|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
75.161.9.23
In Reply to: Ping Allen Wright: RIAA trimming posted by cheap-Jack on April 16, 2007 at 12:33:22:
IIRC, Allen wrote his comments in the context of passive EQ split into two parts, one part between the first and second tube stage, and the other between the second and third. Trimming is tricky if you're thinking of the usual 4 component RIAA network, whether passive or active, because changing any one component will affect all the time constants. The definitive paper on this was written by Stanley Lipshitz and published in the Journal of the Audio Engineering Society many years ago. If you're using a 4 component network it's best to just use his equations to design an accurate network.With a split EQ you can certainly do trimming since the time constants don't interact, however, remember that the best reverse RIAA networks are probably only good to within plus or minus 0.1 dB and Llipshitz equations allow you to design an RIAA network that is as accurate IME. You have to incorporate the output impedance of the preceding tube and the Miller capacitance of the following tube to get that degree of accuracy, however.
Follow Ups:
1st,
I agree with most but not all of what has been said below in response to your question.1/ If the design is really good and the given values of the RIAA network are EXACT, then building it from 1% parts MATCHED to 0.1% (channel to channel) parts will be IMO as close as is worthwhile trying to obtain.
2. It's easy enough to buy 1% R's - even 0.1% R's, but it's still not easy to get 1% caps in the values that may be needed, and of the sonic quality you want - so tweaking of the values in circuit may well be the easiest option.
3/ But to do this - you will need a VERY good inverse RIAA network, and while the Lipshitz papers tell you how to do it, again caps to the required accuaracy may not be eay to get.
4. My key focus is to MATCH the channels EXACTLY - this maintains the stereo image so that the harmonics of a given musical note are all coincident with the fundamental, not spread out in an unfocussed haze across the stereo picture. This can make a stunning improvement in your presentation, and also gives you a reference that allows the cartridge to be accurately aligned in the headshell 0 which is near impossible to do if the preamp has phase errors itself!
5. Having the curve less than perfect is not the big problem, as all recording/cutting engineers have messed with the frequency response from the moment the audio signal first came down the mic cable. So lacking a perfect inverse RIAA network, the focus should be on matching the response channel to channel.
***But there is a pretty good inverse RIAA available commercially - try the link below***.
6/ If you are building a tube phono with active negative feedback RIAA, you may be making a fairly hi impedance network - ie big r's (maybe 1 meg/82k and small caps (3n3/1000pF))- which means circuit stray capacitances may.will have a noticable influence on the response - or at least make important channel to channel differences.
Then you will need to use very small tweak caps, or best of all, some small value air spaced trimmers across the RIAA caps to balance up the hf end frequency responses at (say) 20kHz.
Good hunting, Allen
.
Hi.I am interested to trim resistors of the loop FB RIAA as Allen suggested due to tolerance of the RIAA caps being uncontrollable.
Any suggestion along this thought? Also where to get a reasonably accurate reverse RIAA amp which Allen did not elaborate.
The Lipshitz article is included in an anthology of AES articles on disk recording.Title: Disk Recording
Volume 1 Groove Geometry and the Recording ProcessLibrary of Congress Catalog Card No. 80-53466
Printed: October 1980
Publisher: Audio Engineering Society, Inc.
60 East 42nd Street
New York, NY 10165
By the way:The article is very math intensive and includes differential equations , which is beyond my skills. Computing the "exact values for the curve is no light weight task.
However there is a bit more to designing EQ networks than in these papers as he deals with, shall we say, "idealised amplifiers" compared to real world valve based designs.I've found the best way to design phono stages is with a simulator. Trim if you have to after that. I personally don't trim.
BTW, why would you want a loop feedback phono stage anyway?
Hi.Who wants to trim RIAA given proper design done by stimulation or whatever.
For RIAA EQ, we got to make sure the Cs & Rs are dead accurate to the designed values. But noway we can get any without tolerances.
What else we can do to kill such tolerances assuming you can afford the most exotic parts once we measure out such ever existing tolerance.Trimming is the practical answer.
There are still many good designs of all-in-one loop FB RIAA EQ phonostages available, being simple & less active stages which I personally prefer.
Though split RIAA EQ may give better transient response, but it "may or may not sound better" as Allen once quoted.
c-J
Please see my post below. I don't think you need to trim. I think parts tolerances (1% or better) are about as good as you can get in the "real world" without super expensive calibrated test equipment.OK so you say there are many good designs of all-in-one FB RIAA EQ stages available. Could you therefore please name one at least?
The RIAA standard is comprised of three time constants:T1: 3180 microseconds
T2: 318 microseconds
T3: 75 microsecondsThe recording curve and the reproduce curve are identical(unique to the phonograph). The curves are the same but opposite.
