be an increase in current causing the voltage to drop.....how can I measure grid current on the 6L6s?.....and could grid current there be pulling the bias voltage down?.....I'll go read more

Let me know if you want to see anymore scope shots 
Welcome back....I found your blog while you were away

[jazbo8]

The bias shift is not suppose to happen, since it is the primary reason for using a cathode follower to drive the output tubes in the first place. Here is a quote from R.G. Keen's MOSFET Follies:

"When output tubes are overdriven, their grids are often driven positive. If the coupling to the grids is by capacitor, this can lead to grid blocking/ bias shift as the grid current skyrockets when the grids go positive. This current builds up as a negative charge on the coupling capacitors and causes the tube to effectively be biased more negative (towards class B) than the static bias setting. So on overdriven notes, the output tubes are pushed into a harsh-sounding crossover distortion. In addition, the usual 12AX7 or 12AT7 phase inverter plate impedance is so high that it can't pull the output tube grid positive in the face of grid current, so the grid of the output tube is hard-clipped at Vgk=0V. This clipping can be as harsh as any solid state situation.

The solution to the class B bias shift and grid current clipping is to drive the output tube grids direct coupled from a cathode follower. The follower is set up so that its cathode sits at the normal negative bias voltage of the output tube, maybe -10V to -40V depending on the tube type. This means that the cathode resistor is tied to a more negative voltage of something more than twice the bias voltage, so it can pull the output tube grid negative..."

Here is the whole article.

It just drives me batty not knowing what's causing the shift... Could you please do me a favor and change the grid load resistors of the 6L6s to 100k and see if the bias shift still happens?

[SILVERGUN]

Great article there...thanks AGAIN

It just drives me batty not knowing what's causing the shift... What size is the grid stopper now? Does changing it alter the amount of bias shift?

How much shift could we expect to see in a standard AC coupled output stage? 

The grid stoppers are still the 1.5K and I havn't tried any other value.....that will be a simple exercise tonight
I've got the scope hooked up for a dual trace now,,,,,, if there's anything you want to see, just let me know

The amp definitely stays cleaner longer now since the PI tweaks,,,,,but now after 8 on the vol. I can really hear the 6L6s getting ugly

***Amp sounds great set up next to a Deluxe Reverb re-issue, and you can really tell the difference that the CF makes in bass response 

[jazbo8]

The grid stoppers are still the 1.5K and I havn't tried any other value.....that will be a simple exercise tonight
I've got the scope hooked up for a dual trace now,,,,,, if there's anything you want to see, just let me know

The amp definitely stays cleaner longer now since the PI tweaks,,,,,but now after 8 on the vol. I can really hear the 6L6s getting ugly

***Amp sounds great set up next to a Deluxe Reverb re-issue, and you can really tell the difference that the CF makes in bass response  

Great to learn that the amp is "bettering" the DRRI, at least things are going the right way I think the 1.5k stopper is ok, but the the grid load resistor for fixed bias 6L6 should be <100k per the datasheet. Even though the simulations showed no grid current even with large input voltage, on the bench, there is got to be some grid current flowing when the power tubes are over-driven, so the larger the grid load resistor, the more the grid will get pulled down. Also, you can verify if the bias shift increases the crossover distortion, when both tubes are driven to cutoff - this is a good chance to use the dual trace on the scope.

[SILVERGUN]

I think the 1.5k stopper is ok, but the the grid load resistor for fixed bias 6L6 should be <100k per the datasheet. Even though the simulations showed no grid current even with large input voltage, on the bench, there is got to be some grid current flowing when the power tubes are over-driven, so the larger the grid load resistor, the more the grid will get pulled down. Also, you can verify if the bias shift increases the crossover distortion, when both tubes are driven to cutoff - this is a good chance to use the dual trace on the scope.

