parallel cathode followers, MB1 schem (now DC NFB too)

[greenmanhumming]

in the Pultec MB1 schem, what are the resistors linking the plate and grid of the 2 triodes in the parallel cathode follower for?

why not just link them directly with wire?

is it to make the CF not quite parallel? if so why?

basic question I guess, but I can't find any info on it...

 

[jdbakker]

General suggestion: not everyone will have the schematics of the MB1 memorized. For a faster/better answer, it helps to include a link.

HTH,

JDB.
[Don't take it personally, you're definitely not the first or the only one doing this]

 

[dfer]

Schemo is here - http://www.geocities.com/rafafreddy/DIY/pulteq-mb1/schem.jpg

 

[greenmanhumming]

thanks dfer, I was just about to post that link...

so, does anyone know why the CF is not just linked with wire? why the 1k between the grids and 470R between the plates?

I have made 2 channels of a preamp with a CF and done one each way, and can't hear a difference so far...

 

[PRR]

> why the CF is not just linked with wire?

Two tubes in parallel "can" take-off as a push-pull oscillator, resonated by stray capacitance and inductance, ringing in the 30MHz area.

High-current parallel tubes often have small isolation resistors to break-up the push-pull mode.

 

[greenmanhumming]

> why the CF is not just linked with wire?

Two tubes in parallel "can" take-off as a push-pull oscillator, resonated by stray capacitance and inductance, ringing in the 30MHz area.

High-current parallel tubes often have small isolation resistors to break-up the push-pull mode.

OK, i thought it must be something precautionary like that, great answer thanks PRR!

while we are on the MB1 - anyone care to explain how the DC coupled NFB works? whats that .25uF cap to ground and the whole network where the NFB meets the 1st stage cathode doing? i don't really understand the non grounded transformer etc

feel free to point me elsewhere if this has been discussed, i've seen a lot on building the MB1 but little discussing how it works

and.... if you put a resistor between the points marked to alter the gain, aren't you changing the 1st stage bias too? I've seen a schem someone on here has done for a vol control that is cap coupled at that point which makes more sense to me...

 

[emrr]

The network floating the input transformer above ground is, I think, a bootstrapping network to improve bass response, and the cap is a part of that.  This is common before 1950, hardly ever seen after that time.  I believe higher frequencies see ground directly through the cap, lower frequencies that won't pass the cap see both the grid and the cathode for additional gain.  Though, not much, considering the voltage divider network of 120/1500+ any additional gain resistor.  I think the DC coupled NFB also provides fixed bias on top of the cathode bias, which lessens any potential variation caused by the NFB shunt of the gain strap.  I'm sure there's more than that going on, and some of that may be incorrect. 

 

[Jonte Knif]

It might be that I didn't get every part of emrr's post, but I think things are not quite like that.

I looked at the schema and thought: there is something very familiar.
Two things:
I had reverse engineered another variant of this a couple of years ago. It had EL84 output tube and none of the complexity of the first stage arrangement. I liked the topo, DC feedback, handy, but soon realized that you can not adjust the gain much. The first stage bias resistor voltage is determined by the output stage current and any adjustments in the amount of feedback mess with the bias. If I remember the gain was about 40dB without the trannies, and could really not be adjusted many dB:s.

Then, enter the "ungrounded" (for audio that is grounded) input trannie. This is seen for example in V72 and V76. (But in a bit different purpose)  A very handy way to make feedback adjustment easy down to very low gains. Look, the 160k resistor and .25 uF cap have very slow time constant and don't matter at all for audio. They just pick up the bias voltage from the 120 Ohms resistor, voltage which is mainly determined by the ca. 10mA follower current. 1,2V. Quite familiar value for 12AX7.

What ever you put across 1500 Ohms resistor doesn't affect the bias, because the transformer potential is picked up from the 120 Ohms resistor through the subsonic RC-filter. Shunt the 1500 Ohms and you get higher gain.

How ever, this gain adjustment won't work well for anything else than occasional trimming, because the bias will for a moment get out of hands until the cap has reached the "right" voltage. Extremely elegant topology for "fixed but adjustable" gain block.

-Jonte

 

[greenmanhumming]

hmm, signal ground is effectively the top of the 1.5k resistor not the ground end then?

 

[Jonte Knif]

Yes,

To not cause any more confusion I would say that the grid signal is in reference to that point. Both at DC and AC (flat down to a couple of Hz). Just look where the potential seen by the transformer "ground" side is and calculate the -3 point of the RC network of 160k and 0.25uF to see that you are out of audio band.

-Jonte

 

[PRR]

It is not significantly bootstrapped. The transformer secondary is AC-grounded at the 0.25u cap.

DC bias may be adjusted without significant effect on the AC gain.

V1a bias is just the 120 ohm resistor. This gets to its grid through a 160K/0.25uFd filter, to remove signal and to hold the "cold" end of the transformer winding at AC ground. It is not directly affected by voltage across the 1500 ohm resistor. It is indirectly affected because V across 1500 subtracts from V1a plate-cathode voltage, but for any reasonable values it causes no great change in V1a performance.

The AC gain is roughly (17.5K/(120+1500))+1 or just over 11. The 1500 ohm may be adjusted over a wide range.

It can be set to zero to maximize AC gain. This will increase V1a current about 5%, and V2 current about 8%, neither is a problem. AC gain would be near 145. This is still a High-NFB amp with useful input overload. If you want "more", shunt the 120 ohm with a big electrolytic. Assuming 1:10 iron, input overload is 3mV.

The 1500 could be made larger, though for large increase you might want to reduce the 17.5K in addition or instead, to keep V2 current similar. If you changed 17.7K:1500 to 10K:10K, the AC gain would be 2, and V1a current would be down 25%. Trying to get closed-loop gain down to unity would require re-thinking... but at that point it would be whole-lot-simpler to make V1 and V2 both 12AU7 as quad cathode-follower.

The top of the 0.25u sits at about +14V as drawn, at zero with the 1500 shorted, and near +80V with a 10K:10K NFB loop. Compared with voltage breakdown limits on the 0.25u, the transformer, and V1a heater, this is safe.

The dead men knew what they were doing. Often including the idea that gain might have to be changed per customer needs, without re-design effort.

 

[emrr]

Excellent.  I was typing more, but can now withdraw my wild theories.  Some were on target, some were not, never mind.

If you cap couple the NFB, the cap is itself another frequency tailoring network, and must be sufficiently large to avoid changes in response with increased NFB.  At the time this was designed, no one would have trusted such a large electrolytic cap in that position; many don't trust it these days.  Plus, it would completely change the biasing scheme as described by PRR. 

To add a bit, the 160K/0.25 removes AC signal for the bias path, while isolating that AC signal which exists as NFB at the cathode.