INTRODUCTION
Over the years, I have generally defended the Fisher phono preamp design -- not to extol its absolute EQ accuracy, but simply to say that keeping it in perspective, it is not all that bad at all, and in fact, very much in keeping with the performance of that from many other manufacturers of the day. I've also defended the design against a rather (in)famous modern day review by a well known expert, as executed in a 500C receiver, wherein he rather summarily trashed their performance, based on testing conditions that were capable of anything but accuracy.
In the early days of stereo where the RIAA curve's beginnings can be found, most manufacturers (including Fisher) didn't even provide specifications of EQ compliance, rather choosing to simply indicate that their unit offered a setting to produce the RIAA EQ curve. Where it was given (HH Scott for example), "good" compliance was indicated by a response curve that was specified as being within +/- 1 db of ideal. In the succeeding years of those heady days, the best manufacturers tightened this specification to produce within +/- .5 db of ideal as accuracy drew more attention, but measured accuracy within +/- 1 db was still quite common. Remember, the engineers of the day did not have CAD capability (let alone a computer), or even a pocket calculator for design work! All they had was a slide rule, chalk boards for group work, and pencil and paper for their notes and personal calculations.
On the production end of things, the early stereo units (including Fisher) often employed discrete components in the EQ networks, which just as often, were often not high tolerance components, let alone matched between the channels. Some manufacturers (like Dynaco for example) did use close tolerance parts. But most (including Fisher) simply used standard 10% components, which then added their own flavor of error into the mix. Of course, time and use then further works its own brand of error into these components as well.
In later years, Fisher and others used PECs for the EQ networks, which eased production issues, and also afforded more consistency between the channels and units of a given model. At the time that Fisher made this change, they also took the opportunity to tweak their (by then) almost universal RIAA network design (the X-202 however used a notable variation of it), to tighten the resulting curve produced for closer compliance. It wasn't a major change at all. In fact, it was small enough that so many decades ago when I first noticed the difference in component values, I assumed it more had to do with the manufacturing of all the EQ components into one small PEC package, rather than being the result of improving compliance accuracy. Today however, I realize that the move was indeed one of multiple goals, designed to ease production costs, improve consistency, and improve compliance.
Overall however, when the front end errors from limited design aids were coupled with the back end errors produced from component tolerances (discrete or PEC), its no wonder then that a specification (in the rare cases where provided) -- and "good" measured performance in the lab -- only achieved compliance within a +/- 1 db bracket. Add in also a generally casual attitude towards compliance at the time, and you ended up getting what you got, relying on the better manufacturers to do a better job of achieving compliance, which they did, but which was still only typically within the aforementioned +/- 1 db. With a 2 db (total) spread however, the errors become audible. Today, this can be improved on.
But improving the RIAA performance of yesterday's classics with today's technology can be fraught with its own set of problems. AK is filled with posts of those who have modeled various popular circuits, which they then proclaim as defective because the computer says so. Of course, these same people rarely if ever provide any proof of any such defect, instead relying on the computer to be judge, trial, and jury all at the same time, with their prowess to use it being king.
Done properly, modeling can be a truly wonderful design AID, but should in fact be considered as just that, and hardly be counted on for the final word all by itself -- at least when used with vacuum tube circuit design. Accurate modeling in and of itself requires great knowledge of: (1) The modeling software capability, (2) The hard data it has been programed with, (3) A thorough understanding of electronic circuit design, which affects (4) The accuracy in which the soft data (the circuit to be modeled) has been input into the system.
This is all on top of the realization that the results produced from modeling vacuum tube circuits should be considered as a ball park verification -- one that can be very close if the skill and effort are present to address all the myriad of variables involved -- but still, best used to verify actual results achieved, rather than taken as gospel all by itself. It's like a chain, which is always only as strong as its weakest link. Three of the four areas can be handled with incredible accuracy, but if one is off even by just a little, the results can then be way off from reality.
These requirements then -- for me -- potentially bring into question 99.9% of the modeling results heralded by most diyers which is why I pay only so much attention to modeling results. Instead, I have always preferred to have measured results be the outcome of my efforts -- which is the safe bet, but surely is more time consuming without enlisting the aid of modern design tools. However, I am a creature who is lucky enough to know what I don't know, and since I don't know the intricacies of ACCURATE modeling, I simply don't do it -- but I do endorse it in the hands of those who truly know what they're doing with it.
Against this backdrop, some of you may have followed my efforts recently with a Fisher 400C preamplifier ("A 400C Transformation"), and my goal to have it perform more in keeping with that of a Fisher 400CX/2. To date, those efforts have centered on improving the line stage performance, and eradicating the hum that so many of the stock units display. Along the way, I posted the some scope shots depicting the performance of the modified line stage, which perform admirably now with very little noise. I also posted some scope shots of the RIAA performance of the phono preamps in this unit, which still employed stock, but generally within tolerance EQ components for the RIAA network. They were less than admirable.
To put a wrap on that project, I wanted to address the RIAA results produced. There was not only a discrepancy between the channels, but a notable discrepancy between both of them, and ideal. It just wouldn't be right to have the tail end of the audio chain then perform so well, but not the front end.
As luck would have it, my efforts in working to improve the RIAA compliance in this unit were enhanced immensely by the efforts of George Ronnenkamp, who founded and operated Audio Regenesis. The quality of his work and the products of that company are without equal.
George was indispensable in first accurately modeling the originally 400C phono preamp design (as judged against my own measurements), then comparing it to that produced by the updated design as used in the 400CX(2), and finally modeling the results of my work to improve them, by providing validation of the measured results I achieved.
Modeling the basic Fisher preamp design presents problems from the get go, as apparently none of the modeling programs could properly deal with a grounded cathode (i.e. contact bias) 12AX7 circuit as employed in the second stage of the Fisher design. This was but one variable that had to be chased down, as were things like determining any differences between the reverse RIAA network he was using in his virtual models, versus the real one I was using in my lab. As I said earlier, modeling is fraught with potential problems, let alone those piled on by injecting vacuum tubes into the exercise.
Through out the process then, and parallel with my own real world work, George was independently working on optimum values for the network based on what the model was indicating. Ultimately, the level of conformity we both achieved independent of each other, but verified both by his modeling work, and my real world measurements, was really quite amazing.
Stay tuned!
Dave