Hi there:
I am a retired professional circuit designer. I have an K-H4400A and a K-H4402A oscillator sitting next to me. Both oscillators produce Sine and Cosine
signal outputs. But they both use a rather clumpsy peak detector for deriving the AGC signals. We know from High-school geometry that the
Square root of Sine squared plus cosine squared is = 1. Why not use that for AGC, is my question. I therefor plan to connect some 4 quadrant multipliers up as two squarers and one square router to produce the AGC needed to set the gain. That aught to be better than detecting a peak with some diodes. The diodes are most probably accurate as far as output voltage is concerned, The squarers and square router are probably not going to
be as peak-preside as the diodes, but tests wil show how much they are off. This modification will make the AGC much more trustworthy.
Then it is the Operational Amplifiers used, I agree they were the best available in the 1970-80 time but much better ones exist now. I used them myself then. After i get my improved AGC to work, i will tackle those problems. I know there are a couple of circuits in there that are guarantied to produce even order distortion products. More later when I know more.
Hans Jørgen Weedon.
I am a retired professional circuit designer. I have an K-H4400A and a K-H4402A oscillator sitting next to me. Both oscillators produce Sine and Cosine
signal outputs. But they both use a rather clumpsy peak detector for deriving the AGC signals. We know from High-school geometry that the
Square root of Sine squared plus cosine squared is = 1. Why not use that for AGC, is my question. I therefor plan to connect some 4 quadrant multipliers up as two squarers and one square router to produce the AGC needed to set the gain. That aught to be better than detecting a peak with some diodes. The diodes are most probably accurate as far as output voltage is concerned, The squarers and square router are probably not going to
be as peak-preside as the diodes, but tests wil show how much they are off. This modification will make the AGC much more trustworthy.
Then it is the Operational Amplifiers used, I agree they were the best available in the 1970-80 time but much better ones exist now. I used them myself then. After i get my improved AGC to work, i will tackle those problems. I know there are a couple of circuits in there that are guarantied to produce even order distortion products. More later when I know more.
Hans Jørgen Weedon.
Interesting approach. There are (were) TRMS detectors in chip form as well however they are not that fast so the settling time is long or you have instability. The best solutions I have seen are sample and hold peak detectors, involved but work well and a really interesting solution using an ADC to sample the wave. It would be possible to use the ADC into a small micro and a DAC to drive the AGC element or even create the correction signal that would be coming out of the AGC element into the feedback loop.
The existing solution on the 4400 works OK. On mine the settling time at lower frequencies can be pretty long. I am able to get 10V RMS with the increased rails and all the harmonics at below -120 dB. Its more than good enough for most tasks.
The existing solution on the 4400 works OK. On mine the settling time at lower frequencies can be pretty long. I am able to get 10V RMS with the increased rails and all the harmonics at below -120 dB. Its more than good enough for most tasks.
Hi I am still at it. I have for various reasons, primarily for temperature and accuracy reasons discarded the true RMS way. I do not think that the Barry Gilbert cell from 1968 was ever intended for high precision. I asked him after the IEEE talk when he presented the paper if Tectronix would ever use that method as a Wideband vertical amplifier. He asked me what I thought, and I answered probably not, because there was no precision negative feedback possible with all the natural group delays in the circuit. He was a little surprised and said "You are probably right how do you know that?"
Yes I know that detecting the peak seems to be good enough, but I have mentally gone through the RMS method and discarded it for conceptual reasons. It will produce harmonic components that will be difficult to filter out and at the same time have reasonable settling times. With a sharp filter the servo-loop will have too much delay and the loop will ring, which is not good. But today i figured out a method using peak integration timed from the zero-crossings of the quadrature phase. I have not fully analyzed that method, but I am working on it. The way I do it is to breadboard it in my mind and run a mental analysis of the circuit. Not many people I know can run something like a SPICE analysis in their mind. I believe I can do it because I learned to read electronic schematics before i learned to read English words. My Dad had a collection of Popular Electronics that I studied over and over again. I grew up in Norway, so I could not read the words in English. I remember I was very upset when my dad would not tell me what a µµF was. He Was an electrical engineer as well, but told me to figure that one out for myself. They did not use pF until later. This was probably 1947 or so. So i date myself. Just hang on, I have the parts on order. Hans Jørgen Weedon
Yes I know that detecting the peak seems to be good enough, but I have mentally gone through the RMS method and discarded it for conceptual reasons. It will produce harmonic components that will be difficult to filter out and at the same time have reasonable settling times. With a sharp filter the servo-loop will have too much delay and the loop will ring, which is not good. But today i figured out a method using peak integration timed from the zero-crossings of the quadrature phase. I have not fully analyzed that method, but I am working on it. The way I do it is to breadboard it in my mind and run a mental analysis of the circuit. Not many people I know can run something like a SPICE analysis in their mind. I believe I can do it because I learned to read electronic schematics before i learned to read English words. My Dad had a collection of Popular Electronics that I studied over and over again. I grew up in Norway, so I could not read the words in English. I remember I was very upset when my dad would not tell me what a µµF was. He Was an electrical engineer as well, but told me to figure that one out for myself. They did not use pF until later. This was probably 1947 or so. So i date myself. Just hang on, I have the parts on order. Hans Jørgen Weedon
I'm looking forward to seeing your idea. I hope it works. There was a lot of discussion about this issue in the low distortion oscillator thread https://www.diyaudio.com/community/threads/low-distortion-audio-range-oscillator.205304/post-2878236 some 1000 or so posts back. Its a huge collection of ideas and implementations but probably a "War and Peace" level slog to read through.
I'm rehabbing an early 4400 (1980 board). There are a few differences in the schematic, like the "Main Output Amp" area of the schematic uses one MPS6566 (NPN) and one '4392 jFET, ......and it is 600 ohm output only.
But that aside, the LME497010 is out of production now. Any alternatives ? OPA1611 perhaps ? (needs SOIC adapters).
Also, U103 was an LM310n, a true voltage follower, where inverting pin 2 is left open. Keep it open with the newer chip ?
THX
But that aside, the LME497010 is out of production now. Any alternatives ? OPA1611 perhaps ? (needs SOIC adapters).
Also, U103 was an LM310n, a true voltage follower, where inverting pin 2 is left open. Keep it open with the newer chip ?
THX