Hi All,
I have a small 2XEL84 amp that I decided to experiment with a DC coupled output stage on… So far I like the results but I'm having trouble balancing the bias voltage at the cathode followers and I'm not sure why. I don't have schematic drawn up yet so I'll describe the sections below, maybe someone will spot an issue.
The phase splitter is a standard 12AX7 with 100K resistors and .022uF coupling caps, there is no feedback fro the output. These caps feed the grids of a set of cathode followers (12AU7) via 150k grid stoppers. A negative bias voltage of about -25vDC is delivered to the 150k resistors via 475k resistors. The cathode followers' cathodes are tied to a -40vDC rail via 47k resistors. The cathodes are tied directly to the power tube grids.
The strange thing is that the bias outage that shows up at the cathodes is not equal on both tubes… one tube is always about 1.2 voltages lower that the other. This results in a big disparity in plate current on the el84s. I've verified all the resistors and they are all matched within 1%, there is no evidence of leakage in any caps, and I've tried a few different AU7 tubes… the bias creeps up and down but it's always imbalanced by about the same amount. The power tubes are a fresh, matched set.
I've noticed that other cathode follower drivers tend to have separate bias pots - maybe this is why - but I figured that internal feedback in the follower would effectively force the cathode to follow the grid. I could try to hack a second bias pot but I've built myself into a corner and I'm not looking forward to the surgery.
Does it sound like I messed up by not using separate bias pots or is it possible I have some other issue here? Would it make sense to seek out a matched pair 12AU7?
Sorry for no diagram… I'll try to sketch something and post it.
Brian
I have a small 2XEL84 amp that I decided to experiment with a DC coupled output stage on… So far I like the results but I'm having trouble balancing the bias voltage at the cathode followers and I'm not sure why. I don't have schematic drawn up yet so I'll describe the sections below, maybe someone will spot an issue.
The phase splitter is a standard 12AX7 with 100K resistors and .022uF coupling caps, there is no feedback fro the output. These caps feed the grids of a set of cathode followers (12AU7) via 150k grid stoppers. A negative bias voltage of about -25vDC is delivered to the 150k resistors via 475k resistors. The cathode followers' cathodes are tied to a -40vDC rail via 47k resistors. The cathodes are tied directly to the power tube grids.
The strange thing is that the bias outage that shows up at the cathodes is not equal on both tubes… one tube is always about 1.2 voltages lower that the other. This results in a big disparity in plate current on the el84s. I've verified all the resistors and they are all matched within 1%, there is no evidence of leakage in any caps, and I've tried a few different AU7 tubes… the bias creeps up and down but it's always imbalanced by about the same amount. The power tubes are a fresh, matched set.
I've noticed that other cathode follower drivers tend to have separate bias pots - maybe this is why - but I figured that internal feedback in the follower would effectively force the cathode to follow the grid. I could try to hack a second bias pot but I've built myself into a corner and I'm not looking forward to the surgery.
Does it sound like I messed up by not using separate bias pots or is it possible I have some other issue here? Would it make sense to seek out a matched pair 12AU7?
Sorry for no diagram… I'll try to sketch something and post it.
Brian
Hey,
I'm not surprised I didn't get any replies…. wordy post, no pictures. Well I went back to this build with a match pair 12AU7 and things worked a it better. I'm getting a little disparity in plate current on my el84s. There's exactly -35 volts on both grids of the 12au7 followers, -17.4 and -17.0vDC on the el84 grids, but I'm getting a huge disparity in grid current: 12 and 20ma. I'm pretty sure these are a matched pair but I'm wondering if anyone with more experience thinks that the .4vDC different could account for the nearly 50% difference in plate current?
Any thoughts?
Brian
I'm not surprised I didn't get any replies…. wordy post, no pictures. Well I went back to this build with a match pair 12AU7 and things worked a it better. I'm getting a little disparity in plate current on my el84s. There's exactly -35 volts on both grids of the 12au7 followers, -17.4 and -17.0vDC on the el84 grids, but I'm getting a huge disparity in grid current: 12 and 20ma. I'm pretty sure these are a matched pair but I'm wondering if anyone with more experience thinks that the .4vDC different could account for the nearly 50% difference in plate current?
Any thoughts?
Brian
Brian,
You want individual bias adjustments for each side of the push pull.
I can see a few other things which are not optimal.
First the cathode followers with a 47K load to a -40V rail: The EL84 will require approx -10 to -12V on its grid (assuming no more than say 10 Ohms between EL84 cathode and 0V), check this out first, the -17V you read seems wrong.
That means a say 30V drop across the 47K resistor => 640uA. 640 microamps in a 12AU7 is just not enough, it is also not enough to get decent high frequncy performance (slewrate) from the amp particularly if you have higher Miller Capacitance at the EL84 grids from Ultralinear or Triode Mode operation.
Your negative rail neds to be enough voltage for the output tubes to be driven into cutoff. The "rule of thumb" is 3 times the normal bias voltage, so -40V should be OK (just).
