High speed signal rectification for accurate amplitude measurement.

[Odysseus]

I'm trying to design an analog circuit to make an accurate measurement of sine wave amplitude, lets say usable to at least 1MHz or more.  A high speed ADC would certainly do the trick, but that's overkill, since I only need to measure the (constant) amplitude of the signal.  Instead, I'm thinking I should perform AC-DC conversion followed by a precision low speed ADC, like one in a multimeter.

Since I know the signal is a sine wave of a single frequency, a true RMS converter unnecessary, and they typically have limited bandwidth anyway, so I'd rather implement ordinary , full-wave rectification.  Textbook precision opamp rectifiers like these don't work very well at high frequencies, since the diodes eventually just become capacitors.

At this point, I'm thinking of using an analog switch/mux driven by a high speed comparator.  Something like this is an option: http://electronicdesign.com/analog/high-speed-full-wave-rectifier-requires-no-diodes-few-parts

However, I still feel like I'm missing something.  Is there a better way to accomplish this task?

[KerryW]

I think the easiest way would be a peak detector using an LM311 comparator.  A schematic is shown in the data sheet.

Kerry

[Odysseus]

That's also an option.  I just simulated the same circuit using the faster LM319, and it works perfectly around 200KHz, but the output is 20mV low at 1MHz.  Not bad I suppose.  Interestingly, that circuit doesn't work as drawn.  The inputs to to comparator need to be swapped.

Also, it apparently makes a good half wave rectifier if you omit the cap and change R1 to about 1K.  I find that suspect, though and I think it needs to be physically tested.

[SeanB]

What about using an analogue multiplier to square the input ( connect both inputs together) and this will give a positive going waveform only that is then easier to peak detect and measure.

[Odysseus]

What about using an analogue multiplier to square the input ( connect both inputs together) and this will give a positive going waveform only that is then easier to peak detect and measure.

True, but then I could add a filter and take the square root, and now it's computing RMS,  I've looked at several IC's from ADI and LT that can do this into the many hundreds of kHz.

However, I want something that is accurate to at least 1MHz, and usable beyond that with reduced accuracy.  Rectification with an average responding filter seems to be the most straight forward method.  I'm just unsure about the best among several methods to do this. 

I guess it would be nice if there was a complete integrated solution to perform signal rectification, instead of requiring separate comparators, analog switches, opamps, buffers, etc.

[free_electron]

Make a full wave rectifier using an opamp. Get a couple of fast schottky diodes and a fast opamp and off you go.

http://en.wikipedia.org/wiki/Precision_rectifier

[Lukas]

You won't need an highspeed ADC for determining the amplitude since you're not interested in the actual waveform. A slow ADC with a sufficiently fast sample&hold does the job. It's a good idea to modulate the phase of the sampling clock to avoid mixing effects. As a bonus, you get the histogram of the signal.

[dannyf]

lets say usable to at least 1MHz or more. 

At least 1Mhz? That means from 1Mhz -> infinitely high. Not doable.

A high speed ADC

Fat chance.

Make a full wave rectifier using an opamp

Difficult as well. A discrete solution with Ghz transistors is possible.

I would take a different approach: use a programmable dac + a high speed comparator: compare the input signal with the "reference signal" from the dac. The output should be pulses. Increase the dac's output until the pulses just disappear -> the dac's output voltage would be the peak input voltage.

Fast comparators are fairly cheap.

[Vgkid]

How accurate does this need to be, what about an average responding meter. they seem to have simple schematics.

[Len]

Make a full wave rectifier using an opamp. Get a couple of fast schottky diodes and a fast opamp and off you go.

Sounds familiar...

[Odysseus]

At least 1Mhz? That means from 1Mhz -> infinitely high. Not doable.

I meant DC-1MHz or so.

I would take a different approach: use a programmable dac + a high speed comparator: compare the input signal with the "reference signal" from the dac. The output should be pulses. Increase the dac's output until the pulses just disappear -> the dac's output voltage would be the peak input voltage.

