fixed bias and paraphase inverters

[Andy Le Blanc]

many old fashioned designs use variations of paraphase phase inverter
where the inverted signal is derived via a resistor divider after one side of the
inverter/driver tube, and then re amplified on the other side to provide the
two equal but opposite phase signals needed for push pull.

The simplicity and the tone color of these are always attractive.

see fender 5a3 to 5d3, pretty common designs by many mfg.

I ran into an issue with el34 and el84 and the fender circuits, the control grids
(G1) develop enough current to alter the bias voltage on the inverting tube's grid,
which hinders the push-pull signals from the inverter.

This doesn't happen with 6v6 and 6l6 types.

The fix is to place a coupling cap and a ground return resistor on the grid(s) of the inverter, to block DC
components from affecting the inverter.

Another common feature associated with old paraphase is cathode biasing
of the power side, simple, reliable, but without the advantages and flexibility
of an adjustable fixed bias.

With the addition of an extra coupling cap, the same resistor network that
derives the inverting signal can be referenced to a -V bias instead of
ground, to get both a paraphase inverter and a fixed bias.

[Colossal]

Hey Andy,

Thanks for the discourse as always. I have to agree with the cool sound of the paraphrase. A neat, simple design which speaks to its elegance. I like the common (unbypassed) cathode resistor on the driver tube. Degenerative feedback compensates for mismatching in the triodes and other component variances.

[Andy Le Blanc]

I like local degeneration myself too, do watch out for el34 and el84 with paraphase
It was a strange surprise that a know good circuit would seriously crap the bed
with different tube type, fender 5d3 worked great with 6v6/6l6, as per the fender scheme
but don't assume it cover all types

[Colossal]

I like local degeneration myself too, do watch out for el34 and el84 with paraphase
It was a strange surprise that a know good circuit would seriously crap the bed
with different tube type, fender 5d3 worked great with 6v6/6l6, as per the fender scheme
but don't assume it cover all types

Andy,

That's very interesting about the EL34 and EL84 not sounding good with a paraphase. Were you running them in Class A or AB push-pull? The 5D3's paraphase is interesting; one cathode fully bypassed and the other not.

[Andy Le Blanc]

It wasn't a matter of class or load, the tube type has a just enough natural
control grid current to fubar the grid bias of a small signal triode if you omit
the extra cap with the inverter.

I found it raised in one reference covering inverters for HI-FI, after struggling
with the issue in a build, it caught me by surprise. The input sensitivity of the type maybe.

I'll do some digging in the AM, see if I can't find it again.

The bypass in 5d3 is set up like that for balance, I think.

[renshen1957]

It wasn't a matter of class or load, the tube type has a just enough natural
control grid current to fubar the grid bias of a small signal triode if you omit
the extra cap with the inverter.

I found it raised in one reference covering inverters for HI-FI, after struggling
with the issue in a build, it caught me by surprise. The input sensitivity of the type maybe.

I'll do some digging in the AM, see if I can't find it again.

The bypass in 5d3 is set up like that for balance, I think.

Hi Andy,

The late Gar Gillies has a self balancing Floating Paraphrase splitter used with cathode and fixed bias in his book examples 87 and 88 on pages 93 and 94, respectively. Might be worth a look, it works very well.

Have you thought of using the connection recommended by Mullard application notes (in Wireless World circa 1955) which Garnet and Traynor used with EL34/6CA7, tieing the raw bias to the Suppressor Grid/Beam Forming Plates (G3)? (Pin 1 would have raw bias, pin 8 would be grounded.)This would alter the internal resistance of the tube and reduce the current through the tube. The effect would linearize the tube. This might provide some protection during fault conditions, too. Both Garnet and Traynor were known for their loud clean amps.

In a Cathode Bias amp, what value are using for the EL34, EL 84 power tubes cathode resistor? These usually have to be tweeked upward, most amp designers were using resistor values at (or in Ampeg case, above) design Maximums. The 5D3 amp cathode resistor should be around 270 ohms the 250 ohm in the original at or beyond the limilts of the tube. A screen resistor of at least 1k would be a good idea, too.

Best Regards,

Steve
I have found both the 6CA7 and the KT77 tube behave better than the EL34. Might be worth a try.

[Andy Le Blanc]

I found part of it, the issue is related to the high transconductance of the tube type.

6v6/6l6 have a fairly low transconductance... 3.7 ma/v and 4.7 ma/v

where el34 and el84 have 11 ma/v and 11.3ma/v, if you push the tube it sets
up just enough grid current to tweak the bias of the inverting stage if theres
no cap to block DC, especially with floating paraphase inverters where the
resistance to ground is high enough and conducive.

I select a bias point with a 1r in the cathode circuit regardless of bias type
and measure the difference between the plate and screen grid to set the
screens voltage negative relative to the plate regardless of whether I go
with a series resistor or a voltage divider.

the transconductance and grid current issue in some circuits was a surprise

Thank you very much for bringing up references, I have seen either specifically.
Please post a link, or a quick scan I'd love to see them.

