So what, this was still a listening test (no ABX) with statistical controls and data processing.
Do-you really think the placement of speakers in a room are of such an importance to evaluate the quality of a speaker, as long as the triangle is respected ?
On my point of view, and apart extreme cases, it affects the listening comfort and brain fatigue, nothing more.
Curious to read your opinions.
On my point of view, and apart extreme cases, it affects the listening comfort and brain fatigue, nothing more.
Curious to read your opinions.
In regards to correct RIAA equalization, it was not easy or even obvious, 45 years ago. This was because the different RC time constants were not buffered from each other, and so interacted, making more careful adjustments of the RC to deviate slightly from first computed values. We 'fixed' the Levinson JC-2 in a magazine article in the 80's with just one resistor value change for the most critical deviation. Of course, I should have carefully checked the RIAA rather than just copy it from the more expensive Levinson LNP2 that Richard Sequerra used, but I trusted him.
Gee, that Black hole picture, sure looks a lot like what we figured it would look like. Still, I am impressed that they got it anyway. I would have been an astronomer if I had not be dissuaded in high school that it there was no employment in the field, so I drifted for awhile, until I found electronics.
In the 70s, I used either a programmable calculator or an IBM AT, for RIAA and other calculations.
Before that, I did the calculations by hand. A slide rule was woefully inadequate for precision work.
Also in the 70s I designed and built an inverse RIAA and used it routinely. In fact I still have it.
J. Gordon Holt of Stereophile also used an inverse RIAA at that time (built for him by Ed Dell of
The Audio Amateur, who lived across the street), but for listening tests. He connected his
tape playback deck to the inverse RIAA, and used that source for phono stage evaluation
using his own master tapes, in addition to playing LPs.
Last edited:
But his results can be done with pencil and paper. There are not many nodes you can write the equations and solve them symbolically. The idea of "inventing" inverse RIAA seems a silly way to put it IMO. The inverse came first and I doubt any cutting lathes are extremely accurate either.
The basics at the level of the hi-end.
Last edited:
The FORTRAN-II programming language included the "COMPLEX" datatype in 1963. Making it a straightforward exercise to calculate
H(s) = Vout(s)/Vin(s) = Av(s) / (1 + Av(s)*Beta(s))
Where "s" is the Laplace variable, s = alpha + jw
Av(s) is the open loop transfer function of the amplifier used in the RIAA stage
Beta(s) is the transfer function of the RIAA feedback network (calculated using the design values of R1, R2, C1, C2, etc.)
A program in FORTRAN-II would be a quick and easy way for 1963 era EEs to doublecheck that their RIAA equalizer circuit, had a frequency response very close to the ideal RIAA curve. The program could of course calculate the ideal RIAA curve at the same time, since the three RIAA timeconstants are known, fixed quantities. Then it calculates and outputs the "error", the deviation from ideal, at each frequency: error = [Hckt(s) - Hideal(s)] / Hideal(s). Duck soup.
If a 1963 era EE modeled her amplifier as a finite gain block with single pole response, she would get better accuracy in this approach, than Stanley Lipshitz's 1979 equations in JAES, 16 years later. In uncorrected form, Lipshitz assumes an ideal amplifier having infinite gain and infinite frequency response. Because that assumption makes the expressions simpler. Programs in FORTRAN-II, on the other hand, only grow by five or six lines of code when they include Av(s) = Const / (1 + s*tau) instead of Av(s) = HugeNumber.
_
H(s) = Vout(s)/Vin(s) = Av(s) / (1 + Av(s)*Beta(s))
Where "s" is the Laplace variable, s = alpha + jw
Av(s) is the open loop transfer function of the amplifier used in the RIAA stage
Beta(s) is the transfer function of the RIAA feedback network (calculated using the design values of R1, R2, C1, C2, etc.)
A program in FORTRAN-II would be a quick and easy way for 1963 era EEs to doublecheck that their RIAA equalizer circuit, had a frequency response very close to the ideal RIAA curve. The program could of course calculate the ideal RIAA curve at the same time, since the three RIAA timeconstants are known, fixed quantities. Then it calculates and outputs the "error", the deviation from ideal, at each frequency: error = [Hckt(s) - Hideal(s)] / Hideal(s). Duck soup.
If a 1963 era EE modeled her amplifier as a finite gain block with single pole response, she would get better accuracy in this approach, than Stanley Lipshitz's 1979 equations in JAES, 16 years later. In uncorrected form, Lipshitz assumes an ideal amplifier having infinite gain and infinite frequency response. Because that assumption makes the expressions simpler. Programs in FORTRAN-II, on the other hand, only grow by five or six lines of code when they include Av(s) = Const / (1 + s*tau) instead of Av(s) = HugeNumber.
_
Last edited:
No, they are not. The Westrex 1700 system is basically a bunch of op amps inverse to the ubiquitous preamps of the day. That was what came "stock" with the Scully lathe. Neumann and Ortophon had their own electronics.
Capitol, RCA, CBS, etc used Scully lathes for much of their existence as vinyl companies.
Mostly they were built out of 5% and 10% parts. Some alignment was possible, but it was not "precision". It was accurate enough.
Cheers
Alan
Last edited:
Back then I first made symbolic calculations on a programmable calculator, and soon after that
on an IBM AT using Maple, a nice symbolic math program. If the closed loop gain factor at DC
is defined as Adc (a constant), then the ideal RIAA phono equation for playback is a zero and
two poles: T(s) = Adc x (1 + s x 318uS) / ((1 + s x 3180uS) x (1 + s x 75uS))
The circuit analysis will find a similar equation in terms of the Rs and Cs, and you then equate
like terms (in s) of the two equations, and solve the resulting equations for the R and C values.