Here's what I saw...when I increased the grid-stoppers to 100K, the bias drift appeared worse as a result...
So I added in the 470K "grid-leaks" (pictured)...and they introduced some hum....I put them in as 500K pots and tried adjusting different values to "clean up" the signal getting to the grids of the 6L6s (and I could use them to help balance),,,,,and what I noticed was that one side was acting differently than the other....so I switched the position of the 6L6s, and the odd variation stayed with the (left side) socket.....what I saw was that the bias voltage and current were the same at idle, but at max signal, one side was drifting a lot further than the other, and reading much more cathode current......and when I swapped tubes, the variation stayed with the socket, not the tube
I kinda think that the signal coming into the grids was "hot" and imbalanced, and was causing more grid current on one side than the other and that was causing more bias shift on the one side, and in turn, pulling down the whole bias circuit......I should have mentioned this before,,,but the bias shift always appears a little worse on one side......I was just discrediting that as an imbalance in the triodes of the CF and PI....but know that I think more about it,,,it could be the result of more grid current through one 6L6 than the other....
So then I got determined to figure this out and took the output tubes out, and checked and removed and reconnected all connections on the breadboard and sockets, just to be sure....
While I had everthing all apart, I measured the OT and got equal readings on both sides of the center tap...
One other very important note....obviously,,,, grid current is directly linked to the input signal voltage......so I decreased the signal coming from my sig, gen to the input and was able to eliminate the grid current and resulting  bias shift that way, by decreasing the signal that was reaching the grids...
---How to you determine what signal voltage to feed the input?
---When I tried to reduce the signal at the PI (in my earlier attempt to reduce clipping),,,I lost a lot of gain there
    Is there a "normal" amount of gain that should be acheived coming out of the PI?......voltage wise,,,,,or is it normal to just push it to where it's causing grid current, and then just back it off slightly?
---What is the max. signal voltage that should make it to the grids?
---How do I measure grid current?.......when using ohms law, how do I know where to insert the values

i.e.-----if I measured a 5mV drop across those 100K grid-stoppers at -24vdc bias, at max. signal,,,what does that equal?
I'm assuming it's the same method as figuring cathode current,,,but the extra 5 zeros throws a wrench in it for me


Of course,,,, I got rushed away from it and had to go help my kid move a refridgerator   ,,,,,so I'll try to confirm some things tonight
I can look for crossover distortion on the scope, and try to get a better grasp on the big picture

Sorry this is like tube amps 101 jaz,,,,thanks again for hangin in there with me  

[Willabe]

Of course,,,, I got rushed away from it and had to go help my kid move a refridgerator   ,,,,,

Gotta love those father/son bonding times.


            Brad         

[jazbo8]

First some clarifications, when I said to change the grid load resistor to 100k, I meant the 470k shown in the SSS schematic and not the grid stop resistor which is 1.5k, you can leave it as is.

---How to you determine what signal voltage to feed the input?
If you are refering to the power tube, then then its input should equal to the bias voltage, so 24Vpeak, this should give you a relatively un-distorted sinewave at the speaker terminal (assuming none of the other stages are over-driven). Again, since I did not simulate the rest of the amplifier, I could not tell you what level the signal generator should be set at the input - but you can use the scope to set the level at the power tubes´grid - they should be nearly 24Vpeak sinewave.

---When I tried to reduce the signal at the PI (in my earlier attempt to reduce clipping),,,I lost a lot of gain there, Is there a "normal" amount of gain that should be acheived coming out of the PI?......voltage wise,,,,,or is it normal to just push it to where it's causing grid current, and then just back it off slightly?
Loss of gain? Do you mean when you try to balance the PI, the overall gain got reduced? If so, that is a separate issue. For now, you can watch the scope at the power tubes grids while increasing the signal generator´s output until the waveforms clip (near the 0V line) and back off a bit. I would not worry too much about the PI´s balance for now, since we are trying to identify the source of the bias shift. If the waveform at either one of the grids isn´t clipped, then there should be little or no grid current flowing, so the bias should stay put as you vary the signal level, from zero upto and perhaps even a bit beyond 24Vpeak.

---How do I (can I) measure grid current?
This is tough one, I would like to know as well...

I hope we can get to the bottom of this puzzle soon and thank you for putting up with my pestering  

[SILVERGUN]

---How do I measure grid current?
This is tough one, I would like to know as well...

I went out there at lunch time and tried to do it just like you would try to get a cathode current measurement...
I put a 1R resistors in palce of the 1.5K grid stoppers, and read the V drop accross each one....

At first I wasn't getting any reading, but I was able to get up to 7mV (=7mA) by forcing grid current with a hotter input signal

I might have been forcing grid current all along by not regulating the input signal, and just guesstimating that the volume I heard matched the vol. of my guitar signal (or at least close enough for testing)

THAT'S IT....I just went back out and confirmed 100% that when I force grid current,,,that forces bias drift,,,and one side is definitely more affected then the other,,,,and it doesn't "follow the tube" when I swap them,,,BUT, it does drag the whole bias circuit down with it

[Glennjeff]

You could have left the 1500 R grid stoppers in there and measured the voltage across them and used ohms law to figure out the current. Ohms Law: Voltage = Resistance * Current.

