Split Plate-Load Resistors............

[Jack_Hester]

Good read on this topic.  This was inspired from a schematic posted by JoJoKeo, over in the Parallel FX Loop thread.  And, an expressed interest by Tone Junkie.

Jack

[John]

Jack, thanks!

[jojokeo]

That's one of the papers that helped me too Jack 

[terminalgs]

I'm not exactly sure where the author is coming from, or going with that, so help me out.

he mentions the split-load design as having been around a long time, but (correct me if I'm wrong) they were always used as phase-splitters,  not "gain reducers". While Fender by-and-large moved to the LTPI around 1960, Ampeg continued to use them, and I believe, still does in the SVT.

So on to the meat of the author's point: using split-load for fine tuning gain.  Why not simply use triodes with different characteristics (namely Mu), or voltage dividers?   

"replacing the typical 100k plate resistor with a single 51k plate resistor; however, that would have a drastic effect on the bias of the preamp stage.".  No matter what the Ra and Rk happen to be, to consider bias, one must know what is going on at the grid.   He doesn't even mention the grid in any terms at all !!!

Certainly all kinds of characteristics of the triode will change if you go from Ra=100K/Rk=1500 toward Ra=Rk=51K while holding all other component values constant.   Output impedance, Frequency response going into RC network (he mentions, but doesn't go into),  current, bias,  everything...  But I'm confused why you'd use split-load to tweak whatever you are after (he says gain) instead of tube type, voltage dividers,  coupling cap selection load selection, etc.. 

I don't mean to beat this guy up, but I don't think he's providing the full story, or even a synopsis of the full story...  help me understand

[jojokeo]

I've been playing around with this technique for a while and like the results. It gives us another option and tool in our tool box for designing and/or tweaking. I've mentioned this several other times in the past and cannot remember which threads but read here a bit: http://www.el34world.com/Forum/index.php?topic=16606.msg164923#msg164923

[HotBluePlates]

I'm not exactly sure where the author is coming from, or going with that, so help me out.

he mentions the split-load design as having been around a long time, but (correct me if I'm wrong) they were always used as phase-splitters,  not "gain reducers".

You're getting tripped-up by the names.

A "split-load phase inverter" is also known as a cathodyne or concertina phase inverter, or as a "phase splitter" in RDH4. This circuit divides the load resistance into 2 parts, one in the plate circuit and one in the cathode circuit, and generates 2 equal and opposite-polarity outputs.

Steve goofed when he used the term "split-load plate resistors" as he should have said "split plate-load resistors." This circuit spplits the plate load resistance into 2 parts, but both are in the plate circuit. This circuit also has but a single output.

The bottom line is that a designed stage has too much output signal level (which could be a result of too much gain, or because of something else), which then needs to be reduced. There are several ways you could get the needed loss of signal:

• Use a pot (volume control)
• Omit the cathode bypass cap
• Use a feedback loop
• Make the plate load resistor very small
• Use a lower mu tube
• Use 2 resistors as a voltage divider after the stage

There are tradeoffs and/or problems with all of these. Let's assume our original stage is perfect in every respect except for the too-high output. The problems with the other approaches then are:

#1. A pot is expensive compared to resistors. It needs chassis space. If you don't plan on the gain loss being variable, you waste space and money.

#2. Omitting a cathode bypass cap will cut gain of a stage; however, the gain loss will depend on the transconductance of the tube at the operating point and on the value of the cathode resistor. If your needed signal loss doesn't match the gain cut due to the omitted bypass cap, you're out of luck. The local negative feedback will likely change input and output impedances, maybe in a way that doesn't suit the rest of your circuit. You may need a cathode bypass cap for tone shaping.

#3 Requires (probably 2) external resistors.  The local negative feedback will likely change input and output impedances, maybe in a way that doesn't suit the rest of your circuit.

#4. If you make the plate load resistor very small, voltage gain of the stage will drop. However, distortion may go up, and the stage may not be able to accept as-big an input signal. Plate current will go up, meaning you may have to redesign power supply components because of additional voltage drop or because (in a mass-production environment) you paid for a power transformer able to deliver exactly what the original circuit demanded and not a volt or milliamp more.

#5. You could use a lower-mu tube. But each 9-pin socket typically holds either a pentode or a dual-triode. If we started with a 12AX7, the pentode will very likely be an increase of gain for the stage in question, while using a lower-mu dual-triode means we also lose gain in a different stage which was perfect before our tinkering. You could use a dissimilar triode like a 12DW7/7247/ECC832, but they were never very common (I haven't seen a brick-n-mortar music store in the last 15 years that had anything other than 12AX7's) and what happens if you need more gain than the 12AU7 section will deliver?