For an ideal non-inverting circuit, there is an extra 1+ term added to T(s), so T'(s) = 1+ T(s).
This forces the gain to unity at high frequencies, instead of the ideal -6dB/octave asymptote.
The algebraic equations are much more involved if the circuit's finite open loop gain curve is
taken into account.
Last edited:
I was saying the RIAA calculations are not beyond doing by hand literally. I regret throwing away a yellow engineering pad from 1969 (or so) with a complete hand analysis of the 741 by the great Alberto Billoti, I assume done at the behest of Dave Fullagar.
Funny how everyone here is pratting on about 'how easy it is'.
Yes, from the viewpoint of 2019 RIAA is easy - hell, even I can do it and get reasonable measured results.
The simple fact - and I said this to Mark Johnson about a year or so ago - is that NO ONE DID IT RIGHT until Lipchitz came along in 1979 and went back to basics and derived the formulas in all their glory and everything after that was easy.
There are plenty of 65+ year olds around here besides JC. Why didn't you put the equations up in 1975?
Remember, the issue with RIAA before Lipshitz was with all active EQ - the split passive designs were ok from what I have read.
(Anyone know of he is still with us?)
Yes, from the viewpoint of 2019 RIAA is easy - hell, even I can do it and get reasonable measured results.
The simple fact - and I said this to Mark Johnson about a year or so ago - is that NO ONE DID IT RIGHT until Lipchitz came along in 1979 and went back to basics and derived the formulas in all their glory and everything after that was easy.
There are plenty of 65+ year olds around here besides JC. Why didn't you put the equations up in 1975?
Remember, the issue with RIAA before Lipshitz was with all active EQ - the split passive designs were ok from what I have read.
(Anyone know of he is still with us?)
I did, at the University of Illinois-Urbana, even in CS101. We had an IBM S/360, and ran ECAP on BASIC. We also used PL/1 and FORTRAN. Yes, all on punched cards, run only in huge batches about once a week.
Last edited:
Computer programs allow you to split up the work into individual blocks. Small individual blocks. The amplifier block is K/(1+tau*s). The feedback leg is Z1(s) = 1 + K1*s / (1 + K2*s + K3*s^2) where K1, K2, and K3 are easily derived by hand. The series leg is Z2(s) = (1 + pole(s))/(1 + zero(s)).
Then the cool thing is, you let the computer calculate Beta(s) = Z2(s)/Z1(s) without ever writing out a closed form algebraic expression for Beta. Eliminating a source of mistakes.
Then you let the computer calculate Av(s) / (1 + Av(s)*Beta(s)) without ever writing out a closed form algebraic expression for this. Eliminating another source of mistakes.
And respectfully spoken, just because you are unaware of any 100% error free published equations, does not mean that people didn't design error free RIAA circuits. Either by their own unpublished hand analysis, by FORTRAN calculations, or even by painstaking tweaking in the lab. In fact, reading the introduction to Lipshitz's paper, it seems clear that Lipshitz himself had seen and had measured RIAA circuits which were error-free. It's just that they were an embarrassingly small number of the units he measured. Attachment below.
(A person who wrote FORTRAN code in 1964 was probably born before 1940, and thus >79 years old today)
_
Then the cool thing is, you let the computer calculate Beta(s) = Z2(s)/Z1(s) without ever writing out a closed form algebraic expression for Beta. Eliminating a source of mistakes.
Then you let the computer calculate Av(s) / (1 + Av(s)*Beta(s)) without ever writing out a closed form algebraic expression for this. Eliminating another source of mistakes.
And respectfully spoken, just because you are unaware of any 100% error free published equations, does not mean that people didn't design error free RIAA circuits. Either by their own unpublished hand analysis, by FORTRAN calculations, or even by painstaking tweaking in the lab. In fact, reading the introduction to Lipshitz's paper, it seems clear that Lipshitz himself had seen and had measured RIAA circuits which were error-free. It's just that they were an embarrassingly small number of the units he measured. Attachment below.
(A person who wrote FORTRAN code in 1964 was probably born before 1940, and thus >79 years old today)
_
Attachments
Last edited:
In 1974 I was working part time in the stereo division of a parts distributor. We had a giant switch board that allowed a preamplified turntable to connect to any amplifier or receiver to be switched to any pair of loudspeakers. The problem was that HH Scott receivers expected an auxiliary input level of 1 volt versus all the others 100 millivolts for normal listening level.
So I did the math to do an inverse RIAA curve to feed the phono inputs equalized and volume adjusted signal to the RCA jacks.
The other guys were amused to see me turn the paper sideways to fit my equations. Arithmetic was done with a then new electronic calculator. We also sold those.
So I did the math to do an inverse RIAA curve to feed the phono inputs equalized and volume adjusted signal to the RCA jacks.
The other guys were amused to see me turn the paper sideways to fit my equations. Arithmetic was done with a then new electronic calculator. We also sold those.
Last edited:
Yes, I did calculations by hand for RIAA (at first), but it was very difficult to not make mistakes
when writing down all the algebraic equations by hand. Think of all the parentheses you have to
keep track of. For paper, I used the backs of old computer paper printouts. Some of these
calculations written by hand were a dozen or more pages long, and I write small.
Last edited:
- Status
- Not open for further replies.