The inductance rolls off at low frequency - and so the current must increase during big disc warps. This makes the B field increase, to the point where Warp + signal approaches saturation of the trafo core; this is sure to increase distortion.
If the warp signal is large enough, you reach the nonlinear part of the B-H curve. The (eg mumetal) core in MC trafos is sure to saturate at a low level - so it will be like having some dc current in the trafo.
I don't think Walt talked about it, at least not as far as I know.
The Hofer patent is 4614914 (check at PAT2PDF - Free PDF copies of patents: Download and print! or I can send it).
Lundahl has a one-sheet app note about the same:
http://www.lundahl.se/pdfs/papers/feedbck.pdf
OK, now it's your turn to send me something
jd
Transformer non-linearity at low levels
Menno van der Veen has an interesting article on "Low level audio signal transfer through transformers conflicts with permeability behavior inside their cores" on his WEB page. Menno van der Veen, audio electronic research & consultancy . He is looking at output transformers but this may still be relevant for input transformers.
Menno van der Veen has an interesting article on "Low level audio signal transfer through transformers conflicts with permeability behavior inside their cores" on his WEB page. Menno van der Veen, audio electronic research & consultancy . He is looking at output transformers but this may still be relevant for input transformers.
According to Sy's phono article:
"As an interesting exercise, we will compute the signal to noise of the cartridge itself, just to see how quiet the preamp needs to be in order that it does not add any significant noise. The cartridge resistance is 15R. Plugging that value into equation 1, we find that the cartridge has 72nV of thermal noise at room temperature over the usual 20Hz-20kHz audio bandwidth. The cartridge’s nominal output is 0.2mV. Doing a quick decibel check, signal to noise works out to -69dB."
Sounds like you've already decided that pV level signals are audible over noise.
we should also be careful about mixing/misapplying material state dependent effects when searching for explanations - hard vs soft magnetic material - analog tape um particles vs annealed NiFe alloy have very different behaviors
my understanding is that quality transformer core material has large oriented crystals (or none for amorphous) specifically to minimize the "quantization" of fully saturating a crystal - and that ferromagnetic domain wall movement within a crystal is very low energy - and as Sy alludes the motion may be dithered by thermal noise power at room temperature
we shouldn't make the mistake of thinking of "The hysteresis curve" or that it shows a "deadband" that necessarily affects low level signals - we usually see plotted a high peak excitation cycle that does show saturation effects
but at any starting (low) magnetization point there is a incremental u_r and local "hysteresis loop" for a small (far from saturation level) signal
and the initial u_r may be very high for low excitation in (su)permalloy tape cores and the local loop is vanishingly "narrow" for low signal excitation
material selection is clearly critical, from the Menno van der Veen paper: (starting with Grain Oriented Silicon Steel showing low intial permability)
“The results of this comparison are clear and totally as
expected; higher perm cores have enough perm at
threshold levels to create almost no weakening between
20Hz and 1kHz.”
He also mentions the possibility of analog magnetic tape style dithering of the core with a inaudible bias frequency
and returning to record cutting: a 80 kHz bais was added to the cutter drive with Teldec's Direct Metal Mastering – although their justification was that the bias frequency “polished” the groove
I think Hawksford's noise paper misses the Ohmic connection from the external signal source into the transistor base region - the conduction band electrons are often modeled as a "gas" with a temperature dependent "pressure" and net current is drift velocity - not exactly quantized integer number of electrons per second
this model fully allows for averaged "fractional charge" to create a continuous - i.e. not observably quantized - electric field modulation in the transistor base region by the source voltage – due to thermal “Brownian motion” in the “electron gas”
he also just "pulls out of his hat" a number for the quantization "sampling time"
my understanding is that quality transformer core material has large oriented crystals (or none for amorphous) specifically to minimize the "quantization" of fully saturating a crystal - and that ferromagnetic domain wall movement within a crystal is very low energy - and as Sy alludes the motion may be dithered by thermal noise power at room temperature
we shouldn't make the mistake of thinking of "The hysteresis curve" or that it shows a "deadband" that necessarily affects low level signals - we usually see plotted a high peak excitation cycle that does show saturation effects
but at any starting (low) magnetization point there is a incremental u_r and local "hysteresis loop" for a small (far from saturation level) signal
and the initial u_r may be very high for low excitation in (su)permalloy tape cores and the local loop is vanishingly "narrow" for low signal excitation
material selection is clearly critical, from the Menno van der Veen paper: (starting with Grain Oriented Silicon Steel showing low intial permability)
“The results of this comparison are clear and totally as
expected; higher perm cores have enough perm at
threshold levels to create almost no weakening between
20Hz and 1kHz.”
He also mentions the possibility of analog magnetic tape style dithering of the core with a inaudible bias frequency
and returning to record cutting: a 80 kHz bais was added to the cutter drive with Teldec's Direct Metal Mastering – although their justification was that the bias frequency “polished” the groove
I think Hawksford's noise paper misses the Ohmic connection from the external signal source into the transistor base region - the conduction band electrons are often modeled as a "gas" with a temperature dependent "pressure" and net current is drift velocity - not exactly quantized integer number of electrons per second
this model fully allows for averaged "fractional charge" to create a continuous - i.e. not observably quantized - electric field modulation in the transistor base region by the source voltage – due to thermal “Brownian motion” in the “electron gas”
he also just "pulls out of his hat" a number for the quantization "sampling time"
Last edited:
I don't think Walt talked about it, at least not as far as I know.
The Hofer patent is 4614914 (check at PAT2PDF - Free PDF copies of patents: Download and print! or I can send it).
Lundahl has a one-sheet app note about the same:
http://www.lundahl.se/pdfs/papers/feedbck.pdf
OK, now it's your turn to send me something
jd
I'm about to release a product based on this (or the Lundahl variant) idea. The patent has expired. It works but you either compromise (like we did) or make a very special transformer with a dead winding since its very sensitive to the dcr of the winding and that is very temperature sensitive. It works but I'm not sure how much benefit it brings yet.
There is another input transformer variant on the Lundahl site using a similar idea, but it only works with low source Z and some drive capability.
Last edited:
I'll have to look harder at the "Zero Field" idea
the Lundahl low frequency distortion cancellation technique employs "mixed feedback" to synthesize a negative resistance that needs to fairly precisely cancel all of the primary/source loop real resistance - for a mc cart and stepup xmfr this would include canceling the (much larger) cart resistance - and not meeting the recommended cart load impedance, if it could be done at all without a noise penalty (I'd look for Nordholt and "Electronic Cooling")
this is mentioned in Walt's latest "Op Amp Applications" book in the audio xmfr line driver section - available online at analog.com
another approach is using a "current transformer" with virtual gnd op amp on the secondary shorting the winding and keeping the V (and hence dB/dt) low
you can't set cartridge load impedance - although mc cart may work well into virtual gnd anyway
the Lundahl low frequency distortion cancellation technique employs "mixed feedback" to synthesize a negative resistance that needs to fairly precisely cancel all of the primary/source loop real resistance - for a mc cart and stepup xmfr this would include canceling the (much larger) cart resistance - and not meeting the recommended cart load impedance, if it could be done at all without a noise penalty (I'd look for Nordholt and "Electronic Cooling")
this is mentioned in Walt's latest "Op Amp Applications" book in the audio xmfr line driver section - available online at analog.com
another approach is using a "current transformer" with virtual gnd op amp on the secondary shorting the winding and keeping the V (and hence dB/dt) low
you can't set cartridge load impedance - although mc cart may work well into virtual gnd anyway
Last edited:
Zero field -is- a current transformer and should be ideally, fed from a high
impedance. The mixed feedback just makes it a more ideal virtual ground
at the IP and hence more linear. It can be used with or without mixed
feedback. Both are not suitable for MC input.
I believe Rupert Neve was one of the first to commercially implement a ZF
IP product, but it was a line level product and most likely had 10k or so IP
resistors.
T
My experience with the distortion cancellation has been that it does reduce distortion below 30 Hz, some, depending on the adjustment which is very sensitive. That's why it's so temperature sensitive. It has no effect on mid-band distortion. The distortion products meet the John Curl criteria with almost all being low order harmonics.
On the input side you have a tradeoff between shielding and coupling not to mention all the other aspects. Plus, John doesn't want to use feedback so tricks like the Zero Field would be out for his project. However I'm sure others are less doctrinaire on this subject and would be very interested in exploring zero field and other tricks for reducing the effects of the input transformer.
On the input side you have a tradeoff between shielding and coupling not to mention all the other aspects. Plus, John doesn't want to use feedback so tricks like the Zero Field would be out for his project. However I'm sure others are less doctrinaire on this subject and would be very interested in exploring zero field and other tricks for reducing the effects of the input transformer.
AP has the compensation resistor somewhere inside the winding pack, so that it tracks with winding temperature.
jd
This raises an interesting issue that often gets confused. The nominal output is often rated at -20db from clipping at 1000 cycles/sec (hz.). Comparing this to the full bandwidth noise is misleading as would be comparing it to the bandwidth of the test signal (around 1hz allowing for mechanical errors). Using a critical bandwidth around 1000hz would be a more appropriate comparison, but still not exact.
As a real music signal contains more than a single tone there would also be another way to calculate the signal to noise (s/n) ratio based on that.
Using +15db as a modest estimate of the ears filtering and 20db as the headroom then the signal to noise ratio would be 104db.
But that is an estimate IN MY OPINION.
This is an issue of much debate. The classic a weighting of the noise floor was called into question by Dolby, pointing out that hearing is more sensitive to noise at 3 KHz. They promoted and adopted along with THX the CCIR-ARM method. AES E-Library: CCIR/ARM: A Practical Noise Measurement Method.
However for our purposes two issues need to be resolved. First- max signal. For a digital system that's a pretty easy and obvious answer, for analog recording its always been one needing a judgement call. 1% THD or 3% thd (or 10% as in digital amps?). On the other side is it noise as above or do you account for the dithering effect of noise and reach as much as 10 dB into the noise? And noise level relates to measured bandwidth. 20-20K will have much higher level than 400 to 3000 (A weighting) on white noise. Further with vinyl and with eq etc. the noise is shaped so how do you account for the shaping. This was covered to some degree in a national app note AN-104 http://www.national.com/an/AN/AN-104.pdf 35 years ago before spreadsheets were commonly available.
The number mentioned above is more of a dynamic range number than a s/n number I think.
However for our purposes two issues need to be resolved. First- max signal. For a digital system that's a pretty easy and obvious answer, for analog recording its always been one needing a judgement call. 1% THD or 3% thd (or 10% as in digital amps?). On the other side is it noise as above or do you account for the dithering effect of noise and reach as much as 10 dB into the noise? And noise level relates to measured bandwidth. 20-20K will have much higher level than 400 to 3000 (A weighting) on white noise. Further with vinyl and with eq etc. the noise is shaped so how do you account for the shaping. This was covered to some degree in a national app note AN-104 http://www.national.com/an/AN/AN-104.pdf 35 years ago before spreadsheets were commonly available.
The number mentioned above is more of a dynamic range number than a s/n number I think.
- Status
- Not open for further replies.