#6. You could just use 2 resistors after the stage as-is to form a voltage divider which knocks down the signal by a fixed amount. The seems to address all our other gripes, and is very commonly done by Marshall, as well as almost every high-gain amp ever built.

[HotBluePlates]

Our "2 resistor" solutions (negative feedback, voltage divider) are really 2 resistors plus the original stage's plate load resistor. If you divide the plate load resistance into 2 resistors whose sum equals the original resistor value, you can do the job of the external voltage divider but with 2 total resistors instead of 3. Yes, it's penny-pinching on that scale, which matters if you're building 10,000 amps.

But it also simplifies the designer's task because nothing about the original gain stage has to be re-thought. Biasing and tube current doesn't change, input and output impedances don't change. Stage linearity and acceptable input signal doesn't change. Stage tone-shaping by way of cathode bypass cap doesn't change. The calculation involved is also very straightforward, and predicted results match build-reality very closely.

So there are other ways to achieve the same end result of reduced signal level/output. Simple schematic surveys tell us most often, amp companies used a method other than a split plate-load resistor. But if you have an otherwise perfect design, made in a production environment demanding every cent squeezed out of the cost, no wiggle-room on power supply draw, no budget to pay your engineer for a redesign and/or adding 1 resistor saves time/money over pursuing other methods to achieve the same goal, then the split plate-load is your ideal solution.

[Tone Junkie]

Good read gentleman . Thanks for the info i will use it to my advantage, one can never have to many arrows in his sling.
Bill

[terminalgs]

ahh,

so why doesn't he use a schematic to illustrate his point? 

I also tripped up on this:

"A common request is to increase the headroom  of a blackface Fender amp so that it doesn’t start breaking up so soon; adding a split load to the initial preamp stage would allow you to reduce the signal level going into the tone stack."

I read this and thought to myself "there is no distortion introduced in a typical BF amp prior to the 2nd gain stage, so why fool with the 1st stage (any part of it, including its output impedance)  headed to the BF TS when it can be done elsewhere with much less confusion and potential side effects."

Namely, if you want to shave 10% of the gain headed to gain stage #2, put a 100K resistor on top of the volume pot. right?

Either way, HPB,  thanks for clarifying what the author was getting at!     

Since then, I went and read the thread jojokeo mentioned, and I understand what jojokeo and sluckey's points were...

[jojokeo]

HBP,
Thanks for the great explanations. But I would like to say you have to include "the 7th way" of lowering the signal voltage to your list (even though you sort of did) it may be a little confusing for others being the split plate load resistors (SPLR) method.

Another important point is that whenever using a resistance in the signal path (whether grid stoppers, vol pots, commonly used voltage dividers (signal & load resistors), etc), no matter where they are, you lose an amount of high end to your signal. How much high end depends on how much resistance. Also the total resistances in the signal path can have a cumulative affect. So we employ the use of the famous bypass & treble bleed caps - to help retain high end (in simple voltage dividers) and the resulting "restored" and possibly "enhanced" high end to even unpleasant shrillness they provide depending on how, where, and value(s) used. So we and/or large manufacturers are also saving extra money by not having to use them to "make up" for this "problem". But don't also forget that less turrets/eyelets and board space is also realized by using less parts. So now with these "extra" capacitors they come with RC time constants, high-pass filtering, shelving effects, and other issues that enter into the equations that ultimately can unbalance our signals with too much bass and/or mid cut. This is important in the way the controls & amp responds and sounds at various settings. It may be acceptable to some but to me this many times is not as it's a PITA to dial in, and usually always comes with compromises. Many of these situations can be avoided with the use of SPLRs where they can be used and taken advantage of. There's nothing worse (to me) as using treble bleeds on vol pots where the tone changes significantly from nice & balanced to all highs & little lows down through the rotation. Using SPLR's the signal definitely remains more transparent and unaffected w/out need for the extra caps, IMHO.

Lastly, there IS an effect on lowering of output impedance depending on values used (and you don't get this added benefit with most of the other methods).
Zout=[(rp+1)*R2]/(rp+r1+r2)
Using 100K for a stock Fender plate load resistor total, 62.5k=rp of a 12ax7 @ 250V, R1=plate resistor, R2=B+ resistor;
Zout=
38k for R1=0, R2=100K, signal out =100%
35k for R1=10k, R2=91k, signal out = 91%
31k for R1=22k, R2=82k, signal out = 79%
26K for R1=33k, R2=68k, signal out = 67%
21k for R1=47k, R2=56k, signal out = 54%
19k for R1=51k, R2=51k, signal out = 50%


Respectfully,
keo

[HotBluePlates]

so why doesn't he use a schematic to illustrate his point? 

