They way I tested AD797 (and other opamps) yesterday was quite similar or probably more stressing. Gain 40dB and linear load of 1kohm.
How did you like the National Semiconductor opamp you have used to replace the AD797? BTW, in my measurements the AD797 has lower noise than 4562 and is more resistant to HF interference.
Ohh? I thought we had agreed here that measurements say nothing about sound quality? Are you now saying that the rank of 7th determines the rank of sound quality, even if it is below the -110dB audibility threshold John gave?
To paraphrase: dazed and confused ...
jan
CCIF2 19+20kHz measurement of opamps
Today I have prepared a 2nd set of opamp measurements. Now it is CCIF2, 19+20kHz. Only difference tone (1kHz), i.e. 2nd harmonic component, is calculated in 0.000X% results. But we have the FFT spectra.
Test conditions: Gain = +101 (40dB), Vout = 2.9Vpp, loaded with feedback network 1kohm + 10 ohm. The schematics is in the attached file. I have also added the measurement limit - card loopback measurement. It is another card than yesterday, for it performs better in CCIF2 test.
One can see that AD797, AD844 and OPA637 have had excellent results in this CCIF2 test. OPA134 has higher 3rd order products - the distortion lines close to 19kHz and 20kHz spectral lines.
Regarding THD1kHz and CCIF2 measurements, the winner is AD797. But, as correctly described by Richard, this opamp needs much more care regarding PCB and decoupling layout and is more sensitive to RF interference than JFET input opamps and than AD844 CFA opamp.
Today I have prepared a 2nd set of opamp measurements. Now it is CCIF2, 19+20kHz. Only difference tone (1kHz), i.e. 2nd harmonic component, is calculated in 0.000X% results. But we have the FFT spectra.
Test conditions: Gain = +101 (40dB), Vout = 2.9Vpp, loaded with feedback network 1kohm + 10 ohm. The schematics is in the attached file. I have also added the measurement limit - card loopback measurement. It is another card than yesterday, for it performs better in CCIF2 test.
One can see that AD797, AD844 and OPA637 have had excellent results in this CCIF2 test. OPA134 has higher 3rd order products - the distortion lines close to 19kHz and 20kHz spectral lines.
Regarding THD1kHz and CCIF2 measurements, the winner is AD797. But, as correctly described by Richard, this opamp needs much more care regarding PCB and decoupling layout and is more sensitive to RF interference than JFET input opamps and than AD844 CFA opamp.
You are perfectly right, Jan. My setup erases influence of CM distortion and non-linear capacitances.
1) Resistors in -IN and +IN are only in tens of ohms. Too low to see non-linear capacitance effect.
2) Gain +101 and output voltage slightly above 1Vrms results in 10mVrms input voltage only. Again, influence of CM distortion is suppressed.
I wanted to measure noise (SNR) - thus low resistance values.
If I measured at gain 1 - 10, the result of good opamps would be below my distortion and noise measurement threshold limit.
So it goes, nothing is perfect.
1) Resistors in -IN and +IN are only in tens of ohms. Too low to see non-linear capacitance effect.
2) Gain +101 and output voltage slightly above 1Vrms results in 10mVrms input voltage only. Again, influence of CM distortion is suppressed.
I wanted to measure noise (SNR) - thus low resistance values.
If I measured at gain 1 - 10, the result of good opamps would be below my distortion and noise measurement threshold limit.
So it goes, nothing is perfect.
http://www.google.com/url?sa=t&rct=...QlnxKNwN49oLehmBETsY7dw&bvm=bv.41248874,d.dmg
Pay specific attention to the textual content within the first paragraph. (or, you could work the equations, your choice. Conceptually, eq 4 and 5 say it all, and note the text immediately following eq 11. Note also that the closed form solution for the resistor scaling ratio now has a form which includes lambda to the fourth power, never negative.
For a tubular resistor, the film conductivity will weigh heavily..high resistivity material will have a lower J2, whereas low resistivity will have a more significant j2.
Note that this is bog standard stuff for high speed switchmode supply coil design. High current slew rates directly impact conductor dissipation even if the switchmode supply is constant frequency. High current slews also tend to blow up 3 and 4 inch diameter hockey puck scr's if the external circuit allows too high a dI/dt, even with spiralling gate finger geometries. It blows them up underneath the gate wire pad.
Your first figure is electrostatic potential, not magnetic field density. Static B field in a cylindrical conductor is exactly straight line, zero at center max at surface. Once time varying is introduced, the straight line will be depressed at center, reinforced at surface, and certainly not straight..
Oh, while I'm at it. The current density of the core of a coax goes to the surface, we call that skin effect, think of it as the current redistributing to lower inductance, that being the 15 nH per foot inductance consistent with a cylindrical conductor. The current density of the shield goes to the inner surface...that is NOT skin effect, it is proximity effect as a consequence of the magfield being generated by the core wire. When asymmetrical core to shield occurs, this proximity effect will force the shield current to redistribute such that magfield outside the shield goes back to zero, and the current takes the path of lowest inductance (reactance).
jn
ps..excellent..I wasn't sure if the link would retain integrity. Let me know if the link doesn't work..here's the title..
SERBIAN JOURNAL OF ELECTRICAL ENGINEERING
Vol. 7, No. 1, May 2010, 13-20
13
A Closed Form Solution for the Proximity
Effect in a Thin Tubular Conductor
Influenced by a Parallel Filament
Dragan Filipović1, Tatijana Dlabač2
Last edited:
Chris, don't let them 'shake you up' for your personal convictions. What works, works.
Measurements can often predict that something will work well, BUT NOT ALWAYS.
For example, as an input stage device, I had tested a third device with similar specs to the other two. It DID NOT sound as good as the two others, so it was put aside. WHY? I don't know, and I don't have to go further to find out, because I have found 'successful' IC chips that sound darn good in the application with a gain of 40.
Measurements can often predict that something will work well, BUT NOT ALWAYS.
For example, as an input stage device, I had tested a third device with similar specs to the other two. It DID NOT sound as good as the two others, so it was put aside. WHY? I don't know, and I don't have to go further to find out, because I have found 'successful' IC chips that sound darn good in the application with a gain of 40.
Thanks for finding that. It confirms what I have said all along: that proximity effect depends on frequency, not magnitude. Eq (11) shows that the distribution of currents depends on material properties, geometry, and frequency. Current does not appear.jneutron said:
To clear up one possible confusion: the Fourier series they use is a series in theta (angle around the conductor) not frequency. This of course fits nicely with the original pupose of Fourier theory: solving differential equations in heat conduction. No harmonic generation is implied!!
Another possible confusion (it temporarily confused me!): lambda is not wavelength, but a function of geometry, material properties and frequency given by eq (2).
I would like to thank jn for presenting evidence which disproves his theory. Many people would have quietly buried it.
Actually, that is incorrect. I really expected you to see it, but am not suprised.. Lambda has current embedded, although the authors do not state it..sigma.
The final form for J2 they show as integrated, so they've already assumed a steady state current equivalent. Try looking at the sheet currents on the cylinder with no transport current..just the eddy currents.
Certainly a point worth mentioning. I can see others being confused by it.
And sigma. That being current density.
While I always work that way, in this case you missed the current magnitude the authors casually forgot to mention embedded in that weird lambda. That is why I mentioned the fourth power of lambda, as current is in there.
jn
Last edited:
- Status
- Not open for further replies.