The long lost linear gain stage - diyAudio
Go Back   Home > Forums > Amplifiers > Tubes / Valves

Tubes / Valves All about our sweet vacuum tubes :) Threads about Musical Instrument Amps of all kinds should be in the Instruments & Amps forum

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools Search this Thread
Old 8th September 2004, 03:36 AM   #1
diyAudio Member
smoking-amp's Avatar
Join Date: Dec 2001
Location: Hickory, NC
Default The long lost linear gain stage

Back in the '30s, according to S. Okamura in the book "History of Electron Tubes", several techniques were invented to linearize vacuum tube gain circuits. Then, shortly after, negative feedback theory was invented, and these linearization ideas were forgotten. Long forgotten! One of these ideas is used, in slightly altered form, in modern IC circuitry design extensively today. And most people think of it as having been invented in the '60s. It was never patented however, for good reason. Today, this technology is called the current mirror (not the same as your usual current source by the way). And the current mirror can be configured to give linear current gain as well as unity gain. One never sees this circuit in tube designs today, despite all the talk of linearizing the circuit before resorting to NFB. So, its about time it got some publicity. Attached, and in the following posts are some schematic diagrams and explanation.
Attached Images
File Type: gif vtcurrentmirror.gif (8.5 KB, 2088 views)
  Reply With Quote
Old 8th September 2004, 03:39 AM   #2
diyAudio Member
smoking-amp's Avatar
Join Date: Dec 2001
Location: Hickory, NC
Attached are .gif images of schematics for various current mirrors including
a vacuum tube design.

CurrentMirror1 -- a bipolar current mirror with current gain of two, higher
current gain with use of more output transistors (or just use a physically
BIGGER trans. for output).
CurrentMirror2 -- bipolar current mirror, improved linearity via emitter
degeneration resistors.
CurrentMirror3 -- The Wilson current mirror, one of many variations on the
basic current mirror to improve some characteristic or other.
VTCurrentMirror -- the basic Vacuum Tube current mirror, the number of
diodes in series will vary depending on output tube geometry and diode
construction geometry (cathode area, and cathode to plate spacing are the
important parameters), R1 and R2 are optional for increased linearity, their
ratio must be in inverse accordance with the basic current gain ratio
produced by the diode/pentode so as to give equal voltage drops in
VTCurrentMirror2 -- just some biasing applied to the basic design
note: MosFets can be used in current mirrors also and their circuits are
similar in concept to the Vacuum Tube design