You & I see it quicker on a schematic. Steve used a physical layout to show it. Maybe he or his anticipated audience are more comfortable reading layouts? I dunno.

... thought to myself "there is no distortion introduced in a typical BF amp prior to the 2nd gain stage, so why fool with the 1st stage (any part of it, including its output impedance)  headed to the BF TS when it can be done elsewhere with much less confusion and potential side effects."

Namely, if you want to shave 10% of the gain headed to gain stage #2, put a 100K resistor on top of the volume pot. right?

Okay, there's another good way of reducing signal (resistor at the volume pot to set a minimum amount of loss that's always in the circuit).

And I wasn't exhaustive in my list... I can think of a few other obscure ways to reduce the gain of a given stage.

Bottom-line, maybe Steve chose a bad example if you can poke holes in the reasoning. It might have been easier to point at the Fender 6G16 Vibroverb and the split plate-load resistor at the 2nd stage of the Reverb channel. It uses 22kΩ and 82kΩ resistors instead of a 100kΩ resistor to reduce the dry signal by ~80% before mixing with the reverb output.

Non of us were there, but you might make a case that the 6G16 Reverb channel was designed by making an exact copy of the Normal channel, creating the reverb circuit, and mixing the two with 470kΩ resistors. Then perhaps the Fender designer found the reverb output was too-weak (or dry too-strong), didn't want to re-jigger everything, and used the split plate-load to knock down the dry signal enough to balance the reverb output.

Maybe Steve didn't think that was a compelling case, or perhaps that readers wouldn't see that as necessary (the 6G16 is the first-stab at putting reverb inside a Fender amp, and the answer for all later amps was to drop the split plate-load and use the 3.3MΩ resistor and 10pF cap instead).

Keo,
All great points. I kept away from some of the subtleties of the split-plate load, because explaining treble roll-off requires getting into tube Miller & stray wiring capacitance, while reduced output impedance lowers the total resistance the coupling cap is working with and so shaves bass response a bit.

You are right though that if you can achieve your objective with fewer parts or fewer (cap) circuits which make sweeping EQ changes to correct for minor deficiencies, then you usually get a better overall sound.

[Jack_Hester]

You're getting tripped-up by the names.

A "split-load phase inverter" is also known as a cathodyne or concertina phase inverter, or as a "phase splitter" in RDH4. This circuit divides the load resistance into 2 parts, one in the plate circuit and one in the cathode circuit, and generates 2 equal and opposite-polarity outputs.

Steve goofed when he used the term "split-load plate resistors" as he should have said "split plate-load resistors." This circuit spplits the plate load resistance into 2 parts, but both are in the plate circuit. This circuit also has but a single output.

Maybe one of the Admins can change the topic name to Split Plate Load Resistors, for the sake accuracy in discussion.  This was my goof when posting the topic.

Jack

[Willabe]

Jack I think you can do it because you started the thread. Just go back to the 1st post and change that.


              Brad     

[Jack_Hester]

Jack I think you can do it because you started the thread. Just go back to the 1st post and change that.

              Brad     

I gave it a try.  It will only let me change my post topics.  Everyone else's stayed the same.  Thanks for the suggestion, though. 

Jack

[Willabe]

Well at least now it shows up in the index page as Split Plate-Load Resistors. And my post just now shows the same, so all further posts will have it too?


            Brad      

[sluckey]

Not necessarily.  

If someone quotes one of the posts with the undesired subject title (like I just did on this post), that same title will carry over into the new post. Now if someone quotes this message, the undesired subject title will carry over as well. It's no big deal though because the index of messages will always show the title of the very first post. But if you reply the title from the first message will be used.

Since you can only edit your own posts, you cannot control the title of other people's posts. A moderator would have to edit every message individually to change the title of every message. I'm not aware of a "global subject title change command".

[Willabe]

Oh well, at least it's on the title page index.


            Brad   

[Jack_Hester]

And.........  Something else that I can't change.  Look at the title within the PDF that I posted.  It's still a good read, but the topic is better defined in the thread. 

Jack

[jojokeo]

I modified the thread and renamed the little schematic from my link above that illustrates what we're talking about here to help minimize the confusion there too.

[terminalgs]

I was working on a phase-shift oscillator, and the sine wave had more amplitude than I wanted.  I remembered this thread, and thought 'hmm!!!', I'll try that!  So the plate resistor was a 220K, I inserted a 47K between the 220K and the plate, and let the output signal "exit" at the junction of the 220K/47K resistors as described in this thread. 

plate resistance became 267K, so i suspect a bit more gain came form the triode, but the output signal is attenuated to about 80%.   

This did the trick!