I am looking for guidance on how best to measure the low frequency amplitude stability of a 300kHz sinusoid. The 300kHz signal is generated by a Wien-Bridge oscillator. It is being used as a signal source for a sensor whose output is eventually demodulated with a diode detector. The bandwidth of the system is DC to 100Hz. I would like to be able to measure the low frequency changes in amplitude up to 100Hz to a level of 10ppm or better, relative to the oscillator amplitude.
Measuring Amplitude Stability
Started by ●April 28, 2011
Reply by ●April 28, 20112011-04-28
smc123 wrote:> I am looking for guidance on how best to measure the low frequency > amplitude stability of a 300kHz sinusoid. The 300kHz signal is generated > by a Wien-Bridge oscillator. It is being used as a signal source for a > sensor whose output is eventually demodulated with a diode detector. The > bandwidth of the system is DC to 100Hz. I would like to be able to measure > the low frequency changes in amplitude up to 100Hz to a level of 10ppm or > better, relative to the oscillator amplitude.Doing all of this in analog sounds like very naive idea. Neither the oscillator nor the diode detector could be any close to 10ppm accuracy. The best accuracy that you could expect from your configuration is going to be at the order of 1% or so. If 10ppm of accuracy is required, the signal generation and the amplitude measurement should be done in digital; using reasonably good DACs and ADCs. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●April 28, 20112011-04-28
On Apr 28, 9:31�am, "smc123" <xmacleod@n_o_s_p_a_m.yahoo.com> wrote:> I am looking for guidance on how best to measure the low frequency > amplitude stability of a 300kHz sinusoid. �The 300kHz signal is generated > by a Wien-Bridge oscillator. �It is being used as a signal source for a > sensor whose output is eventually demodulated with a diode detector. �The > bandwidth of the system is DC to 100Hz. �I would like to be able to measure > the low frequency changes in amplitude up to 100Hz to a level of 10ppm or > better, relative to the oscillator amplitude. �Ouch! They used to make crystal ovens that mounted in an octal socket. Maybe if your diode were in one of those, its threshold would be stable enough to make your measurement possible. Of course, since you want the level relative to the oscillator, you have to measure that too. Both diodes in the same oven might cause crosstalk, and separate ovens will cycle differently. (A total error of 10ppm means that each must be stable to 5ppm if the drifts are independent.) How is the oscillator stabilized? The classical way uses a tungsten resistor (light bulb) to control the feedback gain. (That was Packard's and Hewlett's patent that started a company.) Might a table- driven DAC be better? I'd need to pore over to spec sheets to know. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●April 28, 20112011-04-28
Thank you very much for your suggestions. Here is some more info on the system .... The Wien-Bridge oscillator is configured in a control loop, with the amplitude set by a low noise, filtered voltage reference (ADR441). The oscillator output amplitude is diode detected, filtered, and fedback to the error amplifier. The circuitry is ovenized to within 1degC. Also, for clarification, the 10ppm is a peak to peak noise requirement. The accuracy is not so much important because the system gets calibrated. As of now, I'm getting about 20ppm of variation through the system, with the bw limited to about 10Hz. So, trying to determine how much of the total noise is being contributed by the signal source. Thanks
Reply by ●April 28, 20112011-04-28
smc123 wrote:> Thank you very much for your suggestions. Here is some more info on the > system .... > The Wien-Bridge oscillator is configured in a control loop, with the > amplitude set by a low noise, filtered voltage reference (ADR441). The > oscillator output amplitude is diode detected, filtered, and fedback to the > error amplifier. The circuitry is ovenized to within 1degC. Also, for > clarification, the 10ppm is a peak to peak noise requirement. The accuracy > is not so much important because the system gets calibrated. As of now, > I'm getting about 20ppm of variation through the system, with the bw > limited to about 10Hz. So, trying to determine how much of the total noise > is being contributed by the signal source.The standard way of measurement of the precision oscillator quality (i.e. amplitude and phase noise) is by correlation it against itself with some timing lag. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●April 28, 20112011-04-28
