[reVintage]

Simpler and even nicer figures!

[reVintage]

Check differential squarewave between the 6B4Gs grids! Looks descent and symmetrical. Does it matter how the individual grids(measured to ground) look?

EDIT: Dave, added what could be a really good way to use your concept IRL, only the autoformer is missing........

I am all fire about this, hope you don´t find my brainstorming to tiresome !

[reVintage]

One with current-mirrored 6B4Gs. The other one with (partly) good ole´cathode-bias.

[reVintage]

Hey Dave,
This is what I indicated in my mail earlier today. Load can be from 10k up to the CCS impedance. The opamp is not critical. The two 47ohms must of course be matched.

[dave slagle]

chances are this will become a new topic but for now...

OK here is some info on the IIT albeit a few days late. I did a design that has given me some interesting results so the past few nights I have been trying to track down the measurement error. I couldn't find it so I'm reasonably trusting of the plots below. What puzzles me is why the loading doesn't have anywhere near the effect on phase that the previous samples had.

In the measurements for the first design I had a substantial peak at 70K which was easily tamed by increasing the source Z or the load on the secondary. All of this fits expected theory.



when looking at the phase behavior of this device I noticed the "best" phase came from the worst response in the frequency domain which mimics experiences in spice. As you load things to fix the frequency behavior the phase gets worse.



Onto a completely different design. This one had really nice frequency behavior unloaded. The problem is loading just reduced gain and had little effect on any anomolies that secondary loading would quickly take out. Note how the -6dB trace is essentially flat to 100K (save the 63K hiccup) so loading does have an effect but nowhere near what happened in the previous version. (JJ can you say A2 driver)



not that anything ever comes for free, the phase difference between the various loadings shows nowhere near the difference expected.



my problem with this is that all of those plots are considerably worse than the ones from the "ringing" response of the first set of plots.

dave

[reVintage]

Hey Dave,

Great work! The first must be the way to go. But the second one wasn´t to impressing as I believe phase is utterly important!

What happens when you load less with, lets say 1k/100k or 1k/300k? Could only find 1k/25k.....

In my last schematic the load can be in form of a potentimeter(500k?). This way you could maybe "tune in" the optimum value !

How much phase difference is acceptable at 20k, 5 degrees max?

About the PP-cousins with cap output from driver and grid-splitter. Do they measure well, and do they sound good? If so, a 1:1 bifi (instead of Ra and cap) followed by the grid-splitter could be the way to go.

[dave slagle]

[dave slagle]

Secondary loading and phase of the original transformer, You can either get good phase or good amplitude with this version based on load. The other showed less dramatic response to loading. (do bear in mind that I am assuming the phase numbers audiotester gives are reliable)

Also note how the 10K load will lose a bit over 1dB of amplitude so the turns ratio would need to be adjusted to get proper amplitude match (if that is desired)

I also want to mention that I find loaded transformers to sound dull and lifeless and do not like designs that require external loading beyond what is required by the following stage. Obviously OT's need an 8 ohm load but for anything that sees a grid of a tube, I go for a design that operates "acceptably" with just that load.

I know you will probably ask for a RC network across the secondary to try to help things out and I'll look at that later.

dave

[dave slagle]

tried a few snubbers/

1500 +68p / 5000+500p||68p means

1500 ohm source with 68p in to ground (Rp of source || Grid of direct coupled tube)

5000+500p || 68P means 5000R resistor in series with 500P which is in parallel with 68P.

Still have some odd phase behavior...

It is interesting to note that the snubbed secondaries do give what appears to be full gain.

[reVintage]

After a quick read-through: Also noted a small cap on the primary to ground helped a lot. Will do some more sims with snubber secondary load!

[dave slagle]

There will be around 68pf to ground on the primary side from the miller capacitance of the non-inverted (direct coupled) 6B4 grid.

dave

[reVintage]

Dave,
Weird things happen when I sim. So I doubt the results from my sims in LTSpice have any value. Await your definitve results and conclusion.

