About Feedback

It's been quite a while since my byline last graced the pages of Positive Feedback, but I have been keeping quite busy in the land of audiophile nirvana (or not), and this time I would like to share some thoughts on the design of pre-amplifiers, and power amplifiers utilizing little or no global feedback. I have long felt that feedback in amplifiers is somewhat analogous to herbs and spices in good cooking, they make a good thing a little bit better, but their injudicious use can ruin the result, and like salt, excessive reliance on global feedback distorts the perceived quality of the reproduced sound.

Global feedback, while it acts to reduce the overall quantity of distortion, has another interesting effect: the summation occuring in the inverting input also implies a multiplicative process, and the resulting product is then fed back from the output where the process repeats itself until hopefully the resultant product becomes insignificant. This error becomes highly significant under the highly non-linear operating conditions found during clipping, and can readily be observed by defeating the feedback path and adjusting the input drive levels downward by the ratio of open loop gain to closed loop gain and observing the difference in distortion spectra...

All amplifiers exhibit some non-linearity even under normal signal conditions, and all precautions to reduce this error contribution need to be taken during the design of any amplifier circuit, and my main criteria for the use of global feedback is does it confer some advantage other than a simple reduction in static distortions? Bear in mind that all amplifiers exhibit some delay from input to output, and in bandwidth limited designs this means that the delay from input to output can be long enough that the application of feedback may result in very poor performance in the time domain, and also result in dynamic intermodulation distortion. So if global feedback is to be judiciously applied, obviously the stage around which it is to be applied should be good enough without it, and then the following questions should be asked: Does it for example lower the output impedance for driving long interconnects without extreme sensitivity to shunt capacitance? Or rduce noise? Does it make the amplifier appear as a near ideal voltage source for driving a loudspeaker so that output voltage is not directly a function of output impedance? Or does it extend the power bandwidth of a power amplifier with a linear output stage? Global feedback is not a panacea, but a tool of valuable, albeit intelligently limited utility.

Design approaches exist for many problems that require no use of global feedback whatsoever. Consequently I am a strong advocate of local feedback, both for parameter control, and for controlling gain and bandwidth of individual stages, and additionally I strive to avoid the use of loop feedback around more than one current augmented voltage amplifying device. (Read mu follower or amplifier stage driving an improved cathode follower.) I also strongly advocate the use of triodes in low level circuitry, with appropriate types chosen for the particular application , so that global feedbackis not required solely for gain reduction. Balanced topologies have great applicability in power amplifier design due to their common mode cancellation (read power supply noise immunity) as well as their cancellation of even harmonics. I do use global feedback in some of my high power amplifier designs, but I have found it is possible to design moderate power level pushpull triode amplifiers without the benefit of any feedback at all except for a few dB in the driver circuitry, and still get excellent results. (See 6CK4 amplifier) Now I will focus on what in my view makes for a superior phono pre-amplifier.

Please note that this article was written about 15 years ago now and my thinking and circuit design philosophy has changed significantly. However these designs are still a good and relatively low cost entry to playing LP's using tubes. They are quite amenable to component upgrades and in most cases 12AY7/5751 and some other members of the 12AX7 family may be substituted for different required gains and per listening preferences.
- KRK 11/27/2005

Phono Preamp Time!

Like many commercial outfits I have chosen to design many of my circuits without the panacea (or Pandora's Box) of global feedback. I find it provides a level of harmonic truthfulness mostly lacking from conventional designs, and noise isn't really a problem in phono stages with paralleled triodes as input devices, as triodes generate noise which in the broadband sense is essentially random in nature, and the noise [power] contribution of the devices is halved each time the number of devices is doubled. In addition the smaller load resistances allowed by paralleled devices contributes significantly to lower noise as does the use of good quality metal film/bulk foil resistors - all carbon resistors generate far more noise, as much as 20dB worse than the average mf resistor. Equalization is accomplished with passive equalization, the accuracy of which is not influenced by gain variation or aging in the tubes, also it avoids the use of [extremely] high levels of negative (global) feedback at high frequencies, and the attendant ills of high levels of global feedback outlined previously.

