Can someone help me with basic math please? I calculate that the highest frequency for this oscillator is 7,962Hz but they list the range as 1,500 to 15,000Hz. When the 2kohm pots are at zero, I used 200 ohms and 0.1uF and got this result. sump'in ain't right.
I calculate that I need 100 ohm resistors with 1kohm pots to get the range 1.5kHz to 15kHz. Is that right? I cheated and used this online calculator:
http://mycircuits9.blogspot.com/2012/06/wein-bridge-oscillator-calculator.html
I also need a recommendation on a stepped attenuator for the 2 gang 2k (1k?) pots.
I calculate that I need 100 ohm resistors with 1kohm pots to get the range 1.5kHz to 15kHz. Is that right? I cheated and used this online calculator:
http://mycircuits9.blogspot.com/2012/06/wein-bridge-oscillator-calculator.html
I also need a recommendation on a stepped attenuator for the 2 gang 2k (1k?) pots.
Attachments
Last edited:
It's f = 1 / 2(PI)RC. But you may have to account for some input resistance of the op amps. This will cause a deviation from the nominal. Stray capacitance will cause a shift as well.
I don't think you want an attenuator here just to switch in different resistor values for desired frequencies. You can use the above formula to find the resistor values by rearranging to... R = 1 / 2(PI)fC. Likewise C = 1/2(PI)fR.
You may need to trim the resistance to bring the frequency closer to what you want.
Most designers put a frequency vernier pot in somewhere but some believe this pot can cause additional distortion and for good reason, it can.
David.
Last edited:
OK, so you're suggesting that I make this a fixed frequency oscillator? I think a variable oscillator would be more useful which is why I'm pursuing this. It's really hard to get a dual gang pot to track exactly unless you use a stepped pot with precision resistors. I think that's where the distortion that you mentioned comes from.
thanks. I wonder why the screwed that up so badly? I guess the need an editor at Linear Technology to verify their application notes.
I should look up what Walt Jung has to say about oscillators. He seems to be a reliable source, among many of course.
That's not what I'm saying at all.
An attenuator is for attenuating. What I'm saying is to make up you own switched resistor network, like an attenuator, but customized it for the frequencies you desire.
A stepped attenuator won't work properly for frequency adjustment.
I think your intention is, is to replace the gang pot with a switched network.
David.
Last edited:
@dirkwright -- the output level can be adjusted (within limits) by altering the reference DC voltage for the AGC loop provided by the 5V zener diode -- for lower DC level and lower output level, an LED works very well as a zener -- a red LED is around 2V, green up toward 3V; white, a bit more. Less good, but lower DC references would be series 1N4148s -- say two of them, but their knees are not sharp.
Do not change the ratio of the 10k and 14k resistors -- the DC through the 14k corresponds in heating to the heating value of the oscillator signal through the 10k. Putting the two diodes in close proximity (VERY close -- I would wrap them together in some insulation) provides first-order temperature compensation for the signal level. These resistor values are chosen to keep currents through the diodes moderate but high enough to provide some heating. So if you're going to try for lower output, try just reducing them by 1/2 to start.
Some Wien bridge tuning resistor (or capacitor) mismatch doesn't generally lead to an increase in distortion, just a shift in frequency -- the effect on gain is too small; and most of the distortion comes from the amps and the AGC loop anyway.
My opinion is that this oscillator leaves a lot to be desired, especially compared to Victor's design.
RE editing of the Linear app notes, yeah that needs to be better -- note that in the app note this oscillator came from, figures 47 and 48 are reversed with the respect to the descriptions...
Do not change the ratio of the 10k and 14k resistors -- the DC through the 14k corresponds in heating to the heating value of the oscillator signal through the 10k. Putting the two diodes in close proximity (VERY close -- I would wrap them together in some insulation) provides first-order temperature compensation for the signal level. These resistor values are chosen to keep currents through the diodes moderate but high enough to provide some heating. So if you're going to try for lower output, try just reducing them by 1/2 to start.