[dave slagle]

I'm a bit torn, Lately I have been more concerned about phase response and less concerned about absolute amplitude accuracy.

What does conventional wisdom say about a 15dB peak at 70Khz? I'm leaning towards the idea that the peak is inaudible and if the net result is better phase at 20Khz then what is the problem? Of course if this were a splitter with a CT secondary and one side peaked at 70K and the other at 72K I wouldn't be surprised to see a 2K artifact and I think that is the problem. In this approach, we have one side peaking at 70K and the other flat to 1meg with Zero peaking. (ie no sum and difference artifacts)

The loading obviously causes phase issues so I think your idea of sand to provide a high Z for the offset current flow is probably the best choice.

If someone wants to build the circuit and listen to both (in mono) the bobbins are here.

I think you nailed it with a ST-70 to 6B4G conversion. Do a simple circuit board to keep things easy and we are ready to roll. If it works as we expect, it could actually be a product Do the version with the current mirror to assure matched currents and any of the currently available "CT choke/ autoformer" splittes could be used.

dave

[reVintage]

Na, will not take my old ST-70 as one OPT measures weird. Rewind?

So I will take the Anderson(Audion) PP 6B4G I have on the bench. Will add a 6080 tube regulator so there are no caps in the PSU output.

Also have a pair of QuadII that could be rebuilt this way. Have already removed the potting and rewired the OPTs to get rid of the cathode winding, so all primaries are in series.

Are these bobins for the cores we spoke of initially?

Have just another idea to show you where the phase inversion is in the output stage working as a LTP. For that one, only one single choke is needed.

[dave slagle]

They are for the cores that have a 22 X 60 mm top and that would be 80% nickel. With both bobbin wired up you will get stereo, can you listen in mono to each to compare?

dave

[reVintage]

Now I got it ! You have plotted whats coming out of the seconadry. NOT the resulting combined response?

[dave slagle]

yes,

I was going under the assumption that the non-inverted response will have a decade better performance so i think you can just divide all the numbers in half for the summed response.

dave

[reVintage]

So what about single driver 813 PP ?

[reVintage]

See next post!

[reVintage]

Hey Dave,

Took a while until I finally understood why your original idea was so forgiving about differences in inductance !

With an unbypassed cathoderesistor on the output tubes, the amp will work even if we change "upper coil" to zero H. It will then function as a self-split purely Class-A amp with 6dB lower gain!

So if we add a cathodecap we can go into AB but then inductance gets critical. Are we interested to invest in this cap ?

[reVintage]

Hey Dave,
Finally I think the current-balance issue is solved. By inserting small current-sensings resistors in the anodecircuit you can get rid of the crap below cathode. Have not had the opportunity to try it yet but it sims like it should work well. Better precision is achieved if the resistors are bigger. Any idea how big they can be relative to RDC?

[reVintage]

Lifting this thread as I was given a pair of PP transformers for making a proto in my Andersson chassi. Also hoping to get some help with the servo from a sand-guru.

[reVintage]

Hey,

With the help of sand(and tube)-guru Rod Coleman the servo problem is now solved. Simple and straightforward. And without the current-sensing in the cathode circuit we are free to use whatever biasing we want: Diodes, batteries, resistors etc.

These are Rods words albeit slightly adjusted to fit the current schematic.

"Anode resistor R1 senses the 6E5P current (I assume this is about 15mA). So you get about 3V.

Q1 emitter is now about 2,4V so the current source is about 1mA. This 1mA flows through Rtrim and gives about 22..23V, and this programmes M1 to draw the same 15mA in the other choke half.

BJTs will give better balance than FETs, but a M1 FET allow high value of R14, which will filter the audio signal out of the current source better!"

Must say this looks promising!

[reVintage]

Why not give the amp a chance to go A2?

[reVintage]

Finally I got my act together( I hope) and got rid of the separate feed for the driver.

Have tried to shoehorn the PP-chokes into my chassies today with no luck. Have another that maybe fits, will get it from my warehouse tomorrow.