Is there a high-precision way to convert voltage to frequency?

[RandallMcRee]

Just curious. I have an accurate frequency counter--can it be used to monitor an LTZ1000, for example using a precise V/F converter?

The best circuit I know of is the one below. It claims 0.005% precision. That's nice. But then there is that pesky LT1009 in there with its 15-25ppm TC....and there is an LF356 which hardly seems a precision opamp--but perhaps that is telling me that one is not needed?!

So, first of all, is this totally bonkers, or has someone thought about this? If not totally bonkers is there a circuit known to be a good starting point? If its the LTC1043 one can it be tweaked to perform as well as a typical 6.5 digit multimeter?

As I say, just curious. 

[Vgkid]

I believe Solartron used them in their higher end dmm's.
Edit , it was a voltage to time. It created a pulse train, that was scaled to the input.
I will post all the related pages either later tonight/tomm.
See Solartron 7065/55 manual.

[thermistor-guy]

V-to-F circuit with claimed 15 ppm linearity: http://www.linear.com/solutions/1325

[Echo88]

V -> f 7ppm linearity: http://cds.linear.com/docs/en/application-note/an14f.pdf Page 7

[T3sl4co1l]

Yes.  The above are good examples.  The ultimate is the Josephson junction, a tunneling superconducting structure which oscillates proportional to voltage.  It is exact, to the extent that quantum mechanics allows.  (Primary voltage standards use this.)

Curiously (or perhaps not), the inverse operation (F to V) is slow, and usually imprecise.  You can easily pair a precision V-to-F with a PLL to get the same precision, but you have the tracking speed and settling time of the PLL to deal with, so it's not fast.

Tim

[TiN]

tunneling superconducting structure which oscillates proportional to voltage

.
Isn't that the other way around, RF field in the cryojunctions cause voltage steps?

[T3sl4co1l]

It works both ways, absorption or emission.  As far as I know...

Tim

[RandallMcRee]

Thanks,
I seem to have missed that app note...for those who are following along here is the best so far; Jim says: "This V?F has a resolution of 1ppm, with linearity inside
7ppm (0.0007%). When combined with a processor-driven gain/zero calibration loop it has negligible zero and gain drift. "

[JohnPi]

Those circuits can be very linear (and Jim Williams' circuits are very good), but they have offset, and scale errors.

Fundamentally they depend on an R and a C to give the frequency (time) reference. So for x ppm scale precision, you need R's and C's also x ppm accurate. For offset, there are parasitic capacitors which create errors, as well as the offset of the input opamp itself.

[branadic]

Hi,

has someone ever build the Ultra-Linear V→F Converter from AN14? What about noise, real world numbers and graphs on linearity and real world resolution?
While all of the analog gurus passed away will we stuck on further improvements?

I wonder if someone has thought about a voltage to frequency difference converter, based on a frequency reference such as OCXO or GPSDO. I think about "something" with voltage input and two frequency outputs. When input voltage is zero both outputs deliver the same frequency, if a voltage is applied to the input one outputs increases frequency, while the other decreases frequency. This difference can be measured with a time interval counter. Since they are available with 10 or 12 digits high resolution can be obtained. What do you think?

vco, voltage controlled delay line, voltage controlled phase shifter, voltage to time converter

-branadic-

[David Hess]

Has someone ever build the Ultra-Linear V→F Converter from AN14? What about noise, real world numbers and graphs on linearity and real world resolution?

The given specification is linearity and not accuracy or precision.  Offset and gain errors can be calibrated out as shown in figure 10 of AN14.

Because a voltage-to-frequency converter integrates its input, noise is controlled by integration time.  I suppose noise could be significant if a reciprocal counter was used but all of the ones I have seen produced less than 1 count of noise so there was not much to measure.  If the count varied more than 1 count, then something was broken.

While all of the analog gurus passed away will we stuck on further improvements?

There is not much demand now for precision voltage-to-frequency converters because instrumentation delta-sigma converters have gotten so good.

If you are looking for something to evaluate drift of LTZ1000 references, then using a voltage-to-frequency converter is not going to produce a better result than the voltage-to-frequency converter's own voltage reference.

I wonder if someone has thought about a voltage to frequency difference converter, based on a frequency reference such as OCXO or GPSDO. I think about "something" with voltage input and two frequency outputs. When input voltage is zero both outputs deliver the same frequency, if a voltage is applied to the input one outputs increases frequency, while the other decreases frequency. This difference can be measured with a time interval counter. Since they are available with 10 or 12 digits high resolution can be obtained. What do you think?

Voltage-to-frequency converters which produced an A/B ratio output used to be very common up to at least 20,000 counts or 4-1/2 digits but typically were 10 or 12 bits.  They could be directly used with an A/B universal frequency counter.

[Wolfgang]

Just curious. I have an accurate frequency counter--can it be used to monitor an LTZ1000, for example using a precise V/F converter?

The best circuit I know of is the one below. It claims 0.005% precision. That's nice. But then there is that pesky LT1009 in there with its 15-25ppm TC....and there is an LF356 which hardly seems a precision opamp--but perhaps that is telling me that one is not needed?!

So, first of all, is this totally bonkers, or has someone thought about this? If not totally bonkers is there a circuit known to be a good starting point? If its the LTC1043 one can it be tweaked to perform as well as a typical 6.5 digit multimeter?

As I say, just curious. 

IIRC, there was a discussion in the old AoE about how the Keysight 6 1/2 and 7 1/2 digits precision multimeters worked. They used a multislope technique, IIRC.

[Kleinstein]

The circuit from AN14-7 is a good starting point. It is a kind of old type change balancing ADC. However it is by no means perfect. So the 7 ppm claimed accuracy is already quite good for such a circuit. It gets quite tricky if much better accuracy is wanted. It starts with a rather low reference voltage, so the OPs (LT1013, A1) noise would be relative to 1.2 V. The peak current type loading of the reference is also a possible weak point. The VCO used (build around 74C04) also does not look like really low noise.  There is also a dielectric absorption effect on the bucket cap (1 nF PS).
So already the claimed 7 ppm INL level are a low value. There is also gain drift from the capacitor value. So the LT1004 reference is about adequate for the circuit.

The V->F converter way was used in the very early DMMs - so more like 3-4 digits, mainly before the dual slope was introduced. About the only good thing is that signal isolation is easy and an already existing counter could be used.