diyAudio (
-   Tubes / Valves (
-   -   Schade Feedback (

TheGimp 22nd November 2010 03:56 AM

Schade Feedback
I've been working on a guitar amp over at TAG (The Amp Garage) and changed a design by Ken Fischer which is known as a "Trainwreck Liverpool"* to use Soviet tubes.

The Amp Garage :: View topic - Russian TrainWreck - Expresso Lite(Liverpool), Soviet tubes

I found that the background noise (hiss) was unacceptable unless I was very careful in my selection of the input tubes (originally 12AX7s which I replaced with 6N2Ps).

I first tried paralleling two triodes to reduce noise and this had limited success, in that it was still sensitive to tube selection.

In order to minimize the noise and make the circuitry less susceptible to "tube" to tube variations, I then tried a j-FET front end in conjunction with lower first and second stage gain which was quite successful.

Lastly I replaced the first triode stage with a small signal Pentode. which had phenomenal gain. I used Schade feedback to reduce the gain to a manageable level (Av70) along with reducing the second stage gain by a factor of two. The two in combination reduced the hiss to about 10mV which is quite acceptable for a guitar amp (although too much for an audiophile amp).

It appears to me that Schade feedback is classical feedback with the output impedance of the preceding stage as the input impedance in the classical equation:

A' = A *( Rfb/Rfb+Ri)

The Schade resistor can be tied to the gate of the Pentode if a coupling cap is used to isolate the DC voltage and prevent bias shift.

Therefore, the gain of a pentode with feedback Rf (ac coupled) and an input resistor Ri combines Ri plus the output impedance of the preceding stage Ri' for an overall gain of:

Vo= Vi*A*(Rfb/Rfb+(Ri+Ri')) Where Vi is the output of the preceding stage.

or, am I off in the weeds again?

*All rights reserved by the Fischer family

Sch3mat1c 22nd November 2010 04:41 AM

Please, call it "shunt feedback". "Shade" is a useless, arbitrary assignment, and not even consistent with anything else named (for instance, a guy named Thevenin devised the eponymous theorem).

For low noise, two cascaded (not -oded) triodes are better than one pentode.

The feedback equation you are looking for is:
Vo/Vin = G / (1 + G*H)
G = open loop gain of the amplifier
H = feedback voltage ratio
Vo/Vin = overall (system) gain

This also applies to linear, time-invariant systems, where G and H are a Laplace or Fourier domain transfer function.

Applying shunt feedback direct to a proceeding plate forms a voltage divider of H = Rp / (Rf + Rp) gain, where Rp is the total plate impedance (including plate resistor and grid leak).

Note that, for G >> H, this reduces to Vo/Vin = 1 / H, which is familiar from op-amps. A CCS-loaded pentode (mu ~ 4000), with feedback to reach an overall gain of 50 or so, would be an excellent case of this. Incidentially, such a stage would have vanishingly small distortion (< 0.1% at full output?), so naturally, the audiophools don't want you to know about it. :devilr:


Wavebourn 22nd November 2010 04:52 AM

Schade feedback is feedback proposed by Schade to correct 6L6 tube for SE usage.

The problem with parallel (shunt) feedback in low level amplifiers is their input resistance. For less noises you want lower values of resistors. For high input resistance you want higher values of resistors.

Probably, it will work well if you use it with SRPP (for the example, Michael Koster's post in recent Gyrator thread)

Sch3mat1c 22nd November 2010 08:32 AM

I was considering shunt feedback to reduce the output impedance of my oscilloscope stages. Should extend bandwidth better than Rp alone (which requires high current tubes and lots of extra dissipation to accommodate).

More important than noise, large resistors would cause excessive capacitive loading, screwing up bandwidth. Scope circuits require low impedances and high linearity. Open loop impedances are minimized by choice of high Gm, moderate-current tubes (Tektronix was fond of 6DJ8 and 6DK6), and NFB. Cathode degeneration is easiest, but I'm thinking shunt would be advantageous.

Incidentially, 100kohms makes 5.75 uV over the audio band, which is a good -45dB from microphone levels (~1mV). Tubes are pretty noisy in comparison, so don't worry too much about noise from feedback resistors.

Input impedance is a big concern for input stages, and either a conventional stage, or a buffer (CF?) would be a reasonable choice to fix this. (Incidentially, if you use a CF, don't forget the series resistor in the feedback path, otherwise your feedback resistor will try working into the CF's output impedance, which isn't going to work well!)


DF96 22nd November 2010 10:52 AM

So-called Schade feedback seems to be coming as fashionable as SRPP etc. He simply applied the old "anode follower" technique to an output stage. That is basically it. Now, as discussed in another thread, "anode follower" is a daft name for what is essentially Miller-style feedback (think inverting opamp) but it is preferable to "Schade". "Anode follower" is what the textbooks call it, so we should do the same.

