Although the light bulb exhibited extremely low distortion in the mid range up the mechanical sensitivity ruined the whole thing. You would have to find a way of floating the lamp to isolate it from shock.
The ADC + Mdac infinite hold peak detector is by far the easiest to implement and extremely linear. It does have some noise issues but this is easy to over come by maintaining a high SNR and healthy amount of decoupling. This requires a BP filter with an very high Q. The wiens might fair well with their relatively low Q. The gain at resonance is the key.
The down side of the ADC/Mdac is that it's expensive to do.
The ADC/Mdac could be hybrid with the existing peak detector/multiplier in the Shibasoku .
I used two ADC/Mdac in my oscillator. One for the peak detector and a second for the multiplier. The cost got so high that it became not very DIY friendly.
The ADC + Mdac infinite hold peak detector is by far the easiest to implement and extremely linear. It does have some noise issues but this is easy to over come by maintaining a high SNR and healthy amount of decoupling. This requires a BP filter with an very high Q. The wiens might fair well with their relatively low Q. The gain at resonance is the key.
The down side of the ADC/Mdac is that it's expensive to do.
The ADC/Mdac could be hybrid with the existing peak detector/multiplier in the Shibasoku .
I used two ADC/Mdac in my oscillator. One for the peak detector and a second for the multiplier. The cost got so high that it became not very DIY friendly.
Forr beat me to it but oscillator schematic is in the file: http://www.waynekirkwood.com/Images/pdf/Cyril_Bateman/Bateman_Notes_Cap_Sound_1.pdf
Last edited:
Oh, no! Good quality VCAs were a legacy of the music synthesizer chips made for only about ten years in the 1980s by SSM and CEM. The VCO and VCA chips were especially neat and what's left of them sell for high prices. I presume it was studio analog mixing boards that kept the VCA business going, but now even that is digital.
It looks like THAT still makes VCAs as well as some "Dynamics Processors" that might be applicable here:
THAT IC Selection Guide
For that matter, the digital gain controllers might be good for microprocessor control, though 1dB steps looks way too coarse for this app.
One off, pcb and parts not including power supply is about $600.00 Canadian dollars.
This is for a 2.5Vrms source 10Hz to 100kHz. Variably tuned and controlled from an app on a PC via USB HID. No driver required. Windows thinks it's like a custom mouse of sorts.
100% isolated from USB ground. Still don't know what the bottom line is for distortion.
What/where else can be bought for 600 Can Dollars that is similar.... new or used?
THx-Richard
There was some wailing and gnashing of teeth long ago when the SSM2018 went out of new production, but it now appears that THAT semi has some superior offereings.
After experimenting with light bulbs, FETs, CdS cells and thermistors as oscillator amplitude stabilizing devices (which all work in their ways) it always seemed to me that a high quality VCA would be the best solution.
I was always going to built a VCA-stabilized oscillator. Then I discovered computer audio interfaces. The first time I thought of this, the only relevant chips were 4 quadrant multipliers, and they may not have been much better than the others.
Glad to see that the approach could be followed with such outstanding results!
Hmm, given that the job at hand is measuring a really clean sine wave at just one voltage level...
Back in the late 60's (I was a teenager) I used a thermistor as an agc element in an oscillator I designed. DC current through the device controlled its temperature. It was kind of slow, however.
My first oscillator, an EICO 377, used a lamp.
When I began work at Bell Labs, I spent a lot of time looking at the schematics of HP test equipment in the service manuals. When I saw how they were using a JFET with the 1/2 Vds feedback trick I thought it was great! With some tweaks, that's what I've used ever since.
The key issue of importance in many of these agc circuits is the tradeoff of distortion versus amount of noise introduced into the oscillator by the agc circuit. This is easy to see, as in a JFET agc arrangement. You can run the JFET at as low a signal level as you want, getting the distortion way down, but you then need to amplify its agc contribution, and that increases the noise introduced into the oscillator. The JFET agc I used in my THD analyzer provides flexibility in this tradeoff. At the same time, the amount of "authority" (agc range) that the agc needs to be able to exercise is also in the mix. If the oscillator is precise and needs little agc range, then less agc signal needs to be used and the noise contribution goes down, all else remaining equal. All of this has to do with the agc element itself. Matters of the agc detector and injection of distortion via ripple is another matter (although it shares some of the tradeoffs).
One thing of interest, at least to me, is that while people focus a great deal on the distortion of their oscillator, there seems to be little attention to measuring or specifying the noise of the oscillator, part of which results from the agc circuit as mentioned above.
Just some food for thought.
Cheers,
Bob
In the Linear Tech design note they use the true RMS converter to drive the gate of the JFET.
The Tektronix SG5010 is inferior in this respect to the SG505!
When Jell Smith and I did our first sub-ppm oscillator using the SSM2018
rolleyes I had a side experiment using a precision TellLabs PTAT resistor, the idea being thermistors probably had poorer excess noise. I enhanced the heating by gluing a power resistor driven by a buffered output to it. Slow is an understatement, it took minutes to stabilize but the noise sidebands dissapeared.
It would be difficult, but they are almost all based on Dave Blackmer's DBX VCA which is still available from THAT. The single version here is pretty good and trimmable for very low distortion.
THAT Corporation 2181-series Trimmable Voltage Controlled Amplifier ICs
It's in the circuit of the Linear Tech super oscillator with the CFB super gain block -- i could get it (the oscillator) working correctly about half the time.