Mixed bias methods are very usable, I started looking at it as a way to avoid global feedback loop,
a way to utilize local feedback, the degeneration of a unbypassed
cathode resistor, especially with inverter types that aren't feed back friendly.

paraphase inverters have a phase instability at the hi and low ends of their
bandwidth which gives them their tone but also makes global feedback
difficult

[tubeswell]

The bypass in 5d3 is set up like that for balance, I think.

But its on the 1st inverter stage's cathode, which would boost the whole caboodle. So maybe its there just to get the overall gain up (to drive the output tube grids harder).

[renshen1957]

I found part of it, the issue is related to the high transconductance of the tube type.

6v6/6l6 have a fairly low transconductance... 3.7 ma/v and 4.7 ma/v

where el34 and el84 have 11 ma/v and 11.3ma/v, if you push the tube it sets
up just enough grid current to tweak the bias of the inverting stage if theres
no cap to block DC, especially with floating paraphase inverters where the
resistance to ground is high enough and conducive.

I select a bias point with a 1r in the cathode circuit regardless of bias type
and measure the difference between the plate and screen grid to set the
screens voltage negative relative to the plate regardless of whether I go
with a series resistor or a voltage divider.

the transconductance and grid current issue in some circuits was a surprise

Thank you very much for bringing up references, I have seen either specifically.
Please post a link, or a quick scan I'd love to see them.

Mixed bias methods are very usable, I started looking at it as a way to avoid global feedback loop,
a way to utilize local feedback, the degeneration of a unbypassed
cathode resistor, especially with inverter types that aren't feed back friendly.

paraphase inverters have a phase instability at the hi and low ends of their
bandwidth which gives them their tone but also makes global feedback
difficult

Hi,

I will have to wait until tomorrow for a Scan on the Garnet references. Until then, another thought on the EL34 and EL84, Ultralinear connection in the OT might provide some local feedback to linearize the tube.

Best Regards,

Steve

[renshen1957]

The bypass in 5d3 is set up like that for balance, I think.

But its on the 1st inverter stage's cathode, which would boost the whole caboodle. So maybe its there just to get the overall gain up (to drive the output tube grids harder).

Hi,

The cap would boost it higher on the one side, the other side of the inverter has a fraction of the gain; the phase inverter in the 5D3 is imbalanced, gives amp its the vintage tone and warmth.

Best Regards,

Steve

[tubeswell]

The bypass in 5d3 is set up like that for balance, I think.

But its on the 1st inverter stage's cathode, which would boost the whole caboodle. So maybe its there just to get the overall gain up (to drive the output tube grids harder).

Hi,

the other side of the inverter has a fraction of the gain; the phase inverter in the 5D3 is imbalanced, gives amp its the vintage tone and warmth.

Best Regards,

Steve

Yes, but I was not referring to the desirability of the inherent imbalance of the paraphase in these early amps. I was talking about the logic of attempting to balance the signal with the cathode bypass cap on the 1st stage.

Surely the cathode bypass cap is there to give an overall gain boost? If you want to try to balance the signal of a paraphrase, you need to tweak the input voltage divider for the 2nd PI stage, because boosting the gain in the 1st stage will merely result in the proportion of imbalance being the same as if not having the bypass cap there.

On the other hand the un-boosted (2nd) stage is subject to cathode current feedback (which helps reduce the inherent distortion of this type of PI). Was that what you were talking about Andy?

[Andy Le Blanc]

there's several circuit features going on in the old 5D3 schematic.

First off, there's no global feed back, and the cathode biased power tubes are fully bypassed,
and there's no other form of degeneration associated with the power tubes.

moving toward the inverter you next run into the resistor network that provides
the grid returns for the power tubes and the signal for the inverter.
This is a floating paraphase, there is a 220k and a 270k from the legs of the inverter
that are connected to a 56k that is referenced to ground, the junction is the source for the inverting side,
the different values are ment to reduce the imbalance, but it cant be perfect.

And then you have the separate cathode resistors, from what I've gathered
Leo was always attempting to be current with design trends.
Before arriving at the long tailed pair with all the feedback you can see the evolution of the product.

He wasn't after distortion, we like the dirt now, but not then, quite the opposite.

if you go about balancing these things, both AC at the inverter grid,
and DC at the plates, the usable bandwidth widens and you get a cleaner
tone, I can't help but think that is the intended purpose of the design.

but the design didn't last long, how many years did fender market this variation
before going to the next inverter, only a year or so? 1954-1955?

[renshen1957]

there's several circuit features going on in the old 5D3 schematic.

First off, there's no global feed back, and the cathode biased power tubes are fully bypassed,
and there's no other form of degeneration associated with the power tubes.

Hi Andy,

Part of the reason I suggested Ultra Linear connection with the EL's, it provides some local feedback on the power tubes.

there's

moving toward the inverter you next run into the resistor network that provides the grid returns for the power tubes and the signal for the inverter.
This is a floating paraphase, there is a 220k and a 270k from the legs of the inverter that are connected to a 56k that is referenced to ground, the junction is the source for the inverting side, the different values are meant to reduce the imbalance, but it cant be perfect.