> > Thank you very much for your suggestions. �Here is some more info on the > > system .... > > The Wien-Bridge oscillator is configured in a control loop, with the > > amplitude set by a low noise, filtered voltage reference (ADR441). �The > > oscillator output amplitude is diode detected, filtered, and fedback to the > > error amplifier. �The circuitry is ovenized to within 1degC. Also, for > > clarification, the 10ppm is a peak to peak noise requirement. �The accuracy > > is not so much important because the system gets calibrated. �As of now, > > I'm getting about 20ppm of variation through the system, with the bw > > limited to about 10Hz. So, trying to determine how much of the total noise > > is being contributed by the signal source. >another way to look at your problem is you want to measure incidental AM modulation at about 0.001%. The sideband levels on an AM carrier at that modulation depth would be about -100 dBc (if I did my math right). If the incidental modulation were not random but were repetative at oh say 60Hz, you MIGHT be able to see a -100 dBc sideband with a good spectrum analyzer with a narrow res BW. But I'm not sure how you will know if it is AM or PM (I'm not talking about what time it is :-). But all this brings up the real question...if you can hardley measure it, why do you need to do this? Mark
Reply by ●April 28, 20112011-04-28
On 04/28/2011 06:31 AM, smc123 wrote:> I am looking for guidance on how best to measure the low frequency > amplitude stability of a 300kHz sinusoid. The 300kHz signal is generated > by a Wien-Bridge oscillator. It is being used as a signal source for a > sensor whose output is eventually demodulated with a diode detector. The > bandwidth of the system is DC to 100Hz. I would like to be able to measure > the low frequency changes in amplitude up to 100Hz to a level of 10ppm or > better, relative to the oscillator amplitude.I'm with Jerry and Vladimir in questioning why you are generating your signal with a Wien bridge oscillator. Any attempt you make to measure the signal digitally is going to be complicated by needing to know the (ever wandering) phase of the oscillator. I think you'd do a _much_ better job if you generated your reference tone digitally then (if necessary) filtered it to a sine wave. You'd probably still have to measure the reference for amplitude and phase, but at least the degree of wandering of both of these would be significantly lessened. As to measuring the amplitude of a sine wave itself: sample it and convert it to digital at above the Nyquist rate (taking the bandwidth of any modulation added by the sensor into account), then in the digital world quadrature demodulate it. Average the demodulated signals over some integer number of cycles, and get the amplitude (and phase) of the resulting complex number. If you really need 10ppm accuracy then you need a really, really good ADC. If you just need 10ppm _resolution_, with some allowable amount of nonlinearities, then a plain old fast 16-bit (or even 14-bit) ADC may be sufficient. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●April 28, 20112011-04-28
Jerry Avins <jya@ieee.org> wrote: (snip)> How is the oscillator stabilized? The classical way uses a tungsten > resistor (light bulb) to control the feedback gain. (That was > Packard's and Hewlett's patent that started a company.) Might a table- > driven DAC be better? I'd need to pore over to spec sheets to know.I always knew there first product was an audio oscillator, but I didn't know what made it special. -- glen
Reply by ●April 28, 20112011-04-28
On Apr 28, 2:35�pm, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:> Jerry Avins <j...@ieee.org> wrote: > > (snip) > > > How is the oscillator stabilized? The classical way uses a tungsten > > resistor (light bulb) to control the feedback gain. (That was > > Packard's and Hewlett's patent that started a company.) Might a table- > > driven DAC be better? I'd need to pore over to spec sheets to know. > > I always knew there first product was an audio oscillator, but I > didn't know what made it special.They were a garage operation in the right place at the right time. Their first deliverables were to provide sound effects for the movie "Fantasia". Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●April 29, 20112011-04-29
Thank you for the suggestions. I've got an ADI 18 bit converter Eval board on the way. It may not have all the resolution that I'm looking for, but hopefully, it will provide some insight. I'll try Tim's suggestion of demodulating and low pass filtering, deriving a reference signal from a portion of the data.
