Idea for Super Peak Detector

[Mechatrommer]

Anyone have an idea on how to make a 100MHz signal peak detector?

[Simon]

err what exactly do you mean by peak detector ? does it have to pick up if the signal goes over a certain threshold or does it just have to pickup the highest peak yet to be recorded ?

[ejeffrey]

Depends on what you need.  The simplest way is also super-fast, and that is to use a diode plus a capacitor.  The peak voltage is one diode drop greater than the voltage on the capacitor.  Buffer the output with an op-amp and you could be done.  The op-amp can be slow but should have a FET input.  You can use a voltage adder with a second matched diode reference to add back the diode drop to help compensate for the temperature coefficient, but it won't be perfect.

The standard precision peak detector wraps an op-amp around the diode to eliminate the voltage drop.  However, the direct approach is very slow because the op-amp has to come out of saturation.

There are a number of tricks to improve the situation.  Horowitz and Hill has a great discussion of peak detectors.  The parts named are a bit dated but the techniques are still fine -- just have a look at what is available.

[johnmx]

http://www.analog.com/static/imported-files/data_sheets/ADL5511.pdf

http://www.analog.com/static/imported-files/data_sheets/AD8312.pdf

http://www.analog.com/static/imported-files/tech_articles/354831766LogAmpPuls.pdf

Measurement and Control of RF Power by Eamon Nash, Analog Devices, Inc.:
Part I: http://www.analog.com/static/imported-files/tech_articles/561216689RF_Power_Meas_Part_I.pdf
Part II: http://www.analog.com/static/imported-files/tech_articles/530790443RF_Power_Meas_Part_II.pdf
Part III: http://www.analog.com/static/imported-files/tech_articles/153166529RF_Power_Meas_Part_III.pdf

Is this what you are looking for?

[Kiriakos-GR]

Anyone have an idea on how to make a 100MHz signal peak detector?

If we talk about FM Band and radio signals , yes I have some ideas ..

[vk6zgo]

The classic peak detector,as already mentioned,uses a diode & capacitor.The time constant of the circuit is such that the cap's charge doesn't have time to drop substantially between detected peaks,so the charge is very close in level to the peak amplitude(less diode voltage drop).

These work well  with CW signals,FM signals,Analog TV signals, etc,but are less accurate for speech modulated AM & SSB
signals,as the peaks are not consistently time,depending on modulation.
With these signals the value of the capacitor becomes a compromise which has to stay charged longer,due to longer "gaps"
between peaks,but also needs to charge fully on very fast peaks.
For this reason,peak detectors for these types of modulation are usually more complex.

VK6ZGO

[Mechatrommer]

err what exactly do you mean by peak detector ? does it have to pick up if the signal goes over a certain threshold or does it just have to pickup the highest peak yet to be recorded ?

both or either way, anything. i'm currently on a mission to detect/find the "true" peak of a 100MHz signal, preferably faster up to 400MHz, by hook or by crook, not necessarily by the book!. i dont need a very sophisticated automated circuit. and i need linear reading not log or db. the signal is quiter large, not the tiny mV, somewhere 0.5V and up, but the smaller the better.

i googled about diode caps config + opamp as volt peak holder, but at 100MHz, i dont think the diode will be able to switch satifactorily, i can see up to ~30MHz peak detector circuit in the net using fast opamp, but i think the limit is the other component, not the opamp, ie diode.

the idea is like this (picture below), beside the diode cap idea. i have a signal, and i will impose a Vref on it, and at the end, my dso will show a straight line with just a slight bump on it. i then tune/change the Vref until the bump is barely visible, so i know, i'm close to the peak, and later i can read the Vref value using my dmm. and i can later add hi-gain opamp at the output of the peak detector circuit to make the bump more visible. thats my idea, i dont know if it can be done, no such idea on the net i find so far

ps: @johnmx, thanks for the link. but i will study that later as my net is very slow here. but googling about RF detector, it will read out RF power/strength in db, thats not what i want, as i said above, i need linear scale ie the real absolute volt reading. but maybe something about that in your link. slowly i will study that. thanx.