The basic VT Current Mirror operation is to use the diodes as a 2/3 power
non-linear resistor for input current to drive voltage conversion,
the pentode then does 3/2 power drive voltage to output current conversion,
with a net result of linear current gain. It looks simple in principle, but
a number of complexities come into play in practice. The current gain is
determined by geometric considerations between the construction of the
diodes and the pentode. Since 2/3 power and 3/2 power are non-linear
devices, they must also operate on corresponding identical (or inverse
actually) sections of their curves in order to correct each other. This
means not only the right bias points but the scaling of the two devices must
match. This scaling is determined geometrically, so not just any diode will
work with any pentode. In practice, this means stacking diodes in series to
get a close match of the geometric scaling for the pentode used. This can
require anywhere from one to a dozen or more diodes in series depending on
the tubes used. Considerable experimentation is mandatory, including a
variable bias supply and curve tracer or X-Y scope to look at current in
versus current out with a swept current input waveform. In addition, the
geometric spacing of diodes I have tried tend to be rather loosely
controlled in production, so just changing tubes with "identical" tubes can
change the number of series diodes required. A practical approach is to use
a few multi-diode tubes like 6JU8 and provide a jumper to select the number
of sections in series. Obviously, this will NEVER catch on for production
circuit designs, unless some special diodes are designed with strict control
of geometry. But for DIY, anything goes! By the way, using a diode
connected triode or pentode in place of the diodes would work in theory, and
even would allow convenient adjustment of the geometry factor by grid bias,
but in practice, the "inselbildung" or grid "island effect" ruins the
required curvature at small currents.
Attached Images
File Type: gif currentmirror1.gif (3.1 KB, 1616 views)
  Reply With Quote
Old 8th September 2004, 03:40 AM   #3
diyAudio Member
smoking-amp's Avatar
Join Date: Dec 2001
Location: Hickory, NC
more examples
Attached Images
File Type: gif currentmirror2.gif (3.5 KB, 1490 views)
  Reply With Quote
Old 8th September 2004, 03:42 AM   #4
diyAudio Member
smoking-amp's Avatar
Join Date: Dec 2001
Location: Hickory, NC
Attached Images
File Type: gif currentmirror3.gif (7.4 KB, 1511 views)
  Reply With Quote
Old 8th September 2004, 03:51 AM   #5
diyAudio Member
smoking-amp's Avatar
Join Date: Dec 2001
Location: Hickory, NC
I use an assortment of fixtures with a bunch of diode tubes wired in series with selectable series tap points. 6JU8 and 9006 are suitable diode tubes for making current mirrors with gain. A high gm pentode works best for high current gain. Its also possible to use the diode corrector idea on triodes with fixed impedance loads. (triodes are only strictly linear in theory with an infinite impedance load, the corrector allows linearity with a realistic load) Just requires additional resistance in the diode string to compensate for the plate voltage feedback in the triode. Generally, the triode correctors require a LOT of diodes in series however.
Attached Images
File Type: gif vtcurrentmirror2.gif (7.1 KB, 1501 views)
  Reply With Quote
Old 8th September 2004, 08:02 AM   #6
EC8010 is offline EC8010  United Kingdom
diyAudio Moderator Emeritus
EC8010's Avatar
Join Date: Jan 2003
Location: Near London. UK
Default Just so that you don't think nobody's listening...

A fascinating series of posts. As you say, current mirrors are very popular in the input differential pair of IC op-amps, and with good reason. However, their use also gives a clue to their main problem. IC op-amps rely on wrapping plenty of negative feedback around a high gain amplifier. In other words, the signal levels at the input stage are very small - and that's just as well because the semiconductor current mirrors restrict voltage swing to a few hundred millivolts peak to peak. (I experimented with a semiconductor current mirror on top of an ECC83 and was painfully reminded of the fact.)

I'm interested by your thermionic current mirror with diodes and pentode, as with a suitable choice of pentode and operating point, it should be possible to achieve quite a reasonable voltage swing. Have you made any measurements of swing?
The loudspeaker: The only commercial Hi-Fi item where a disproportionate part of the budget isn't spent on the box. And the one where it would make a difference...
  Reply With Quote
Old 8th September 2004, 08:11 AM   #7
diyAudio Member
Sch3mat1c's Avatar
Join Date: Jan 2003
Location: Milwaukee, WI
Send a message via ICQ to Sch3mat1c Send a message via AIM to Sch3mat1c
Looks to me like operation consists of using a non-ohmic device (the diode(s)) to correct for the similarly non-linear curve of a tube. Which depends on type (triode/pentode) and load (particularly with pentodes).

Seven Transistor Labs
Projects and Resources
  Reply With Quote
Old 8th September 2004, 05:56 PM   #8
diyAudio Member
smoking-amp's Avatar
Join Date: Dec 2001
Location: Hickory, NC
Default operational info

"Have you made any measurements of swing?"

Yes, first off, one must get into the mindset of using current levels when using this circuit rather than voltage. Its not a suitable circuit for a high impedance voltage input stage, but more suited to intermediate or even output stages of an amplifier. So, first off, the input of an amplifier would be a voltage to current stage, at small input signal levels, before entering any current mirror stages.