Some Wien bridge tuning resistor (or capacitor) mismatch doesn't generally lead to an increase in distortion, just a shift in frequency -- the effect on gain is too small; and most of the distortion comes from the amps and the AGC loop anyway.
My opinion is that this oscillator leaves a lot to be desired, especially compared to Victor's design.
RE editing of the Linear app notes, yeah that needs to be better -- note that in the app note this oscillator came from, figures 47 and 48 are reversed with the respect to the descriptions...
Thanks for your very helpful comments.
Well, I don't even remotely understand Victor's design, so I'm at a loss to do anything with it. It doesn't appear to be a Wein bridge, and it's not a state variable type so I don't have a clue. I don't think I want a fixed frequency oscillator after all.
Looks like a Wien bridge to me, just configured a bit differently.
I'd agree with Richiem, Victor's design looks better.
I can never understand how a LDR + PI controller can work properly, seeing as the LDR is a PI control element in itself, or at least a leaky integrator. Everytime I've simulated one it's very difficult to get them to stabilise. By the time I got that AN43 design to work the distortion looked poor.
Does anybody have a spice model for the VTL5C10 ?
I only have Thomas Gootee's VTL5C2 model, the 5C10 is a low slope type.
I'd agree with Richiem, Victor's design looks better.
I can never understand how a LDR + PI controller can work properly, seeing as the LDR is a PI control element in itself, or at least a leaky integrator. Everytime I've simulated one it's very difficult to get them to stabilise. By the time I got that AN43 design to work the distortion looked poor.
Does anybody have a spice model for the VTL5C10 ?
I only have Thomas Gootee's VTL5C2 model, the 5C10 is a low slope type.
It's a Wien bridge -- series elements in the input leg and parallel elements in the feedback leg. This is however an all-inverting design, which eliminates a serious obstacle to low distortion in feedback oscillators -- common-mode errors. Jim Williams attempted an apparently successful fix for this in the oscillator from app note AN43.
Victor's circuit (I hope I looked at the right schematic) solves this in a different way by keeping the amps in inverting mode. Huge benefit. A state-variable design does this too, except for the summing stage, but not having any non-inverting amps in the oscillator proper really helps.
Victor's circuit (I hope I looked at the right schematic) solves this in a different way by keeping the amps in inverting mode. Huge benefit. A state-variable design does this too, except for the summing stage, but not having any non-inverting amps in the oscillator proper really helps.
I knew I'd seen the all inverter config before:
Low distortion oscillator tests measurement circuits | TMW
Low distortion oscillator tests measurement circuits | TMW
Has anyone tried using the Linear Tech RMS-DC converters in the detector/AGC loop? Williams wrote an application note (AN-106) and there was another little squib using the LTC1966 or LTC1968.
With respect to the VTL5C10, you can take the values off the data sheet and use Eurequa to determine the spice parameters.
With respect to the VTL5C10, you can take the values off the data sheet and use Eurequa to determine the spice parameters.
I tried the LT RMS-DC converter. I had to filter the output of the convertor so heavily it was only good for 1KHz down. Not suitable for a full range oscillator unless noise is not an issue.
The output looks like a full wave rectifier but with a wave form shape that averages RMS with an LP filter.
I discovered a trick with the 1968. Since the output is switched capacitor, two convertors can have their outputs parallel. One of the two convertors fed from the quadrature. The output wave form is symmetrically overlapped reducing the ripple buy about 80%.
David.
Last edited:
The 'little squib', was this it ? http://cds.linear.com/docs/LT Journal/05_03-12-LTC1968-Pei.pdf
Yes, I expected I'd have to modify or create a VTL5C10 model, I thought I'd ask here first.
OK, please be patient with my ignorance. I do not see the series and parallel elements in Victor's schematic. There are supposed to be two capacitors and two resistors of equal value. I can't even identify the input leg and the feedback leg. Are the resistors in the Wein bridge R17+R18 and R10+jfet? I don't get it.