[Mechatrommer]

ok i read some of johnmx's links. i think i need to learn db and power log as well let say a 100MHz sine signal of 2Vpp, or ±1V. what power is it? simple basic formula? thanx in advance.

[ejeffrey]

i googled about diode caps config + opamp as volt peak holder, but at 100MHz, i dont think the diode will be able to switch satifactorily, i can see up to ~30MHz peak detector circuit in the net using fast opamp, but i think the limit is the other component, not the opamp, ie diode.

jellybean RF diodes like the 1n914/1n4148 or shottky diodes like the 1N5711 have sub-nanosecond switching times, and you can go faster.  Fast diodes are among the fastest devices around, and are often used as switches when nothing else is fast enough.  The limitation of the active peak detectors is always the amplifier, at least as far as speed goes.  The problem is that you are breaking one of the opamp golden rules: you must always operate with negative feedback.  When the signal is far away from the peak, the (+) terminal sees the signal voltage while the (-) input sees the stored value on the capacitor.  The output is at the negative rail, and the diode is switched off, which breaks the feedback loop, and worst of all the input transistor is saturated.  You can see the problem with this quite easily -- make a 2x gain non-inverting amplifier and drive it with a sine wave from a function generator.  Increase the amplitude until you get clipping and you will see the signal 'stick' to the rails.  As the signal crosses the previously held peak, the amplifier input must recover, then it has to drive the output hard enough to slew all the way from the negative rail to the the turn-on point of the diode.

All that said, if your signal is periodic, this is a drop-dead cinch to do with a comparator.  A comparator is basically an op-amp designed to avoid all those saturation problems and to have really fast slew rates at the expense of linearity.  If you can live with manual tuning, just feed your signal to a comparator and put a voltage divider on the negative input.  Adjust until the pulses just go away and read off the voltage.  If you want it to be automatic configure it just like an op-amp based circuit with a diode and capacitor, but put a series resistor in with the diode.  You choose the resistor such that the each output pulse from the comparator feeds a metered bit of charge into the capacitor, enough to move it (say) 5 mV in a nanosecond.  It will take a few dozen cycles, but your capacitor will charge up to the peak value.

One question:  If you are willing to look at the signal on a scope, why not just read the maximum value off the scope?

[vk6zgo]

The comparison idea is good,& is commonly used.
I don't think the diode will be the limitation in a diode capacitor setup. I've seen them used at 200 or so MHz for this type of circuit.

As far as the power & dB stuff is concerned, it all depends on what impedance the voltage appears across.

For instance , with your quoted signal level of 2v p-p,the RMS voltage =0.7071volts.

If this is across 50ohms,from the standard formula:- P = V Squared/R the power will be:-0.01watts,or 10 milliwatts.

If you wanted to double the power to 20mw, you will need to increase the RMS voltage to 1 volt,corresponding to 1.4142

peak,or 2.8284 p-p.

Using the last voltage  of 2.8284 p-p,or 1volt RMS,if this appears across 75 ohms,  P= V squared/R
 which in this case =1/75 =0.133 watts,or 13.3 milliwatts.

From this,you can see that the same voltage across  different impedances represents different power levels, so that if you want to find the power gain or loss of a circuit,you must either calculate the power individually for both input & output,or ensure that both input & output are the same impedance.

dB:-
The human ear,& most physiological reactions are Logarithmic.
It was found,early in Radio & Telephony,that an increase in sound power of 2 times was just barely discernible.
This made it easier to design amplifiers,etc,as it made the specifications easier to achieve.
The original unit used was the Bel,but this was found to be too large for normal use,so the
"decibel" (1/10 Bel),or dB was established.

To express a power ratio in dB,the following formula is used:- dB = 10 log P1/P2.

If both PI & P2 appear in circuits of the same Impedance,you can use the voltage formula :- dB=20log V1/V2

dBm:- expressed in dB above & below 1.0milliwatt.This is used with systems of the same impedance,but the system impedance may differ in different fields
 ( RF  may be 50 or 75 ohms ,audio 600ohms).