The non-linear resistance on the input is not nearly as bad as the exponential non-linearity in SS current mirrors, so actual scope measurements on the input will show some voltage swing, in fact, this swing will obviously match the operating voltage swing of the pentode grid in a given design. The SS current mirrors are much more tricky to get accurate matching because of the exponential characteristic, so usually only appear in ICs where identical transistors are readily available. The V tube current mirror is actually more practical for discrete component design due to its more gradual 3/2 power law. (and a trimpot in the circuit to set grid bias level and a selectable number of diodes in series and a fair amount of patience on the designer's part!)

Since a constant current gain is enforced, the input current will be simply related to the output current by this gain factor.
I have gotten current gains up to around 97 using a 9KC6 pentode (+13V on G3, 100V on G2) and 6x series 6JU8 diode sections. Using 9006 diodes about doubles the gain yet versus the 6JU8 diodes. What voltage appears on the output of course depends on the load resistance (at least for a pentode with sufficiently high output resistance spec) times the current.

Higher current gain comes from using a smaller diode and higher gm pentode. Too small a diode however may get one into noise problems if the current draw is too small. The more series diodes one has to use to get the scaling correct, the higher the input (avg.) impedance will be and hence the input voltage swing for a given input current swing, but this is not a freely adjustable parameter.

When thinking in current gain mode, instead of voltage gain mode, one actually would prefer the input node to be low impedance so it doesn't swing around and affect the previous stage's output resistance. Only the last output stage reverts back to voltage swing into a load. So an "ideal" current mirror amplifier would see no voltage signals whatsoever internally (Notice that this KILLS Miller effects nicely!) , just on the output. Of course, having a pentode for output with very high output impedance is another problem, so usually one would put a cathode follower, or similar, on the output to lower output impedance or use negative feedback to lower Zout.

"diode(s)) to correct for the similarly non-linear curve of a tube"

Exactly. The diode is a 2/3 power law non-linear resistor and the pentode is a 3/2 power law non-linear Gm. The output is the 3/2 power function of the 2/3 power function of the input. For a function of a function, the exponents get multiplied (rather than added as for a multiplication function). More intuitively, for a given pentode grid voltage, the diode's input current will be the 3/2 power of the grid voltage and likewise the pentode's output current will be the 3/2 power of the grid voltage. So input to output current ratio is linear. The current gain comes simply from geometric scaling of the devices, use a bigger pentode (or higher gm) and you get more current out for the same grid voltage. Use a smaller diode and you get less input current.

This current mirror operation is incidentally quite similar to positive grid operation of a triode or pentode, everything just built into the same tube then. Only problem is the diode (grid) current subtracts from the plate current, eventually hogging the cathode current and causing saturation. So, not as linear as the current mirror configuration.

Another variation on this current mirror idea, (in fact it was invented this way I think) is to reverse the order. Pentode first, then a big diode for load impedance. This is the voltage mirror, and gives linear voltage gain. Unfortunately, its not too practical, as the load is the diode, nothing else allowed except maybe just a grid load. Not to mention that it takes a big diode to correct a small pentode. (clunk -- tossing into the dustbin)

Another useful application of the current mirror could be as a linearity corrector for a SET amplifier, the diode input corrector (with additional resistor) gets applied to the input of the triode. This allows true linear operation of the triode with a real load resistance. Unfortunately, this is strictly linear only for a constant load resistance, so depends on speaker impedance variation being small.

  Reply With Quote
Old 8th September 2004, 07:05 PM   #9
diyAudio Member
Join Date: Jan 2002
Location: nowhere
I've been using SS current mirrors as loads on all my amps the last couple of years.
I dont know how EC8010 only got mV of swing with a ECC83 when I get hundreds of volts with every tube I've used so far. It's the SS mirror which is limiting the voltage swings since few P-channel MOSFETs or PNP bipolars go much higher(lower?) than -300V.
To me this topology is the easiest way to make amps with few stages.
  Reply With Quote
Old 8th September 2004, 07:51 PM   #10
diyAudio Member
smoking-amp's Avatar
Join Date: Dec 2001
Location: Hickory, NC
Default details details ...