Fast comparators are fairly cheap.

Hmm, I think I see where you're going with this, but I can't guarantee the phase of the signal to be measured. Besides, that kinda conflicts with the system design.  The analog block of the system needs to be self contained.  I.E. Sine wave in from function generator, DDS, Wien bridge, etc., and DC amplitude out to a multimeter, micro, Integrating or sigma delta ADC, scope, etc.

I'm wondering now if I can compensate for the propagation delay in the comparator and analog switch with a transmission line.  Of course, 10ns is already about 2 meters of mircrostrip.

[Marco]

AD8033/34 datasheet has a circuit for a fast(ish) peak detector.

[nctnico]

Analog devices has al kinds of signal level detectors (log amps) for RF stuff. The random sampling idea is also useful. But for DC to 1MHz the precision rectifier (diodes + opamp) sounds right to me.

[GK]

The AD8307 log detector is a popular part and does DC to well over 1MHz, with good accuracy and dynamic range.

http://www.analog.com/static/imported-files/data_sheets/AD8307.pdf

If you prefer something with a linear response however, something like a "Clamp Amplifier" could do the job (see app. circuit Figure 12 on the bottom of page 19):

http://www.analog.com/static/imported-files/data_sheets/AD8036_8037.pdf

.....just follow it with an averaging filter (either passive or active) and you're done; However that won't have anywhere near the dynamic range of the AD8307.

[Odysseus]

If you prefer something with a linear response however, something like a "Clamp Amplifier" could do the job (see app. circuit Figure 12 on the bottom of page 19):

http://www.analog.com/static/imported-files/data_sheets/AD8036_8037.pdf

.....just follow it with an averaging filter (either passive or active) and you're done; However that won't have anywhere near the dynamic range of the AD8307.

That AD8307 may be just what I was looking for, I had no idea something like that existed.  I'll definitely order one to play around with.  Dynamic range shouldn't be an issue for this application, since the input will be maintained above 100mV amplitude.

[nctnico]

The AD8307 log detector is a popular part and does DC to well over 1MHz, with good accuracy and dynamic range.

http://www.analog.com/static/imported-files/data_sheets/AD8307.pdf

Careful here! With RF parts 'DC' often doesn't mean DC. The AD8307 isn't an exception. The lowest usable frequency is 10Hz.

[calexanian]

As far as rectification up to 1mhz your basic 1n4148 should be adequate. 1N34 germanium could also be used if you want the lower bias voltage. Just make sure you take junction capacitance into account.

[GK]

The AD8307 log detector is a popular part and does DC to well over 1MHz, with good accuracy and dynamic range.

http://www.analog.com/static/imported-files/data_sheets/AD8307.pdf

Careful here! With RF parts 'DC' often doesn't mean DC. The AD8307 isn't an exception. The lowest usable frequency is 10Hz.

Yes, the AD8307 is one of the exceptions. The AD8307 is internally DC-coupled from in to out and will respond to DC if the inputs are directly coupled. I've even used it this way. Where did you get the 10Hz figure from?

http://www.analog.com/static/imported-files/application_notes/AN-691.pdf

"The low frequency performance of the following parts is discussed in this application note: AD8302, AD8306, AD8307, AD8309, AD8310, AD8361, and AD8362. (The AD8314 is not included because it contains a series capacitor at its input, which precludes its use at low frequency.) Using the appropriate precautions, some of these devices can be dc-coupled at their inputs. The external circuitry substantially determines the lowest frequency at which operation is acceptable."

"It is important to note that limitations in the signal generator did not allow for measurements below –35 dBm at very low frequencies. Using the schematics shown earlier and those found in the data sheet, the AD8307 will respond to input signals down to dc."