The advantage of dB use appears when you are using dBm; for instance if you have a signal of level +3 dBm & you put it through amplifier of 7 dB gain, the output is +10dBm.

VK6ZGO

[Mechatrommer]

jellybean RF diodes like the 1n914/1n4148 or shottky diodes like the 1N5711 have sub-nanosecond switching times, and you can go faster.

i'll study this new info to me. i'll check my diode stock if any can do the 100MHz job

All that said, if your signal is periodic, this is a drop-dead cinch to do with a comparator

i bought and done with the so called "fast comparator" LM361, its just sux and cannot do the job at greater than 20MHz or something. i cannot find any better comparator, or just maybe i'm too tired to search

One question:  If you are willing to look at the signal on a scope, why not just read the maximum value off the scope?

you nailed the golden question! well, the answer is, at the limit of my dso BW, i dont trust the peak its showing anymore. this is the golden objective, to find out how much the dso attenuate at that limit. if i can get the peak reading steady at some Vdc, i can simply read that with more accurate voltage reader, ie DMM. it also can tell about the source/FG output flatness.

edit: thanx vk6 for the info. i'll study that, i'm not very good at this power db calculation.

[vk6zgo]

With dBs you will get better with them as you use them.
The best investment is a Scientific Calculator with Logs & square functions on it.
I find if much more convenient than messing about with the calculator function on a computer.

Write down the formula every time you use it,get the required square values or logs from the calculator,write them into the formula,then  work the formula with the calculator.That way,you have a check of the working as you go.
Dave had a blog on dBs,it's well worth revising.Read the comments,too!

VK6ZGO

[ejeffrey]

I have used the max961 which is a good deal faster than the LM361.  I am using it as a threshold crossing detector which is a little less demanding than a peak detector, but it might be enough for you.

Also, I was perhaps exaggerating when I said it was easy.  The idea is simple, but ultra-fast comparators are never easy to use.  They are by nature sensitive to capacitive coupling from output to input, power supply transients, and output loading.  Pure analog or pure digital at 100 MHz is not so bad, but mixed signal is tricky, and peak detection is inherently a mixed signal problem.

You could also rig something up with a fast track-and-hold amplifier to capture the peak.  A track-and-hold is kind of like a peak detector except that instead of using a diode switch it uses an externally gated FET switch.  You would then need to come up with a trigger that accurately coincided the peak, or one that you could vary relative to a known trigger time.  You would be building an extremely simple sampling oscilloscope.  Equivalently you could just use the track-and-hold on an ADC and get the conversion to digital as a bonus.  Analog devices, for instance, has the AD9204-20 which has a 700 MHz but only 20 MSPS acquisition rate.

All that said:

you nailed the golden question! well, the answer is, at the limit of my dso BW, i dont trust the peak its showing anymore. this is the golden objective, to find out how much the dso attenuate at that limit. if i can get the peak reading steady at some Vdc, i can simply read that with more accurate voltage reader, ie DMM. it also can tell about the source/FG output flatness.

Honestly given this I would give up unless the the simple passive diode+cap works well enough for you.  You need to beg, borrow, or steal a faster scope.  You can't debug what you can't see, and no active circuit is going to work the first time.  You need the appropriate tools to be able to develop, debug, and build confidence in your circuit.

[Mechatrommer]

yes i can see the max961 is much superior than lm361. buying additional part is the last resort, esp the pricey adc. i tried 1n4148 diode last night using diode+cap+opamp, the result is not satisfying. i burnt 2 hi-speed opamp already.

[ejeffrey]

Burnt out two op-amps how?  Can you post your schematic?  What opamp are you using?  That shouldn't be possible unless you have something more fundamentally wrong -- like unwanted oscillation.  Maybe a storage capacitor that is much to big for the  opamp to drive.  You should be using a low-leakage capacitor of < 1 nF I would guess.