I have spent spent a considerable number of years designing phono pre-amplifiers that utilize local feedback, and would like to share two relatively simple designs with you, the caveat being that they are the property of Kennedy Audio [currently suspended] AND ARE COPYRIGHTED, and therefore should not be used for commercial purposes without my express permission.

Design Number One: Simple And To The Point

The first design is quite simple, requiring only a small handful of parts and a single 5751 or 12AX7A per channel. The design uses a single triode in the first stage and is therefore most suitable for use with high output mc or mm type cartridges. This stage drives a simple 2 pole passive network which provides the inverse equalization characteristic required for de-emphasis of the RIAA equalized phono input. The [passive] equalizer circuit in turn drives the other section of the tube. Overall gain of the circuit is about 38dB @ 1KHz and the values given should result in performance better than +/- 1.0dB from 20Hz-15KHz without trimming. [ This stage is extremely sensitive to external capacitive loading and I recommend the addition of a follower if driving cables or a capacitive volume control - all except stepped attenuators!] One channel requires 2mA at +300Vdc, and 125-150mA @ 12.6Vdc. YOU MUST USE DC ON THE FILAMENTS OF THESE CIRCUITS, otherwise loud humming noise will eminate from your speakers. The output impedance is quite high at 40K ohms, so this circuit should be used near the associated line stage or amplifier or preferably as part of a modest pre-amplifier. Parts cost is quite low even with premium parts. (I recommend Holcotm resistors **except plate resistors** and MIT tm coupling caps except in the equalizer where Mallorytm 5% or better polystyrenes should be used. [See schematic basic phono]

A Higher Performance Model

The next circuit I am going to share with you provides higher gain, lower noise, and much lower output impedance than the previous design, at the cost of considerably increased complexity. This design features a parallelled triode front end driving a passive equalization network which in turn drives the mu-follower output circuit, and is entirely suitable for use as an external self contained phon stage. With the best parts and attention to detail the performance will rival the state of the art for a fraction of the cost. Gain is approximately 6dB greater @1KHz than the previous design, and the snr about 12dB greater. Equalizer accuracy can be as good as 0.1dB with selected resistors and capacitors in the equalizer network. Power requirements are 4mA @ +300Vdc, and 250 - 300mA at 12.6Vdc per channel. [See schematic deluxe phono.]


I suppose I should be kind and providesome sort of appropriate power supply designs for these pre-amplifier circuits. I would recommend the use of an LM317T (or 7812 + 1 diode drop to ground for 12.6Vdc) as the filament regulator, and any filament transformer capable of providing 1.0A @ 15 - 18Vrms should be suitable for this application. 600VCT @ 15mA is required for the B+, and infact a PAS 3 transformer would be quite suitable if used solely for plate power and filament power for the two 6BQ5 beam power pentodes used as series pass elements. The screen supply is well filtered to reduce output ripple. With care this circuit will produce less than 100uV ripple at the outputs. Power for the 12AX7 (ONLY!) error amplifier should come from the separate filament supply mentioned above. Note: I have observed high frequency instability in the 6BQ5's that can lead to RF oscillation. To deal with this I recommend directly bypassing the screen grids with .01uF /600V caps to chassis at the tube sockets. Properly functioning regulators will produce less than 1mV pk ripple typically. [ See Power Supply] You can also build just one section of the supply to power your phono stage if desired.


None of the designs described are difficult to build and could be successfully constructed on vector board or even point to point as desired. I encourage you to build these for your pleasure...


Note: All Known schematic errors have been fixed. 2/18/98
As promised these files have been replaced with unrestricted/printable PDF files as of 11/27/2005

Single Tube Phono Stage

Two Tube Phono Stage

One Possible Power Supply

©1998, 2001, 2005 By Kevin R. Kennedy