Agreed, this is a floating paraphase, but I would have used different values, as in the upload and the cathode caps.

And then you have the separate cathode resistors, from what I've gathered
Leo was always attempting to be current with design trends.

I disagree on this one:

Separate cathode resistors weren't uncommon, just not found in cost cutting amps.

Leo Fender was a Radio repair, pure and simple. His circuits are Western Electric and RCA. Nothing new or patentable until the Tremelo circuit hit the scene.

Leo went from a common used in radio splitter (paraphase) to a Cathodyne splitter (concertina) in the next incarnation of the Deluxe. The Concertina is one of the oldest splitters to be used, one the first after transformer coupled splitters were jettisoned because of the cost.

The Schmidt Splitter/Long Tail Pair phase inverter had been around since the 1938, as an out growth of the differntial pair. Leo's engineers take on the self-biased version of this circuit that works better with positive-only power supplies and ground-referenced inputs.

The Bassman amp seemed to be the amp that was experimented with the most over the years. Quite a bit of different circuits were used, put into production, and then updated until the next experiment.

So Leo went from an old circuit, to an equally old (if not older) circuit, to a newer circuit. However, the earliest versons of the See Saw and Concertina were much simpler versions than the varieties we are discussing.

He wasn't after distortion, we like the dirt now, but not then, quite the opposite.

In than you have my complete agreement, Leo was never about distortion and for that matter neither was Ampeg (who targeted the Jazz niche).

Leo's market was Country and Western in the beginning and he never really left it, and Fender amps was about clean amps even into the CBS years of decline, even then company was apologetic about distortion. (Leo wasn't beneath borrowing ideas from Standel Amps of nearby El Monte, CA, either).

As the amps were designed to be clean, and to fill the venues that were played to with clean sound, it never occurred to build anything that would dimed, just sufficient headroom. It's in these lower Fi Hi Fi circuits (with cost cutting) that when pushed become interesting (distortion wise).

if you go about balancing these things, both AC at the inverter grid,
and DC at the plates, the usable bandwidth widens and you get a cleaner
tone, I can't help but think that is the intended purpose of the design.

but the design didn't last long, how many years did fender market this variation before going to the next inverter, only a year or so? 1954-1955?

Leo followed the logical pathway of clean (less distortion) that fixed bias and negative feedback provided in his larger high amps (Pro, Super, Bandmaster, Bassman, Twin, 5E series) and the new the 5E11 Vibrolux (Concertina with fixed bias), the Deluxe received the upgrade of only the Concertina Splitter (make sense to have the girls wire essentially the same PIs in all these models), the 5E9-A tremolux being one hold out, with the Paraphase (See-Saw) Splitter and tremolo.

There were circuits that used feedback around the PI, Ampeg had their share. However, according to the late Ken Fisher, Fender and Hull had too much pride to copy either company's design. (That would have to wait for Mesa, Peavey, and Behringer/Bugera)

Here is the uploads I mentioned, from Gar Gillies book (RIP). I am posting in context of editorial comment and fair use with no intent of violating.

Best Regards,

Steve

[Andy Le Blanc]

I appreciate the scans, thats just what I've done, tube and values are different

I spent some noodle time on it this morning...

the plate resistors and the resistors of the divider network are are the same 82k
and the cathode resistor(s) are 1k...

something interesting happens in this inverter when you go from one common
cathode resistor, to two separate resistors, to two resistors with the first stage bypassed.

with 1v 1k cps. ac signal applied to the input of the inverter, and a common cathode resistor,
the resultant ac signal on the inverting stage was .78 vac.

with separate unbypassed cathode resistors, the ac signal at the grid of
the inverting stage increased to .83 vac...

and with the first stage cathode resistor fully bypassed and the inverting stage
cathode resistor left unbypassed the vac present at the grid of the inverting stage was increased further to 1.17 v

Definitely a balance concern being addressed, with a common cathode resistor, you can hear the instability of the circuit.

The separate cathode resistors and bypassing is much cleaner, you can
hear the improved band width and stability.

[renshen1957]

[quote="Andy Le Blanc"]I appreciate the scans, thats just what I've done, tube and values are different

I spent some noodle time on it this morning...

the plate resistors and the resistors of the divider network are are the same 82k
and the cathode resistor(s) are 1k...

something interesting happens in this inverter when you go from one common
cathode resistor, to two separate resistors, to two resistors with the first stage bypassed.

with 1v 1k cps. ac signal applied to the input of the inverter, and a common cathode resistor,
the resultant ac signal on the inverting stage was .78 vac.

with separate unbypassed cathode resistors, the ac signal at the grid of
the inverting stage increased to .83 vac...

and with the first stage cathode resistor fully bypassed and the inverting stage cathode resistor left unbypassed the vac present at the grid of the inverting stage was increased further to 1.17 v

quote]

Hi Andy,

Thanks for the update, especially the values.

Will you please post a measurement with both cathode resistors bypassed, if and when you attempt such?

Also any success with the EL34s?

Best Regards,

Steve