The weakness of the anode follower is that it either requires a big input resistor (which can be noisy, and interacts with Miller capacitance) or it relies on the output impedance of the previous stage (as the OP thought) and so introduces a new source of distortion as output impedance can be more non-linear than voltage gain.

kenpeter 22nd November 2010 02:22 PM

Schade = sameoldsameold anode follower for sure.
But Schade's Fig35 proved that local feedback from
plate to grid is what gave a Triode Mu to begin with.
What the screen taketh away, Schade givith back
in spades! Or Mu or whatever...

Some things called Schade today, deviate pretty far
from any of his schematics. But what really matters
is wether said design adheres to principals of Fig35.
I'd accuse the worst pretenders to Schade approach
GNF. Oft forgotting the key word is "Local".

Michael Koster 22nd November 2010 04:07 PM

There isn't a good name for it. I think we on this forum started calling it "Schade" about the time I started a thread titled "What if O. H. Schade had MOSFETs?".

I've started calling it "stage local plate-to-grid feedback" because that's what it is. Other techniques like feedback to the cathode are different over the wide range of operating conditions (how to get a D3a stage to input 150V P-P signal without an attenuator? good luck using CFB...)

Usually, an input resistor is addeed to the output impedance of the preceeding stage to stabilize the gain and provide a finite impedance for the preceding stage to work into.

There are lots of ways around the input impedance problem. one way is to use a pentode or MOSFET as the first stage and load it with the plate feedback current of the next stage. Examples abound. Another way is to use a mu-follower etc. to buffer the drive to the input resistor. Also, some applications are happy with relatively low impedance input circuits. The practical range in my experience is from 1K to 10K ohms as an input resistor, and about the same equivalent load if using the technique of feedback to the driver plate (drain). Schade's original circuit used a coupling transformer to sum the feedback from the plate, thus avoiding the impedance issue.

Something hot mentioned is that the gm of the tube is preserved, i.e. if you have a 10 ma/v gm pentode you can adjust the feedback to get any ratio of mu and rp that preserves the 10,000 gm. For example, you could design for mu of 10 and Rp of 1000 ohms, or mu of 5 anf Rp of 500 ohms. So of course when using the voltage divider feedback you need input voltage to drive the feedback network accordingly. Using the 2 stage MOSFET or pentode (or even a high Rp triode) driver circuit the needed drive voltage is minimal; the feedback is current based over the gm of the driver.

The noise of the stage is a function of it's input resistance. The input resistance is limited by plate-grid capacitance and desired upper Fc and would not practically be as high as 100K in most circuits. More like 1K to 10K inmy experience.

5.75 uV self noise is about 20 dB too high for a 1mV sensitivity microphone in studio use. Luckily, microphones like 1000 ohms load or so, which reduces the resistor noise to ~600 nV. Tubes can be this quiet or even better (D3a is 65 ohms ENR in triode connection).

But I don't see though how local feedback would extend bandwidth. If you add too much input resistance it could reduce the upper Fc due to Miller effect. I have not had a problem using even triodes up to about 10K grid resistance in these circuits.



smoking-amp 22nd November 2010 05:31 PM

2 Attachment(s)
And then there's the new kid on the block: "Magic" Schade feedbacks or Output Voltage plus Differential Current Sensed Series Feedbacks. OV+DCSSFdbk or "Magic" for short. Gets rid of crossover distortion for class AB. (requires some overlap in output stage conduction, driver is class A differential)

European Triode Festival and Crossover Notch Distortion and New OTL Design
(2nd diagram up from bottom)
Hope it's OK to attach the diagram from Tubecad, Dec 20, 2006.

And then RCA RC-30 Handbook, page 696, attached.
Here the cathode feedbacks to the driver are the equivalent, except would need a CCS tail on the driver. The current sense resistors are implicit in the DC resistance of the primary windings.

edited to remove distortions!

Wavebourn 22nd November 2010 05:32 PM

According to what we learned in Russia, it is "Parallel Feedback By Voltage". No confusion is possible: a fraction of output voltage is applied in parallel with input signal.
When people don't have systematic academic education they tend to view examples in popular books and magazines as revelations, attributing them to authors of books and articles, calling them by their names.

Here again is a picture I drew long time ago for students:

artosalo 22nd November 2010 05:43 PM


But I don't see though how local feedback would extend bandwidth. If you add too much input resistance it could reduce the upper Fc due to Miller effect. I have not had a problem using even triodes up to about 10K grid resistance in these circuits.

If you just add a resistor in series with the grid, it will directly cut the high frequency response since it forms a low pass filter with Miller capacitance. However, when the feedback resistor is also added from anode to grid, the situation changes. The smaller the anode-to-grid resistor is, the lower the gain is, but also with the same the high frequency response will be extended. And distortion reduced, input impedance reduced, output impedance reduced, etc.etc....

All times are GMT. The time now is 10:35 AM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 18.75%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio