Because they are, or at any rate were, when they first appeared. Active one-stage RIAA could now be designed in a minute or two.
Why people seem so reluctant to use them, copping out with 2-stage passive RIAA, is hard to know.
But why on earth would you want to repeat the work of deriving the equations???
And permit me to say that disrespecting a valuable bit of work as you have does not impress me much.
No, that is not the case. In the single-stage active RIAA the time-constants interact, and the solutions for component values are certainly not trivial. And that is before amplifier shortcomings are considered.
A quick way to see this in action is to set an all active RIAA up in LTspice.
Then calculate the values using time constants and the result is you get big errors. The only way out is a lot of tedious experimentation if you want a better result with this approach. This was the state of affairs wrt all active prior to Lipshitz's paper and he notes this somewhere in the document.
Next step, repeat exercise but use his equations. Result is you'll probably be within .5 dB across the audio band - maybe some tweaking on the output post filter to get the response really accurate in the c. 20 kHz region. I routinely get sub 0.25 dB 20 Hz to 20 kHz and 0.1 dB with a bit more patience.
I spent quite some time playing with this (time constant vs Lipshitz) and the Lipshitz stuff works and especially so if you are talking all active.
I've put a write up on my website about it ('RIAA Equalization Amplifiers) it also discusses the Baxandall active/passive equalizer.
Then calculate the values using time constants and the result is you get big errors. The only way out is a lot of tedious experimentation if you want a better result with this approach. This was the state of affairs wrt all active prior to Lipshitz's paper and he notes this somewhere in the document.
Next step, repeat exercise but use his equations. Result is you'll probably be within .5 dB across the audio band - maybe some tweaking on the output post filter to get the response really accurate in the c. 20 kHz region. I routinely get sub 0.25 dB 20 Hz to 20 kHz and 0.1 dB with a bit more patience.
I spent quite some time playing with this (time constant vs Lipshitz) and the Lipshitz stuff works and especially so if you are talking all active.
I've put a write up on my website about it ('RIAA Equalization Amplifiers) it also discusses the Baxandall active/passive equalizer.
Last edited:
Absolutely so.
It should be better than that. I get to within 0.05 dB every time, though to reach 0.01 dB a little tweaking of R1 is required. I don't know why this is, but it is not due to finite opamp gain.
I assume the 'output post filter' is what I call the HF correction pole?
Yes indeed.
I agree and have used them continuously since I designed my first phono stage back in 1988. I wrote a little program to do it in basic and have since forgotten most of the details. I use the single stage passive implementation.
Walter Hauessermann's design report of the Saturn V Guidance System: Useful Audio Links
The issue with RIAA is that people working in isolation sometimes make mistakes - it takes someone with specialist knowledge or insight to improve things. For all active RIAA EQ network design, Lipshitz was the man.
No dig on my part meant at you or anyone else here BTW - no doubt, for low distortion, low noise opamps, you are the man!
No dig on my part meant at you or anyone else here BTW - no doubt, for low distortion, low noise opamps, you are the man!
Didn't even occur to me. My point was simply that by your description folks built something that was easily measurably wrong for quite a few years. In any case I apologize if anything was trivialized, I find the algebra can be tedious but it is just algebra not tensor calculus.
Last edited:
Geez fellers, it's just a noninverting opamp and circuit analysis. You solved these a hundred dozen times as an EE undergraduate at 22 years of age. For capacitors apply Z=1/sC and then derive the transfer function just like all the other homework problems. Presto, ten minutes later you have The Lipschitz Equation.
Want to include the effect of finite opamp bandwidth? Assume a single pole rolloff, insert this pole into the opamp block and re-derive the transfer function. Ten minutes later you have The Lipschitz Equation V.2.
Want to include the effect of nonzero opamp output resistance? Insert it into the opamp block and re-derive the transfer function. Ten minutes later yo have The Lipschitz Equation V.3
22 year old you would consider it a medium difficulty homework problem, not some kind of holy grail. Messy but no creativity required.
Want to include the effect of finite opamp bandwidth? Assume a single pole rolloff, insert this pole into the opamp block and re-derive the transfer function. Ten minutes later you have The Lipschitz Equation V.2.
Want to include the effect of nonzero opamp output resistance? Insert it into the opamp block and re-derive the transfer function. Ten minutes later yo have The Lipschitz Equation V.3
22 year old you would consider it a medium difficulty homework problem, not some kind of holy grail. Messy but no creativity required.
The problem perhaps was not that the math was difficult (although I am no mathematician), but that the incorrect assumptions were made. Incorrect assumptions and good math still give the wrong answer.
Easy to say this is 22 year old stuff. And for math homework perhaps it is.
I'm thinking of the $1.10 bat and ball analogy right now . . . .
Easy to say this is 22 year old stuff. And for math homework perhaps it is.
I'm thinking of the $1.10 bat and ball analogy right now . . . .
In ten minutes? Do me a favour.
Nobody is saying that the Lipshitz paper was bursting with new and amazing insights. Its importance is that L was the first guy to grit his teeth and actually do the sheer hard work of the algebra. It's real turn-the-paper sideways stuff; it's not that it is inherently hard to do, but it's certainly hard to do without making a mistake somewhere. And of course it was all done by hand without symbolic manipulation software.
Let's have a bit less sneering and a bit more respect.
Well the eqns (for the various RIAA networks) are certainly worth including in your book. Not everyone has access to AES papers.
In da early days, only LEAK & Quad had correct RIAA responses. Great Guru Baxandall did the maths for Quad.
I worked through one set of eqns cos I wanted to use a different RIAA network for a single OPA stage. It took me a lot longer than 10min though I had more than 1 brain cell in those days.
It was a great relief when the Lipshitz eqns came out. I wish I'd copied them down when I had access to JAES.
Tried it again just now even though I'm a lot older than 22 years. From start to finish took me 7 minutes and 47 seconds according to the stopwatch on my smartphone. Three pages of handwritten equations @ about 5 equations per page. It's not that hard a problem, publishing the solution is not that worship-worthy an achievement.
Today of course we'd just throw the whole business into a numerical optimization routine and let our multi-gigaflops laptop computer crunch out component values at nine digits of precision. Easier still, if your circuit simulator includes a built in optimization routine (as HSPICE does, see below), you can simply tell it to find the optimum component values that give the minimum error. Couldn't be easier, in today's modern world.
_
Attachments
We grovel at your feet O Guru Johnson
.. and I'm sure everyone who buys Self's new book will be conversant with Matlab, HSPICE etc
Excel (spreadsheet program) is very inexpensive and the numerical optimization package for Excel is free. Quite a number of us diyAudio people use it to optimize SPICE model parameters, i.e., best fit between measured transistors and simulated transistors. Once you learn how to use the optimizer, you can apply it to RIAA network design. It's so 21st century.
Again Python is all free and has a particularly good optimizer module with a lot of strategies to choose from. I tried to tell Charlie L. that Excel can/will give garbage answers for ill-conditioned problems (i.e. poles outside the unit circle for IIR filters) not sure he paid attention.
One issue Doug is that if someone wants to, say, try some of the other older equalizations they might get lost and apprising them of the software tools available could be useful.
Last edited:
and I'm sure everyone who buys Self's new book will be conversant with Matlab, HSPICE etc
Why would you buy a book on making it yourself and not be interested in expanding your horizons?
Is it possible to build a program using Python that will run on somebody else's computer WITHOUT requiring that person to download Python libraries or DLLs or other stuff? I think this is called "fully static linking" and the last time I looked into Python, you couldn't do it. The computer that runs a Python program, must first install one or more helper libraries or environments or both. (Back when I investigated). You couldn't build a standalone .exe that would run on other diyAudio member's computers, UNLESS each and every person who ran that program, also downloaded and installed something like a DLL.
(Tcl/TK does offer this capability. I can send you an executable program that will run on your computer immediately. But Tcl/TK is terrible at numerical computation).
(Tcl/TK does offer this capability. I can send you an executable program that will run on your computer immediately. But Tcl/TK is terrible at numerical computation).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.