The original version of the AD8307 datasheet even had a "DC-coupled applications" section, but for some unexplained reason they have since abbreviated the datasheet and removed some of the applications circuits sections at the end of the datasheet, which isn't particularly helpful:

REVISION HISTORY
7/08—Rev. C to Rev. D
Deleted DC-Coupled Applications Section ................................ 22
Deleted Operation Above 500 MHz Section .............................. 23


I'm pretty sure I have a printed version of the original AD8307 datasheet somewhere around here, as that was my reference when I applied the AD8307 in a DC-coupled circuit.



 

[dannyf]

use a programmable dac + a high speed comparator: compare the input signal with the "reference signal" from the dac. The output should be pulses. Increase the dac's output until the pulses just disappear -> the dac's output voltage would be the peak input voltage.

Coming to think about it now, it can be further simplified: a capacitor + diode driven by the comparator's output can function as an automatic "dac". The peak input voltage will show up as a dc signal on the capacitor.

So a comparator, + a capacitor, + a diode + maybe a couple resistors are all you need.

The detector is also a voltage-to-pulse width modulator too.

[Dr. Frank]

The Fluke 5200 AC calibrator contains such a precision high speed rectifier / AC-DC converter.
Have a look into the manual and schematics, which effort that 10Hz - 1 MHz design requires. More modern components might simplify the circuitry.
Frank

[calexanian]

Here is the HP 410 meter. Thought it would be interesting to look at. My friend Roni at M.U. used to make those input diodes. Still has a few around the shop.


http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CEkQFjAB&url=http%3A%2F%2Fwww.hpl.hp.com%2Fhpjournal%2Fpdfs%2FIssuePDFs%2F1950-11.pdf&ei=kwu5UoCEBcjgoATH5oKIAQ&usg=AFQjCNFbNOpzU1p_fDBDJMDw84wfxC5XCg&sig2=Z2cDaDzmcCKS-j3FesFVAw&bvm=bv.58187178,d.cGU&cad=rja

[Odysseus]

The Fluke 5200 AC calibrator contains such a precision high speed rectifier / AC-DC converter.
Have a look into the manual and schematics, which effort that 10Hz - 1 MHz design requires. More modern components might simplify the circuitry.
Frank

Thank you for the tip, Frank. That manual is beautiful and humbling at the same time. It uses an opamp+diode scheme to do the rectification, although I'm having trouble understanding all the discrete circuitry that comprises the amplifier.  As you said, however, all that mess should be unnecessary with modern parts.  I guess it was done differently back then; you had to make do with what you had.

I had previously written off this method as inaccurate, but I've done some more simulations and come the following conclusions:
1) The reverse recovery time of the rectifier diodes must be very short (or zero with schottky diodes).
2) The junction capacitance of the diodes, which is typically higher with schottkys, must be compensated for by using smaller resistance values in the feedback network, on the order of 100 ohms or less for BAT54s, for example. Otherwise the capacitance begins to dominate the behavior of the diode at high frequencies and it doesn't rectify anymore.

So, obtaining a high speed opamp with the capability to drive low impedances isn't difficult, but I could use some advice in selecting appropriate rectifier diodes.  Maybe I should look into RF diodes, perhaps?  LTSpice indicates the 1N4148 has a significant 20ns recovery time, which results in approximately 1% error at 1MHz.

Here is the HP 410 meter. Thought it would be interesting to look at. My friend Roni at M.U. used to make those input diodes. Still has a few around the shop.

Amazing.  "These design considerations have resulted in the rectifier tube and it's circuitry having a resonant frequency of approximately 1500 megacycles."  1.5GHz seems impressive even by today's standards.  I also can't wait to use the unit "micromicrofarad" while talking to my graduate advisor.

[calexanian]

I would probably not use micromicrofarad. It went out of date with the kilocycle.

[Odysseus]

I would probably not use micromicrofarad. It went out of date with the kilocycle.

Yes, I suppose "puffs" rolls off the tongue more easily.

[calexanian]

There you go....  Much easier.