Just a note, since the current mirror is a current to current gain device, it needs to be driven from a current source. So a typical amplifier circuit will consist of a conventional voltage to current tube stage (a pentode or pentode cascoded triode), but loaded by a current source or high value resistor, not a normal load resistor. Then capacitor coupled to the input of the subsequent current mirror stage. The same goes for mirror to mirror stage coupling. This might look a little crazy since a current source load on top of a current source driver would normally lead to undefined or rail pinned voltages, but the input diode impedance of the subsequent current mirror stage is what imposes order here, at least AC wise. With the capacitor coupling you still have an undefined load for the DC op. level of the current sources, so using a high value load resistor or a high L inductor load is called for. (Similarly, the diode input needs some steady DC current level too, to set the mirror stage DC operating point.)

But with current to current devices, another form of coupling is suitable. Using a current source load or high value resistor load to B+ for the V to current stage, the coupling to the mirror stage can be just a resistor from plate to diode input. This resistor is sized so that the idle current operating point (actually idle current of the driving tube minus idle current of the current source load) times the resistance gives the V drop down to the mirror stage diode/pentode grid level. (This residual current also sets the DC operating point for the diode and current mirror.) A capacitor CAN be placed across the coupling resistor for psychological comfort, but in theory at least, no loss of gain is caused by a coupling resistor in the series path of a current to current connection. The same current in, comes out the other end of the resistor.

This "current mode" thinking can be a little puzzling at times. For instance, to measure the current gain of a mirror stage on an X-Y scope display or dual channel scope .. or whatever, the input to the current mirror must be driven by a current source, not a voltage source. I usually just use a high voltage source with a high value resistor to the currrent mirror input. But remember, it is current you want to measure for linearity, not voltages. (Of course the output current is easily converted to voltage with just a conventional load resistor)

As the above resistive coupling technique hints, things can get complicated fast, since this provides DC coupling between stages.
Operating point drift can become a problem. So, unless you intend to make a DC coupled amplifier with a DC bias servo, I suggest some capacitor isolation or DC only feedback loops, or possibly a DC and AC global feedback loop. This is strange territory for usual tube design.


Oops, Just saw this:

"I've been using SS current mirrors as loads on all my amps the last couple of years. ..."

The circuit most people are using for a current source load for tubes is actually a current source (or load) and not a current mirror in its intended configuration. The confusion of a current mirror load has arisen because current mirrors are often used to make a current source (or load) by programming them with a fixed input level. But just the output connection is being used for the connection to the plate, which is at high impedance with large voltage swing capability. The INPUT terminal for a SS current MIRROR is typically a low impedance with limited voltage swing. Using a SS current mirror input for a plate load connection will generally pretty much freeze the plate voltage, but will echo the current on its output. Very similar to a cascode stage (or grounded grid), except conveniently inverts the current output.
Hope I clarified things, rather than confusing them further. The English language is running out of terms to describe electronics well these days.
  Reply With Quote


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off

Similar Threads
Thread Thread Starter Forum Replies Last Post
The most linear stage for large voltage swings - Lender's Circuit tiefbassuebertr Solid State 47 26th September 2013 04:54 PM
Gain stage with g2-feedback. jane Tubes / Valves 25 3rd July 2013 09:21 PM
LTPZen, Zen with long tailed pair input stage Allen Newby Pass Labs 18 28th January 2007 03:19 PM
Heretical Unity Gain line stage PLUS a little gain ? yagas Tubes / Valves 7 24th January 2006 01:30 PM
oh happy day a long lost cd revisited.. ibanezcollector Music 0 27th December 2005 03:14 AM

New To Site? Need Help?

All times are GMT. The time now is 06:38 AM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright 1999-2015 diyAudio

Content Relevant URLs by vBSEO 3.3.2