It's been fun, time for another core dump of all that I have learned. I hope to put it in a context that anyone with a little incentive can make use of it. 0.0003dB conformance at 192K with only 7 or 8 multiply accumulates is impressive.
a possible gotcha with IIR is the required word length - some biquad accumulators/state variables required resolution scales as the square of the filter fc to fsample ratio - may need to keep more than 32 bit – I had to go with 48 bits to prevent rounding errors becoming visible at 16 bit lsb in a 10 Hz filter with only 16k sample rate in a measurement app
Did anyone ever measure the distortion reduction on MC cartridges when the terminating load value is reduced? Besides distortion reduction, the signal level decreased but a pre-pre brings it back up.
A 4-transistor pre-pre with compl/push-pull topology using gnfb to the input of a grounded base stage is an excellent way to get low noise and low input Z and lower overall distortion level from the cart.
Thx-RNMarsh
A 4-transistor pre-pre with compl/push-pull topology using gnfb to the input of a grounded base stage is an excellent way to get low noise and low input Z and lower overall distortion level from the cart.
Thx-RNMarsh
Last edited:
Why would the distortion go down? I would think pulling current from the coil would cause it to go up. You may get an apparent reduction if the inductance is enough to cause a rolloff but correcting for that would bring the distortion back up.
Separately on the moving magnet/moving coil transducer energy performance, it just occured to me that the low frequency resonance would change if the loading affected the mechanicals. This would be measurable pretty directly and conventionally I would think. I don't have a working phono setup right now so I can't test but I'm sure someone here could.
The effect was reported by Peter Moncrieff in International Audio Review (IAR) mag when he lived in Berkeley, CA. He had some equipment that he used to look at waveforms, did waveform capture and compared... trying various loads on MM and MC to see affects and listened to them under various loads. Otherwise i was not part of his work.
I did design the pre-pre as described and it worked so well I published it... along with some low Z power cable ideas i used and such. Lawrence Livermore Lab released it so i could pursue a patent on it I wanted to. Instead I just published it... dont know if it was patentable or not but thiier patent lawyers though it might but had no application for them and so they wouldnt pay for the patent. Doesnt matter... no one else was doing such things and still arent. But now I dont own any analog record playing equipment... but this forum reminded me... otherwise i forgot about it.
I am wondering if anyone has looked into it in all the intervening years. Sounded great and quiet. John Curl talked to me about it and made suggestions but I never pursued his areas of improvements to it. I had a day job after all.
Thx-RNMarsh
I did design the pre-pre as described and it worked so well I published it... along with some low Z power cable ideas i used and such. Lawrence Livermore Lab released it so i could pursue a patent on it I wanted to. Instead I just published it... dont know if it was patentable or not but thiier patent lawyers though it might but had no application for them and so they wouldnt pay for the patent. Doesnt matter... no one else was doing such things and still arent. But now I dont own any analog record playing equipment... but this forum reminded me... otherwise i forgot about it.
I am wondering if anyone has looked into it in all the intervening years. Sounded great and quiet. John Curl talked to me about it and made suggestions but I never pursued his areas of improvements to it. I had a day job after all.
Thx-RNMarsh
Last edited:
What do you mean down to earth? Something more accessible than a TEK DSA8300? There are many possible tools. And more, what do you want to know? In a sense looking at I2S signals to determine performance is like looking at the torque on a driveshaft to get 0-60 times. Its possible but very indirect. Its a good diagnostic tool but what you want to know is on the analog output.
I've got you covered ;-)
Linear Audio Vol 6 will have an article by Erno Borbely and Sigurd Rushkowski using a complementary grounded gate JFET phono input stage.
jan
Hi Demian
Yes. GHz DSOs are not allowed to enter in my dreams. With my 20Mhz analog scope I can see the SPDIF eye pattern a bit hazy and that’s it.
I wish that there is one up to ~ $200
Ha ha. Correct.
The components in my recent set-up are I2S interconnected. There are many things that I may can do better and this is what I want to look at.
You may be right.
Yesterday I found this splendid paper written by the late Julian Dunn
http://www.electron.frba.utn.edu.ar/~jcecconi/Bibliografia/13%20-%20Medicion%20de%20Amplificadores/Documentos/AudioPrecision_AN5_DigitalAudioMeasurement.pdf
That’s a good start for me. Some of the tests are implementable with what I have (the scope, signal generator, soundcard).
George
Attachments
Thanks jneutron for these.
Am I going too far if I think of speaker/cartridge analogies, (scaled-down, weighted and inverted)?
speaker motor/cart motor,
speaker suspension and cone/cart cantilever assy,
speaker cone-air interface/cart needle-vinyl interface,
speaker chassis-box interface/cart motor–cart body interface,
speaker box-stand interface/cart-arm interface.
George
You may wonder why would I want to do that.
The reason is that there is a lot of published, practical, systematic and detailed research on speakers, done using modern tools (Klippel for example), something that has not be done on cartridges .
To the extend that analogies hold water, one can be helped by that.
George
The computational overhead is so low any CPU can do it in 64bit floats in real time. Apparently even several very long FIR's (Brute FIR) can be streamed real time. SoX does the RIAA at 192K easily.
EDIT - The SHARK processor in the miniDSP is 32 bit floating point which should pose no problem.
Last edited:
I wish someone would send me a link to these magical test LP's. The original RCA series was not up it nor were any others I've looked at, 1% is pusihng it. Identical results on all carts that I measured Monster Alpha 2 no different than the $15 Grado. No VTA tweeking etc. did much to help.
Will it retain interchannel timing at the 2 uSec level?
I face the exact same dilemma. They won't persue a patent unless they envision huge return. We attempted to patent something we developed back in '96, no go. Now, it appears that others are getting close to understanding it, so there goes that edge..