Make sure you have grid stops on the EL84 (4K7 minimum). If they are oscillating at VHF frequencies then the voltage readings will be strange.
For your possible interest:
See post #603 here:
http://www.diyaudio.com/forums/tubes-valves/72536-el84-amp-baby-huey-61.html
Mosfet source follower in lieu of your 12AU7 cathode follower. Current source load in lieu of your 47K else much the same idea.
Cheers,
Ian
You want individual bias adjustments for each side of the push pull.
I can see a few other things which are not optimal.
First the cathode followers with a 47K load to a -40V rail: The EL84 will require approx -10 to -12V on its grid (assuming no more than say 10 Ohms between EL84 cathode and 0V), check this out first, the -17V you read seems wrong.
That means a say 30V drop across the 47K resistor => 640uA. 640 microamps in a 12AU7 is just not enough, it is also not enough to get decent high frequncy performance (slewrate) from the amp particularly if you have higher Miller Capacitance at the EL84 grids from Ultralinear or Triode Mode operation.
Your negative rail neds to be enough voltage for the output tubes to be driven into cutoff. The "rule of thumb" is 3 times the normal bias voltage, so -40V should be OK (just).
Make sure you have grid stops on the EL84 (4K7 minimum). If they are oscillating at VHF frequencies then the voltage readings will be strange.
For your possible interest:
See post #603 here:
http://www.diyaudio.com/forums/tubes-valves/72536-el84-amp-baby-huey-61.html
Mosfet source follower in lieu of your 12AU7 cathode follower. Current source load in lieu of your 47K else much the same idea.
Cheers,
Ian
So you're saying one of the cathode followers is resting with 1.2V more bias than the other? That does sound like quite a large discrepency for a 12AU7, since both presumably have identical anode and grid voltages...
I think he's talking about a 150k + 475k potential divider between stages.
Last edited:
Thanks Merlin, but I still don't understand what he's doing. Why run your audio signal through a 150K grid stopper? . . . . . and as part of a divider, why throw away your signal voltage?
150K x 1uA is .15V, why add that complication to the circuit?
The GE data sheet calls for a max recommended grid resistance value of 250K for fixed bias . . . . .
150K x 1uA is .15V, why add that complication to the circuit?
The GE data sheet calls for a max recommended grid resistance value of 250K for fixed bias . . . . .
Dunno!
A cathode follower has plenty of degeneration, so at least this isn't a problem.
Folks,
Sorry, let me play catch up a bit. Below is a quick (incomplete) schematic and a picture of the build. It's basically a matchless spitfire type amp with a direct coupled cathode follower for each el84 inserted after the LTP phase splitter. I pulled the cathode follower design from the only instrument amp I know of with a negative rail and followers… the SVT. That's where I got the 150k grid stoppers.
There are errors on this. For example, the signal enters and leaves the phase splitter triodes via the grids… rather than leave via the plate (whoops). Aside from that you can get the general idea. I really need to draw something else up that is more representative but the signal leaves the LTP at the plates, passes through the 150k grid stopper to the grid of the followers; and there is a 475k bias feed (shown as 150k per the SVT schematic but increased to not load down the LTP). There should be a ground symbol at the power tube cathodes (far right), and the unmarked resistors are there to set the bias range… I'll have to double check them.
I'm not doing a great job of helping the forum help me I think… that's why I wanted to refocus on smaller pieces of the problem. Does the mis-match in cathode bias voltage in the followers make sense? Should I need matched a matched triode 12au7? Could the large cathode resistors be working against me?
-Brian
Sorry, let me play catch up a bit. Below is a quick (incomplete) schematic and a picture of the build. It's basically a matchless spitfire type amp with a direct coupled cathode follower for each el84 inserted after the LTP phase splitter. I pulled the cathode follower design from the only instrument amp I know of with a negative rail and followers… the SVT. That's where I got the 150k grid stoppers.
There are errors on this. For example, the signal enters and leaves the phase splitter triodes via the grids… rather than leave via the plate (whoops). Aside from that you can get the general idea. I really need to draw something else up that is more representative but the signal leaves the LTP at the plates, passes through the 150k grid stopper to the grid of the followers; and there is a 475k bias feed (shown as 150k per the SVT schematic but increased to not load down the LTP). There should be a ground symbol at the power tube cathodes (far right), and the unmarked resistors are there to set the bias range… I'll have to double check them.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
I'm not doing a great job of helping the forum help me I think… that's why I wanted to refocus on smaller pieces of the problem. Does the mis-match in cathode bias voltage in the followers make sense? Should I need matched a matched triode 12au7? Could the large cathode resistors be working against me?
-Brian
I can see no reason for such a large mismatch between the 12AU7s, assuming you really have built what you think you have built. Are you sure you have enough cathode resistance for the power tubes? Maybe one of them is biased too hot and is drawing grid current? Stick a couple of small grid stoppers on them (100R say). Not only is this good practice, it allows you to check for grid current too.
Are both the V- points actually the same point?
