Rectify 500MHz Small Signal

how is it normally done? input squared with muliplier and then a squareroot ? muliplier with input and input amplified+limited?

-Lasse

On 25 sep, 17:13, Jim Thompson :

If you need accuracy you may use a large bandwidth AOP for this. I guess that a diode would introduce more than 1 mV error here.

Can you fab a 6AL5 on that process? ;-)

Tim

Only 40dB dynamic range, that's easy. 500MHz is going to be the not so easy part.

It's been almost 20 years since I did this (in hardware though) and I don't know the X-Fab process, but have you thought about successive detectors similar to what you'd find in a log amp chip? Basically a bunch of gain stages that each cover a small sliver of the dynamic range and then saturate, with the grand total being the summed outputs.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

What about an inverter and adder with nullifier for the negative halves? You could use zero-crossing detection to nullify the appropriate signal. This would be equivalent to inverting the signal during a zero crossing.

I'm tending toward AC-coupling plus a DC restorer... to "rectify" ;-) ...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

                           LOSE the WUSS
                          BRING BACK BUSH

Can't use some diode path and servo that against another diode path?

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Make two identical paths, each with a diode, that lets you measure the DC current from the diode to ground. Some fixed amount of forward voltage bias (less than "a diode drop") may be helpful.

Drive diode A with your RF, and measure its DC current.

Drive diode B with a DC voltage such that its current equals the current in diode A.

Measure that there DC voltage. There will be some error because of crest factor, but not (I suspect) enough to worry about.

I suspect that this would work just peachy with a pair of unmatched

1N4148s; do it all with a process that lets you match the diodes reasonably well and you should be sitting pretty.

.---. | | === | |\\ GND '--|+\\ ___ | >-------o------o o------|

You are going to use silicon 1N4148s at 500 MHz. with a 50 mV floor?

Jim

Do you also have a 499MHz signal handy?

Do you want an analog voltage out or are you after a digital answer?

If you jiggle the bias of a diode junction with the 500MHz, it will conduct more.

If you jiggle the bias of a matched junction with a lower frequency and adjust the amplitude until it causes the same increase in conduction, you will know that the two waveforms are of equal amplitude.

The adjustable amplitude waveform can be a multiplying DAC that you increase and decrease the number into it until the currents on the junctions match.

Barrie Gilbert has had some ideas. His AD834

could be used to square your 500MHz signal, which automatically rectifies it, though it leaves you with a lot of 1GHz ripple.

If you wanted to do something cuter, you could phase-lock a nominally

1GHz voltage-controlled logic level oscillator to the 500MHz signal, and use it to generate two 500MHz signals mutually in quadrature - nominally square waves, but at least with well-defined amplitudes, and form in-phase and quadrature products.

In a second order phase-locked loop, the quadrature output is integrated to control the VCO such that the quadrature output is precisely in quadrature with the incoming signal while the in-phase output can be used to drive another product detector (aka multiplier) whch gives you your rectified output.

-- Bill Sloman, Nijmegen

The 1mV accuracy requirement rules out the solutions with matched diodes, not mentioning the unmatched ones.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

How about a s&h, a 7 bit ADC, peak hold register and a 6 bit DAC?

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
                     "If it doesn\'t fit, use a bigger hammer!"
--------------------------------------------------------------

Of course, see my WVB receiver (on my website, SED page), dated 1974.

Just ducky when you have ample headroom (±5V supplies). I have a single supply, minimum operating at +2.7V

And Gilbert cells aren't all that accurate without lots of voltage an on-chip trimming... I need accuracy at small signals.

Dream on ;-) ...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

                           LOSE the WUSS
                          BRING BACK BUSH

Everybody is doing this sort of thing these days:

John

I like the electrical substitution idea already suggested (two diodes, drive one with DC to null out the signal from the other). How about ping-ponging a couple of them (say 75% duty cycle each), and measuring the offsets in between? You could measure the delta gain while they're both on (25% of the time) and the offsets when one or the other is off. With 75% duty cycle, they wouldn't ever be off together.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

There's the slideback technique: drive a comparator with RF on one side, DC feedback on the other. Tease the DC appropriately.

I once made a slideback sampling oscilloscope, using tunnel diodes, as my EE senior project. I won an award and had to attend a dreadful IEEE chapter banquet and repeat it to a bunch of old-fart power engineers who didn't understand a word I said. I described the slideback sampling scope in this ng some years back and a certain party loved the idea so much he later decided that he'd invented it himself.

HP used to sell a random sampling voltmeter.

John

I thought about AD log detectors also, and Joerg suggested something similar. However 1mV wrt 500mV = ~ 0.02dB; that's hardly feasible.If such accuracy is really required, a solution could be a FET downmixer.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

I'm sort of "ping-ponging" (diode current) to get chip (and external) temperatures. At 500MHz capacitance screws up rectification accuracy at low levels (BiCMOS process). ...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

                           LOSE the WUSS
                          BRING BACK BUSH