Thanks jneutron for these.
Am I going too far if I think of speaker/cartridge analogies, (scaled-down, weighted and inverted)?
speaker motor/cart motor,
speaker suspension and cone/cart cantilever assy,
speaker cone-air interface/cart needle-vinyl interface,
speaker chassis-box interface/cart motor–cart body interface,
speaker box-stand interface/cart-arm interface.
George
You may wonder why would I want to do that.
The reason is that there is a lot of published, practical, systematic and detailed research on speakers, done using modern tools (Klippel for example), something that has not be done on cartridges .
To the extend that analogies hold water, one can be helped by that.
George
No, I do not wonder, and no, you are not going too far.
For any iron device, there will always be the lenz/eddy concerns. For carts, the wire guage virtually eliminates wire proximity effects, but still ya got the iron lam thickness. Though, big iron is either solid producing eddies and lenz, or lammed at 1mm or .5mm to drop them.
So far, I've never seen modern tools (models) which accurately portray what actually happens in the gap. The correct model has to include lenz inductance reduction and eddy dissipation, and both have to include velocity dependent components to include the vc movement while carrying currents not correlated to the velocity but impressing field into a solid pole piece. Effects which are not quite as prominent in a cart.
jn
Hi John,
The choice of when and where to use ICs or discretes is always an interesting and important one in phono preamp design. It often reflects the philosophy of the designer and the type of engineering choices to be made. Such is the case with the VinylTrak phono preamp.
I find that phono preamp design involves some of the most fascinating analog engineering challenges. Indeed, such challenges often appear at the electro-mechanical interfaces in audio systems.
I chose a two-stage core phono preamp for both the MC and MM designs in the VinylTrak. The first stage, where the initial amplification and 75us equalization occurs, was chosen to be discrete JFET. I believe that a discrete first stage provides the best performance for a phono preamp. I prefer JFET inputs for both MM and MC applications partly because they provide the best EMI resistance and least cartridge interaction (the latter mainly an issue for the MM). The discrete input stages also allowed me to operate the first stage without negative feedback, with a true balanced input (high Z on both sides), and with the ability to incorporate the 75us time constant directly within the amplifier.
JFET inputs allow me to direct couple to the cartridge and avoid the need for a high-quality, and potentially large, input coupling capacitor.
JFET op amps generally don't achieve as low a noise as I like to obtain. They are of course pretty much out of the question for an MC front end.
I believe that BJT op amps do not generally deliver optimal performance for MM preamps, but that is just my personal design philosophy.
We do pay a price for using JFETs in the MC front-end. In order to achieve 0.7 nV/rt Hz noise with long-tailed pairs, four Linear Systems LSK389 dual JFETs had to be paralleled in a way that was stable DC and AC-wise and did not require device matching. These are not inexpensive JFETs. A discrete BJT MC front end can achieve lower input voltage noise, but I don't think that obsessing over noise below 0.7 nV/rt Hz is the best engineering tradeoff, especially if it is at the expense of EMI resistance in a circuit where such small signal levels are present.
Going discrete for the rest of the preamp would result in a level of complexity that I deemed unnecessary given the quality of op amps available. So I made the engineering tradeoff of using both discrete and IC approaches where each seemed to bring the most to the table.
Cheers,
Bob
My skin is thick enough to only feel some mild irritation here, but I still want to set it straight. The reason I jumped in this discussion is that I felt uneasy with the way SY attacked JC for a lack of attention to detail for using the 797 in a MM preamp. So that is apparently the way it goes.
I took opamp current noise into account when explaining that the effect of a Rs of 1KHz, although perhaps a bit on the high side for the 797, only comes with a small noise penalty, relatively speaking. On my internal slide rule I came in an earlier post to a bit over 10 % of the thermal noise of the cartridge. I looked it up with a calculator in my hand today and came to the same conclusion: 4nV/rtHz thermal noise for the cartridge and in.Rs=2E-12.1E3=2nV/rtHz for current noise. Sqrt(20) is a bit over 10% more than sqrt(16). In other words, if there are other good reasons to use this particular opamp, for example the way it behaves in an actual circuit, I don't see how you can call it a lack of attention to detail if you were to contemplate it for this use.
Now, I am always willing to learn, that's why I am presently going through Art Kay's book on opamp noise, and so perhaps I overlook something here that also escaped Kay. I hope you can perhaps give some further explanation of "The noise produced by current flowing through the MM Z is not excess noise, it is just another noise source. 2uA has about .8pA/rt-Hz". What is the nature of that "another noise source", I wonder.
Last edited:
Phase conformity is equally good but how do you get out of phase in this application? The streaming is standard of course there is latency but the timing is out of the filters hands. SoX works like a UNIX pipeline but DOS on a PC. One thing I have not tried is a fancy USB device, I don't have one that will do full duplex at 24/96 or 24/192. I suppse one of those PCI ones would work. You want, record || filter || play.
You are right the op-amp is not appropriate, please don't be anymore irritated than I am when jn points out that copper does not superconduct.
I am merely pointing out the standard definition of excess noise. When you put a (noiseless) battery on a resistor any current noise above sqrt(4KT/R) is excess noise. This is using the Norton equivalent of the resistor. This noise is virtualy always 1/f so it appears at the low frequencies. The bias current of the op-amp has noise sqrt(2qI) which is just another noise source in the circuit. The bias current does not make enough DC voltage on the coil to cause what is called excess noise. Maybe the article does not mention that excess noise has a specific physical meaning in the EE sense.
I was considering the difference between a z=1 FIR to analog. Zero crossing for localization control with no reconstruction math seems an issue, I was wondering how accurate timing would be.
jn
- Status
- Not open for further replies.