Have you measured the actual RMS output power of the amp at the point where bias starts going off. With 300 V DC on the 6L6's you can only get about 20 Watts RMS, so if the max output power gets there at or before bias drift point then the bias is not an issue.

If you run 380 V on the output tubes, you will have to run greater negative bias voltage and thereby change the onset of grid current to a higher volume.

If you want to fix the bias problem with the PI/CF, increase the voltage on the anodes of the output tubes.

[SILVERGUN]

You could have left the 1500 R grid stoppers in there and measured the voltage across them.
V-IR

I just figured the math would be easier with the 1Rs in there......no math

[Glennjeff]

Fair enough. A bit of math comes in handy sometimes and nearly all "do it with a calculator" level stuff.

Will measure my amp a bit more in the next couple of days to see if my thoughts were valid.

From what we've discovered so far the PI / CF can put out about 50 volts peak so it is not surprising that it is exceeding the6L6's -24 V bias on positive swings and inducing grid current.

[SILVERGUN]

Have you measured the actual RMS output power of the amp at the point where bias starts going off.

no

If you run 380 V on the output tubes, you will have to run greater negative bias voltage and thereby change the onset of grid current to a higher volume.
If you want to fix the bias problem with the PI/CF, increase the voltage on the anodes of the output tubes.

I went out and tried that quickly,,, and it appears to be correct.....it seems much more difficult to induce grid current with the higher plate/bias voltage

I'll try to confirm 100% and report back later

[SILVERGUN]

If you run 380 V on the output tubes, you will have to run greater negative bias voltage and thereby change the onset of grid current to a higher volume.
If you want to fix the bias problem with the PI/CF, increase the voltage on the anodes of the output tubes.

I went out and tried that quickly,,, and it appears to be correct.....it seems much more difficult to induce grid current with the higher plate/bias voltage

Nope....I must have rushed the results earlier, and turned down the amplitude on the sig. gen....
It is just as likely to conduct grid current if I have 300, 380, or as high as 420 on the plates  (maybe not in theory,,,but in reality ,yes)
If I overload the grid with a hot signal, it doesn't help with grid current if I have a higher - bias voltage.....it still drifts

I reverted back to getting the input signal correct first and used the scope while I played, and tried to match the amplitude of that sine to the one that I was feeding the amp from the sig. gen.....ok check,,, close enough for rock and roll
I noticed that my signal looked great coming through the PI and CF.......but as I cranked the vol. control on the amp,,,,still grid current, and bias drift

So I used those two 500K pots that I had put in place of the 470K grid leaks to dump some signal just before the grid stoppers,,,, I kept my eye on the bias voltage and monitored the drift from -25v to -30vdc,,,,with the vol. cranked and the bias reading @ -30vdc, I dialed the grid-leak pot down to get the bias voltage back down to -25, and wound up at 180K on the pot....I did the same thing on the other side and wound up with the same 180K over there...

So that got me NO bias shift,,,but resulted in MORE grid current

EDITED: Couldn't go to sleep until I figured out what happened....
I made a mistake...I wasn't compensating for putting the 1.5K grid stoppers back in,,,,so when I thought I was reading MORE grid current, I was actually reading a larger voltage drop across the 1.5K (compared to the 1R I was using for my calculations earlier today)
So now that I have done the math, it appears I'm down to 2.6mA grid current,,,at full signal voltage
  
4 volt drop across 1.5K resistor = 2.6mA grid current....right?

[jazbo8]

I went out there at lunch time and tried to do it just like you would try to get a cathode current measurement...
I put a 1R resistors in palce of the 1.5K grid stoppers, and read the V drop accross each one....

Duh, that wasn't as hard as I thought... so you have verified that grid current was the culprit, so we can check that off. Now if you recall HPB's point on the CF's function to handle the onset of grid current as oppose to AB2 operation, then you need to control the amount of drive at the power tubes' grids, with little or no grid current when the volume is fully dimed - you can do that by adjusting the overall gain in the preamp stages. Love to hear some more clips when you get all that under control.