Check your resistances. Maybe one of them isn't what you think it is.
Nice looking build, BTW.
What is that black twisted pair disappearing off the bottom of the photo? It looks like it should go to the grid stoppers of the cathode followers??
Are both the V- points actually the same point?
Check your resistances. Maybe one of them isn't what you think it is.
Nice looking build, BTW.
What is that black twisted pair disappearing off the bottom of the photo? It looks like it should go to the grid stoppers of the cathode followers??
Last edited:
you are close, but yet so far away.
the feedback you were referring to only will effect the AC parameters.
there is things on the dc that your very close.
I do like the parts you used.
the one thing you have to keep in mind is how linear the amplification is on the signal on top half and bottom half.
btw I do make an amp like this and no you can't have the schematic.
I gonna make you work for this one.
the feedback you were referring to only will effect the AC parameters.
there is things on the dc that your very close.
I do like the parts you used.
the one thing you have to keep in mind is how linear the amplification is on the signal on top half and bottom half.
btw I do make an amp like this and no you can't have the schematic.
I gonna make you work for this one.
Merlin,
Thanks for the compliment… the builds always start neat but things breakdown quickly during aggressive debugging sessions.
I swapped el84s to see if the imbalance would follow the mismatching bias voltage. Ip went from 17/23 mA to 19/18 mA…. which I could live with. So maybe these tubes aren't all the well matched anymore and maybe by swapping them I lucked out and found a workable point. With a B+ of 339vDC I upped the basis current to .030 mA. Sounds good… the clipping is a little rougher than I thought it would be.
I didn't have time today but next I'm going to drop the follower's cathode resistors to up their current.
Oh! And the twisted wires on the front lead to PPIMV and PPI treble cut controls dangling along the front. There is also a 6P1T rotary switch with a handful f different coupling caps for a bass control. The preamp before the PI is actually a standard 12AX7 gain stage DC coupled to a cathode follower with its cathode bootstrapped back to a split load on the preceding plate - straight form the the, uh, valve wizard website. Trying to get a nice high gain, simple stage without fussing with a small signal pentode.
So, lessons learned:
-Brian
Thanks for the compliment… the builds always start neat but things breakdown quickly during aggressive debugging sessions.
I swapped el84s to see if the imbalance would follow the mismatching bias voltage. Ip went from 17/23 mA to 19/18 mA…. which I could live with. So maybe these tubes aren't all the well matched anymore and maybe by swapping them I lucked out and found a workable point. With a B+ of 339vDC I upped the basis current to .030 mA. Sounds good… the clipping is a little rougher than I thought it would be.
I didn't have time today but next I'm going to drop the follower's cathode resistors to up their current.
Oh! And the twisted wires on the front lead to PPIMV and PPI treble cut controls dangling along the front. There is also a 6P1T rotary switch with a handful f different coupling caps for a bass control. The preamp before the PI is actually a standard 12AX7 gain stage DC coupled to a cathode follower with its cathode bootstrapped back to a split load on the preceding plate - straight form the the, uh, valve wizard website. Trying to get a nice high gain, simple stage without fussing with a small signal pentode.
So, lessons learned:
- DC coupled outputs aren't magic and are troublesome
- DC Bias going through followers ought have separate adjustments
- If you want to dodge bias shift from power tube grid current don't go through the hassle of a negative rail and DC coupling if you don't have separate winding for bias… the bias voltage still sags heavily with signal spikes…
-Brian
OK, I made some more changes and I like where things are going.
I reduced the cathode follower resistors to 10k, they're running at about 1.6mA each now.
I reduced there resistance in the bias supply in a couple of places to raise the voltage and reduce sag. I moved the V- rail supply node from the second filter cap to the first. I'm taking the adjustable bias from the second node. The negative rail is around -60vDC.
I upped the current through the el84s to around 30mA (About 10 watts idle dissipation). There is a slight imbalance between triodes but not enough to lose sleep over.
Brian
I reduced the cathode follower resistors to 10k, they're running at about 1.6mA each now.
I reduced there resistance in the bias supply in a couple of places to raise the voltage and reduce sag. I moved the V- rail supply node from the second filter cap to the first. I'm taking the adjustable bias from the second node. The negative rail is around -60vDC.
I upped the current through the el84s to around 30mA (About 10 watts idle dissipation). There is a slight imbalance between triodes but not enough to lose sleep over.
Brian
I have three direct coupled amplifiers push pull amplifiers. All require no adjustment and are universally stable whilst ever the valves stay within spec.
The secret is to fix the current in the finals stage and then let the other stages find a voltage where that occurs. This is achieved with CCS in the output cathodes.
Shoog
The secret is to fix the current in the finals stage and then let the other stages find a voltage where that occurs. This is achieved with CCS in the output cathodes.
Shoog
You can acheive better balance in the output stage by using separate cathode resistors rather than a shared one (or at least some individual cathode resistance for each output valve, plus some shared). This makes the operating points of the previous stages more relaxed, as Shoog mentioned.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.