The best method to calibrate a reproduce amp is to build a record curve network that follows your signal generator. This is easily done using the same basic four part RC network as a passive filter network. Drive the phono preamp with this network and the output of the phono preamp will measure flat. Any errors will be quite easily measured. If a square wave is passed through it will be correct if both curves are identical.
The pitfall here is that you MUST get the record curve filter correct on its own or you will adjust the phono preamp to match the errors.
The reproduce curve is:
Frequency RIAA
20 19.36
22 19.24
25 19.04
28 18.83
31.5 18.57
35 18.29
40 17.88
44 17.54
50 17.03
55 16.61
63 15.94
70 15.37
80 14.59
89 13.93
100 13.18
110 12.54
125 11.65
140 10.85
160 9.90
190 8.67
200 8.31
240 7.04
250 6.77
315 5.27
340 4.80
380 4.16
400 3.87
430 3.49
480 2.93
500 2.74
540 2.38
610 1.86
630 1.73
680 1.43
760 1.02
800 0.84
850 0.63
950 0.26
1000 0.09
1100 -0.23
1200 -0.52
1250 -0.65
1300 -0.79
1500 -1.31
1600 -1.55
1700 -1.80
1900 -2.27
2000 -2.50
2100 -2.73
2400 -3.39
2500 -3.61
2700 -4.04
3000 -4.65
3150 -4.95
3400 -5.43
3800 -6.17
4000 -6.52
4300 -7.02
4800 -7.82
5000 -8.12
5400 -8.70
6100 -9.64
6300 -9.89
6800 -10.50
7600 -11.39
8000 -11.81
8500 -12.30
9500 -13.22
10000 -13.65
11000 -14.44
12000 -15.17
12500 -15.52
13000 -15.85
15000 -17.07
16000 -17.62
17000 -18.14
19000 -19.09
20000 -19.53
21000 -19.95
OK so now you have to build an inverse RIAA EQ better than the design for the amplifier you are testing because the test equipment should have an order of magnitude better accuracy than that being measured otherwise you are introducing errors at the signal end.So if you're aiming for +/-0.1dB, then you gotta have equipment that can measure this EQ deviation accurately. That's equipment accurate from 20Hz to 20kHz to 0.1% or better (0.1dB = 1%). I very much doubt many hobbyists have "calibrated" equipment to that standard. Now you have to ensure that no other strays enter the measurement setup otherwise you're going to be wrong here as well. Just work out how small stray capacitance has to be to put all this effort to waste. It is minutely small.
OTOH, if all you're aiming for is +/- 1dB then the above is a bit much but I think there is an easier and way less expensive way. Just simulate your circuit. The better your model, the better your result. I aim in the model to get the curve inside a 0.02dB window. However once it's as good as I can get it, I then rely on the circuit to look after itself. Buy 1% or better accurate components and bung them in. Chances are the worst it will be is +/- 0.1dB ish.
As most people build high Z EQ networks that are quite circuit dependant, all of this in many cases can seem academic. Fer instance, valve drift can change the EQ dramatically over time never mind strays etc.
For those hobbyists with a battery of test gear to the quality mentioned above (resistance and capacitance meter, freq gen, AC voltmeter t0 0.1% accuracy) then maybe the build and tweak method can work but I certainly don't have such good quality test gear that would better the model and build approach.
OTOH, maybe I got the worng idea here. If so, another viewpoint would be welcome.
I would also want to include for the cutter head rolloff as mentioned by Allen W. Jim Hagerman has a good paper on including this in the inverse EQ.
You don't "need" to build the record curve EQ to get the playback correct. It just make the trimming process easier and faster once you have it in your toolbox. You are correct that the ability to get one correct is no better than the other. I know it is a bit much to assume that everyone has a good set of test equipment and that it works correctly but I want to give the best method I know for the job.Your comment about the record system is valid. I have wondered why I should make such a fuss about the playback when the recorder is nowhere near correct. The Westerex systems are specified at ±3 dB within the audio band. I don't know the specs on the Neumann systems but I suspect that they are only marginal better at best. Add in the reproducer frequency response and it gets worse, especially if it is a high inductance type that is load sensitive.
I lucked onto some great HP test gear (at give away prices) that does allow better than 0.1% AC calibration accuracy across the audio band, but even with this gear it's a bastard of a job to get real accuracy.So, yes, get the best bits possible - match them before you install them, then tweak them to PERFECTLY match the channels to allow for the various wiring strays etc, and then quit worrying and just enjoy the music!
This post is made possible by the generous support of people like you and our sponsors: