Can't-you bypass the problem by making clipping occurs *before* the input stage or at the input point where you take your input signal to compare ?
May-be you will be amused to look at that ?
http://www.diyaudio.com/forums/solid-state/221737-ultimate-amp-protection-circuit-4.html
Mr Marsh, Error correction works in comparing two signals. I am afraid there is noway to use that kind of thing on an oscillator, where there is nothing to compare.
I am afraid that is no simple solution. When clipping occurs, the error level become enormous. May-be detecting its peak level in normal condition and make-it clip in the comparator if it exceeds this level ? But i do not believe-in it: it seems obvious, looking at your waves, that the OPA itself takes more and more time to recover...
Last edited:
Thanks Richard - unfortunately no, as Esperado says. I think the best approach for that is to use a notch filter.
BTW, I still regularly refer to your capacitor articles - they have certainly stood th test of time!
I recall making a phono preamp using the LF156 that way, and subbing the inputs with a LM194.
I don't know. I could ask the rep..
Interesting. But as I recall, many here including you, have been lamenting old parts which are discontinued. I linked them so that you guys could decide if they were worth anything.
jn
Let us again discuss the significant differences between different families of complementary jfet parts:
1. REALLY OLD PARTS (long gate) >40 years old: Can be fairly quiet, however the extremely low Gm limits mid-range noise to maybe 3nV/rt Hz.
2. OLD PARTS (short gate) >30 years old <40 years old: Higher Gm gives promise of lower midrange noise, some devices can achieve 0.7nV/rt Hz in the midrange.
1/F noise is still dominant and can completely destroy the noise characteristic without selection. Parts have always been available from somebody, but some manufacturers discontinued the parts. Some parts have been made noisy by newer manufacturing methods.
3. JAPAN PARTS (meshed gate) <35 years old till recently: Best overall performance, VERY HIGH Gm, VERY LOW 1/F, a designer's dream. UNTIL discontinued by major Japanese companies. NO REAL REPLACEMENT, until LIS gets their act together, and then at a higher price than previously available.
1. REALLY OLD PARTS (long gate) >40 years old: Can be fairly quiet, however the extremely low Gm limits mid-range noise to maybe 3nV/rt Hz.
2. OLD PARTS (short gate) >30 years old <40 years old: Higher Gm gives promise of lower midrange noise, some devices can achieve 0.7nV/rt Hz in the midrange.
1/F noise is still dominant and can completely destroy the noise characteristic without selection. Parts have always been available from somebody, but some manufacturers discontinued the parts. Some parts have been made noisy by newer manufacturing methods.
3. JAPAN PARTS (meshed gate) <35 years old till recently: Best overall performance, VERY HIGH Gm, VERY LOW 1/F, a designer's dream. UNTIL discontinued by major Japanese companies. NO REAL REPLACEMENT, until LIS gets their act together, and then at a higher price than previously available.
Thanks Richard - unfortunately no, as Esperado says. I think the best approach for that is to use a notch filter.
After the notch filter, you have the residual... the distortion products/harmonics which can then be used to cancel them in the original waveform. Im sure it would take a whole new approach to impliment. But, then, super low thd from the source wouldnt be needed.
I suspect the ShibaSoku THD instruments might do something similar to get a THD range of .0003% full scale. Putting it out here in case others think of a way to use the isolated harmonics to cancel them in the source. Thx-RNmarsh
Hi slope pass band around the frequency, followed by trap filters for the nearest frequencies (F2, F3) ?
Still short channel and its called impact ionization (dramatic gate current increase after 5-8V). LIS has it too. Interesting question is that P-channel FET's are immune to this and still N -channels were prefered. It is well known that in general PNP's can have lower rbb all else being equal. I figure it's a right vs. left hand thing, schematics just look right with N everything.
Half serious here.
JC, can you list some examples?
especially the "0.7nV/rtHz in the midrange"
The big Toshiba parts were the 2SJ72 and 2SK147. The gate C of the -72 was 185pF, and the e sub n curves in my 1989 databook show about 650pV/sqrtHz at 10mA. By the 1992 databook these parts had disappeared.
Damn! My interest is for condensor mike use. BF862 still looks best for this in certain circuits like Guru Wurcer's Linear Audio stuff.
The 862 is fine except for the rather high current, given one of Scott's constraints being battery operation. But whether even it is significantly better overall, given the other noise sources like Brownian motion of air and thermomechanical noise of the structure of the capsule is questionable.
Which reminds me: I wonder if Wave ever tested those photonic mic capsules?
- Status
- Not open for further replies.