There is always some circulation current in the ground connections unless everything is battery powered. The best you can hope for is to reduce it to a minimal value. The best way to understand the problem is to draw a circuit of the system and represent each transformer filter etc as a cap or network of caps between the line and neutral to the chassis. Even 100 pF will have an influence since the source voltage is 120V RMS (in the US and higher in EU) and microamps of current will cause noise in a ground link.
The typical line filter will have .01 caps between line and neutral to the chassis. The leakage current will go somewhere. If two chassis have filters and the ground connection isn't real (not connected at the wall socket) everything floats at 1/2 line voltage. If the ground is OK then the currents circulate between chassis over the powerline ground, and the signal ground if there is a connection between them. Medical grade filters have much lower leakage, but not zero. Ultra isolation transformers have much less leakage but cost a bunch and have poor regulation.
Leakage through a turntable and a tonearm can me much higher than one might think, disconnecting the ground wire will illustrate the issue.
The typical line filter will have .01 caps between line and neutral to the chassis. The leakage current will go somewhere. If two chassis have filters and the ground connection isn't real (not connected at the wall socket) everything floats at 1/2 line voltage. If the ground is OK then the currents circulate between chassis over the powerline ground, and the signal ground if there is a connection between them. Medical grade filters have much lower leakage, but not zero. Ultra isolation transformers have much less leakage but cost a bunch and have poor regulation.
Leakage through a turntable and a tonearm can me much higher than one might think, disconnecting the ground wire will illustrate the issue.
1audio said:
I tried to explain this to a couple of "smart" guys (MIT Phd's no less) once. They were convinced the virtual ground type of input for MC provided useful mechanical damping of the cantilever/stylus system, even at the tip resonance.
One of our fab guys did some spook work on laser interferometers for the military and yes you probably could actually measure it if you wanted to and had $$$$ to waste.
Years ago, when moving coul cartridges were the only real option for good sound I had an intern (A PHD student at Standford) do the arithmetic on a cartridge, energy in at the stylus to the energy out of the coil. The transfer was really low. It was clear that nothing at the wire end would impact the stylus.
However the wires are in a nonlinear magnetic circuit. Perhaps, drawing current from them is interacting with the magnetic circuit and introducing some nonlinearity.
However the wires are in a nonlinear magnetic circuit. Perhaps, drawing current from them is interacting with the magnetic circuit and introducing some nonlinearity.
1audio said:
They thought that like a ribbon mic they could fix the high end resonance by electro-mechanical damping. I thought at the time in the case of the mic it is the primary resonance and it is in this case tightly coupled. The tip/cantilever resonance OTOH is secondary and very poorly coupled to the motor, so a double loss.
A quick thought experiment; Put a cartridge down on an LP and drive one channel with a step and look for excitation of the primary tone arm resonance in the other channel. With a lock-in amp this is probably more sensitive than the laser.
syn08 said:
another point is that we're not talking about Hi-Perm transformer style distortion with small DC current - the MC pole piece iron is deep into magnetic saturation due to the permenant magnet bias field so the hysteresis loops should be severely "squashed" compared to the 0 magnetic bias curves
I would guess Eddy current damping would be the largest effect of any DC current in the MC cart coils short of so much current you get heating effects
scott wurcer said:
Finally a use for the lock in amp in storage. However the ambient vibration might swamp out the effect, and the electrical isolation between channels in a cartridge is not very high. On a typical cartridge both channels are wound on the same piece of iron. They are orthogonal but not that perfect.
1audio said:
Thats why I would use a step, you could probably gate away the transient and look for the ~10Hz tonearm/cart resonance. So if you put in a 1Hz step you would look for the 10Hz component modulated by 1Hz. You are right though the noise at 10hz probably has sidebands that obscure the result. Also you would want to use non-harmonically related excitation and resonance.
I have a great book on gravity wave sensing, this is starting to remind me of that.
Hps 4.0
Unless I get a better idea, I think I've decided about the HPS 4.0 bipolar input stage. A few key parameters (the results of breadboard measurements, matching pretyy well the simulations).
- 0.3nV/rtHz noise. With 2SC2240/2SA970, noise is 0.7nV/rtHz, not bad at all, still pretty good for a MC cartridge (it's similar to HPS 2.0).
- under 0.001% distortions at 20KHz and 100mV input (4V output).
- 2MHz ULG and 85 degrees phase margin.
- Input impedance is 4Mohm up to 1KHz, then dropping at 1Mohm @20KHz. Good enough, even for a MM cartridge with 0.5H inductance (which is anyway beyond the scope of this project).
A low noise input stage/follower, biased by a constant current source (mostly for increasing the input impedance). A low noise cascoded gain stage, then the same folded cascode and high output current opamp, feeding the feedback network. The emitter ballast resistors can be omotted, although I decided to keep them to compensate for potential devices little mismatchings (although the Hitachi pair, being ion implanted, are very well matched out of the tube). A total equivalent of 0.25ohm doesn't really matter.
Here's the schematic:
What were the other options?
a) A discrete current feedback opamp architecture. While this option had some nice features (increased input impedance, better thermal stability, Baker clamp, ability to run the output stage in class A providing abyssimal low distortions) I considered it to complicated for very little improvement. The noise performance is about the same as above. Here's a schematic draft (it misses the servo and the frequency compensation).
b) A derivative of Edmond's CMCL input stage. Working fine, but I found little reason why CMCL would help here. Distortions are not really an issue neither is the "fighting VAS" effect. Again to complicated for very little gain. I have also found pretty difficult (although possible) to frequency compensate this circuit. Here's a schematic draft (missing the frequency compensation).
c) A floating input stage. I have tried prof. Leach input stage and prof. Hawksford idea, published in Wireless World. They both work fine (having pros and cons though) but require battery supply, or some sort of floating power supply. Unfortunately this is where I failed miserably, I was unable to create a simple floating low noise power supply. While the supply itself is not a big deal, unless it's fed from a separate mains transformer, the capacitive couplings between this supply and the others on the same transformer are cancelling any "floating design" at high frequencies. Such a configuration would require at least four mains transformers for a dual mono construction, which is to me a to hefty price to pay for zero cartridge current (and that at DC only). Of course, and R-core could be the answer, but I don't have any and I was unable to source some, at a decent price.
I also spent some time modelling the 2SC2547/2SA1085 pair, based on the datasheet and my own measurements. The models below are pretty good, although far from perfect.
.model Q2SC2547 NPN(
+ Is=1.6014e-013
+ Xti=3
+ Eg=1.11
+ Vaf=143.7
+ Bf=420
+ Ise=7.7335e-013
+ Ne=1.6093
+ Ikf=0.024257
+ Nk=0.1249
+ Xtb=1.5
+ Var=100
+ Isc=555.1p
+ Nc=1.796
+ Ikr=5.85
+ Rb=2.5
+ Rc=.1232
+ Cjc=1.1398e-011
+ Mjc=0.40659
+ Vjc=.35
+ Fc=.5
+ Cje=7p
+ Mje=.3333
+ Vje=.75
+ Tr=10n
+ Tf=32.8n
+ Itf=1
+ Xtf=0
+ Vtf=10)
.model Q2SA1085 PNP(
+ Is=1.0651e-013
+ Xti=3
+ Eg=1.11
+ Vaf=85.5
+ Bf=330
+ Ise=2.3301e-013
+ Ne=1.782
+ Ikf=0.074989
+ Nk=.9631
+ Xtb=1.5
+ Var=100
+ Isc=555.1p
+ Nc=1.796
+ Ikr=0.01
+ Rb=2.2
+ Rc=.2032
+ Cjc=2.295e-11
+ Mjc=0.62422
+ Vjc=.3905
+ Fc=.5
+ Cje=9p
+ Mje=.3333
+ Vje=.75
+ Tr=10n
+ Tf=89.5n
+ Itf=1
+ Xtf=0
+ Vtf=10)
Unless I get a better idea, I think I've decided about the HPS 4.0 bipolar input stage. A few key parameters (the results of breadboard measurements, matching pretyy well the simulations).
- 0.3nV/rtHz noise. With 2SC2240/2SA970, noise is 0.7nV/rtHz, not bad at all, still pretty good for a MC cartridge (it's similar to HPS 2.0).
- under 0.001% distortions at 20KHz and 100mV input (4V output).
- 2MHz ULG and 85 degrees phase margin.
- Input impedance is 4Mohm up to 1KHz, then dropping at 1Mohm @20KHz. Good enough, even for a MM cartridge with 0.5H inductance (which is anyway beyond the scope of this project).
A low noise input stage/follower, biased by a constant current source (mostly for increasing the input impedance). A low noise cascoded gain stage, then the same folded cascode and high output current opamp, feeding the feedback network. The emitter ballast resistors can be omotted, although I decided to keep them to compensate for potential devices little mismatchings (although the Hitachi pair, being ion implanted, are very well matched out of the tube). A total equivalent of 0.25ohm doesn't really matter.
