FRECUENCY MEASUREMENT@1GHz

[davicente]

phase matrix/eip 585c

Hi all,

I have to measure frecuency inside a 1Ghz pulsed signal, about 500ns wide. I think it is too narrow to try a traditional PLL, but maybe mixing with a DDS until i obtain a continious DC level (ripple should be specified depending on resolution:10Khz) and then reading their registers to obtain the frecuency
Any other ideas, papers, other examples????

Thanks in advance

DAVID

 

[biff44]

HP 5361B Pulsed/CW Microwave Counter

Phase Matrix / EIP 585C/588C

 

[davicente]

I cann't use an external meter, i have to develop a little module to be integrated into a bigger system.
So I was thinking in mixing or mix down and digitalize in order to use fft , but i haven't seen in the web any aplication or paper or somethinbg similar

thanks again

DAVID

 

[biff44]

What accuracy do you need in the meaurement? Does the pulse repeat with the same frequency? What Duty cycle?

 

[davicente]

accuracy:better than 10Khz
DC: less than 2%

DAVID

 

[biff44]

I guess you could downconvert it to around 20 MHz and then use a frequency discriminator circuit.

If you try to do a frequency discriminator at microwave frequency, you will need so high a Q that the amplitude will not be settled by the end of the 500 nS event.

 

[echo47]

FRECUENCY MEASUREMENT(at)1GHz

Sounds like you have 500 cycles of 1 GHz and want to measure its frequency to within 10ppm. Ouch! That's only a couple degrees of phase shift across the pulse.

Do you have to do the measurement within one pulse, or can you use multiple pulses? If multiple, then what's the pulse rate, and do all the pulses use the same 1 GHz source?

 

[davicente]

Hi again
All pulses have the same frecuency inside near 1Ghz which I have to measure
There could be only three pulses cycles every 2.5-5ms

thanks

 

[sergio mariotti]

Not easy. Are you sure that it is possible?

The simplest way, a counter circuit, give a resolution of 1e9*500e-9=0.002 parts that means the resolution is 0.002*1e9=2 MHz , too large for you.

You may downconvert to lower freq then perform a period measurement, but you'll be slave of the converter stability.

 

[echo47]

FRECUENCY MEASUREMENT(at)1GHz

Those 2.5-5ms pulses are too far apart to be useful.

Here's a maybe-possible method:
- Mix the 1 GHz input down to 200 MHz.
- Multiply the 200 MHz up to 1 GHz (use RF multiplier diode, not a PLL)
- Mix it down to 200 MHz
- Multiply it up to 1 GHz
- Mix it down to 200 MHz
- Multiply it up to 1 GHz
- Mix it down to 200 MHz

If your 1 GHz input shifts 10 kHz, then the 200 MHz output will shift 1250 kHz. You can easily measure that with a digital counter in an ASIC or FPGA.

Any jitter in your local oscillator will be exaggerated and pollute the results.

You can choose different intermediate frequencies to take advantage of your availabe parts.

 

[davicente]

Hi all again,

what do you think about the next idea:
What i am thinking is to make the DDs to sweep its frecuency in order to mix it with the fix frecuency I have to measure.In theory, when both frequncies are pretty the same, I will obtain from the mixer something near a DC voltage, when this happens, DDS should stop sweeping and a I can read DDs frequency register.
Pulse width will be 350ns as minimum and 60us as maximum
with this architecyute i will not need to digitalize.
i need to measure the frequency, not just measure 10KHz shift

SERGIO:
I cannot understand the 2MHz limit, could you explain the ecuation, please?
Thank you very much again

I hope this will hekp you to understand what i have in mind

Thanks

DAVID

 

[echo47]

FRECUENCY MEASUREMENT(at)1GHz

The problem is "something near a DC voltage". If the pulse width is 500ns and you are 10kHz off, the mixer output will still look like "something near a DC voltage".

 

[davicente]

Hi echo47,

I supoose i could calculate the ripple and frequency of the mixer output in order to be able to know that th accuracy is better than 10 khz. Although i still don't know how to do it, is it suposed to be so difficult??i've never done something like a frequency counter in 1 GHz.

thanks

DAVID

 

[artem]

FRECUENCY MEASUREMENT(at)1GHz

May be , you can downmix to 20 MHz , digitize with ADC at 1 GHz, add subsamples with extrapolation to achive needed accuracy and perform autocorrelation to find max .

 

[davicente]

Hi Artem,
any paper explaining autocorrelation methods to calculate frequency
I remember using autocorrelation methods with pulsed signals, but i can't remember if it was to calculate frequency

thanks

DAVID

 

[artem]

FRECUENCY MEASUREMENT(at)1GHz

Unfortunately I do not have a book for that , i even dont know its relevance for reliable frequency measurement)). But logically it seems OK