Here's the schematic:
What were the other options?
a) A discrete current feedback opamp architecture. While this option had some nice features (increased input impedance, better thermal stability, Baker clamp, ability to run the output stage in class A providing abyssimal low distortions) I considered it to complicated for very little improvement. The noise performance is about the same as above. Here's a schematic draft (it misses the servo and the frequency compensation).
b) A derivative of Edmond's CMCL input stage. Working fine, but I found little reason why CMCL would help here. Distortions are not really an issue neither is the "fighting VAS" effect. Again to complicated for very little gain. I have also found pretty difficult (although possible) to frequency compensate this circuit. Here's a schematic draft (missing the frequency compensation).
c) A floating input stage. I have tried prof. Leach input stage and prof. Hawksford idea, published in Wireless World. They both work fine (having pros and cons though) but require battery supply, or some sort of floating power supply. Unfortunately this is where I failed miserably, I was unable to create a simple floating low noise power supply. While the supply itself is not a big deal, unless it's fed from a separate mains transformer, the capacitive couplings between this supply and the others on the same transformer are cancelling any "floating design" at high frequencies. Such a configuration would require at least four mains transformers for a dual mono construction, which is to me a to hefty price to pay for zero cartridge current (and that at DC only). Of course, and R-core could be the answer, but I don't have any and I was unable to source some, at a decent price.
I also spent some time modelling the 2SC2547/2SA1085 pair, based on the datasheet and my own measurements. The models below are pretty good, although far from perfect.
.model Q2SC2547 NPN(
+ Is=1.6014e-013
+ Xti=3
+ Eg=1.11
+ Vaf=143.7
+ Bf=420
+ Ise=7.7335e-013
+ Ne=1.6093
+ Ikf=0.024257
+ Nk=0.1249
+ Xtb=1.5
+ Var=100
+ Isc=555.1p
+ Nc=1.796
+ Ikr=5.85
+ Rb=2.5
+ Rc=.1232
+ Cjc=1.1398e-011
+ Mjc=0.40659
+ Vjc=.35
+ Fc=.5
+ Cje=7p
+ Mje=.3333
+ Vje=.75
+ Tr=10n
+ Tf=32.8n
+ Itf=1
+ Xtf=0
+ Vtf=10)
.model Q2SA1085 PNP(
+ Is=1.0651e-013
+ Xti=3
+ Eg=1.11
+ Vaf=85.5
+ Bf=330
+ Ise=2.3301e-013
+ Ne=1.782
+ Ikf=0.074989
+ Nk=.9631
+ Xtb=1.5
+ Var=100
+ Isc=555.1p
+ Nc=1.796
+ Ikr=0.01
+ Rb=2.2
+ Rc=.2032
+ Cjc=2.295e-11
+ Mjc=0.62422
+ Vjc=.3905
+ Fc=.5
+ Cje=9p
+ Mje=.3333
+ Vje=.75
+ Tr=10n
+ Tf=89.5n
+ Itf=1
+ Xtf=0
+ Vtf=10)
Ovidiu -
excellent work!
Thinking about 4 transformers I see no problems with that when the gain is to have zero DC current through an expensive MC cartridge. The cost for 4 transformers is relatively low!
What DC currents do you suspect are going through the cartridge for the different designs?
BTW,
thanks for the two SPICE models!
SIgurd
excellent work!
Thinking about 4 transformers I see no problems with that when the gain is to have zero DC current through an expensive MC cartridge. The cost for 4 transformers is relatively low!
What DC currents do you suspect are going through the cartridge for the different designs?
BTW,
thanks for the two SPICE models!
SIgurd
Sigurd Ruschkow said:
Certainly under 1uA - this is what I measured on the breadboard. Unfortunately I don't have very low current measurements capabilities, so I can't be for the moment more precise.
I'm still looking for references/discussions about, meantime what I got on this forum convinced me it's not that critical. I also can't find any test data on commercially available MC stages, nobody (including Stereophile) seem to care about such. I can't imagine all high end MC stages are build with low noise JFETs.
Unfortunately, it is thermodinamically impossible to cancel the base currents, so that the input current is null, without impacting the noise. This can be rigurously proved based on the semiconductor transport mechanism (Boltzmann transport, etc...) but before going there, imagine a circuit that would cancel the input bias current. Sensing the base currents (and balancing them) or sensing the input offset (due to the input bias current) would insert at the input either a current noise source or a voltage noise source. Unfortunately, uncorrelated noise sources never cancel, but always add, so the input current cancelling circuit will actually add to the noise... Of course, a manual one time adjusting of the input bias current is always possible, by manually balancing the input stage bias collector currents.
So one way or another, a bipolar MC low noise amp has to live with some input current...
syn08 said:
You don't need much.
Here is what I use to accurately measure / test tubes for grid current.
A precision, jfet input opamp is required here. The caps are just low value polypropylene; required to prevent oscillation (LF pole caused by the large resistors and the opamp input capacitace).
The output voltage is equal to -20mV/nA
Attachments
Your not measuring femtoamps? I thought tubes had really low grid bias current. Oops, sorry, thats electrometer tubes.
Jfets have a low bias current but not zero, and it increases quickly with temperature. I would suggest a chopper amp for that application. The MAX4238 bias current for example is 1 pA so the error will be very low.
Jfets have a low bias current but not zero, and it increases quickly with temperature. I would suggest a chopper amp for that application. The MAX4238 bias current for example is 1 pA so the error will be very low.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.