[Mechatrommer]

i dont know which circuit that blow it, i tried all sort of things, but basically just the opamp and diode+cap on the output and fiddling with feedback (with resistor ofcourse! required for current feedback opamp even for simple voltage follower). its the CFB THS3062 opamp from texas instrument. i dont think wrong wiring, but thats the way of CFB, they use alot of current inside the chip, the ths3062 dont have heat protection, so they just die like that, its mentioned in the datasheet that it cannot do repetetive signal, but i just stubborn like that. i have another cfb opamp circuit finished using ths3095, its amplifying allright, but its hot like hell at some frequency, luckily it got overtemp protection. i simply put a small heatsink for this current sucker. but not the ths3062, a lesson for the next chip. time to read the datasheet again, this time word by word.

[ejeffrey]

In addition to the issues you identified, it could be oscillation due to an overly high capacitive load on the op-amp.  The THS3062 recommends a damping resistor whenever the load capacitance is greater than _10 pf_!.  In this setup, when the diode is reverse biased (most of the time) the op amp only sees the diodes capacitance of a few pf.  However, when the diode is conducting the full capacitance of the sampling cap is loading the output.  Keep the sampling cap low, maybe 100 pF, and you may have to add the damping resistor in any case.

One thing you can do when experimenting with devices like this is to put a couple of resistors in series with the +/- supply rails.  If the amp starts to oscillate and draw too much current, it will pull the supply low and the resistors will take some of the heat.  The performance will be not as good due to the variation in the supply rail, but you can debug the circuit then take them out.

[Mechatrommer]

well under controlled for the damping resistor on the output, i was just only ignorant on heat dissipation previously. thanx for the advice, will keep that in mind. btw, i redo with the second chip, with double insulation for heat absorbtion. tried to be more organized this time. so here it is, the schema (reference OA-07.pdf) , real picture, and sim and reallife result, the system just cannot take it at high frequency.

[Mechatrommer]

it seems i've been hit again the 2nd time, by speed demon geeks, ie the devil "reflection", "terminator" and "standing wave in the way". i tried to make direct connection to scope (without cabling) and terminate line where appropriate, and then i got some usable peak detection (same circuit above), although there is 0.5-0.8 voltage drop from the true peak at high speed somewhere >20MHz (picture at 100MHz), i guess i should believe my scope now Dave's quote is in my mind right now... "Its a trap for young player" huhu

[vk6zgo]

The thing is doing remarkably well,up to about 10MHz,but above that the drop in frequency response of the circuit starts to kick in.
Your wiring is certainly not suitable for VHF.Long leads on components begin to look like inductors,well before they exhibit
transmission line effects.

The first rule of High Frequency circuits:- Short leads!!
The heat sinks on the op amp probably look like capacitors,too!

Try rebuilding the device,with care taken to use the shortest component leads & links possible. You should get better results!

The circuit seems to be quite complex for what it does,the most common circuits used in Radio Transmitters use the diode/capacitor circuit, perhaps followed by an op amp to increase the detected (DC) level to drive a meter.
Obviously,with that type of circuit you have to use a correction factor to account for the forward voltage drop of the diode.

VK6ZGO
 

[Mechatrommer]

thanx for the advice vk6, somehow, i know about the hi-freq rule (stray L&C) but chose to ignore sadly, maybe because lack of practice. you have encourage me to further optimize this circuit, i was about to abandon this idea. what i'm afraid of is that the opamp or diode will deteriorate by themself in performance at hi-freq and attenuate similar to scope behaviour at bw limit (-3db), that i cannot be sure. but we'll see what i can do with it, maybe i should build proper pcb with smd parts and proper inp output terminal so i wont need a single coax cable for connection. thanx alot for encouragement/advice/idea/reminder.

about the circuit complexity, i read in the net that normal diode+cap will have the diode voltage drop in the output reading (as you mentioned), so the opamp+diode+cap config is so called "super diode config" where no voltage drop occur. but thats what i read. and i just follow from existing circuit.

thanx, regard,
Shafri.

[vk6zgo]

These links are worth looking at:-


http://ve2zaz.net/RFVoltMeter/RFVolt.htm

http://www.io.com/~n5fc/rfprobe1.htm

http://www.techlib.com/files/detect.pdf

If you are only trying to check the  frequency response of your DSO,it doesn't really matter if the detector has errors,providing they are the same for all frequencies.
The ratio of DSO deflection to DC reading can be determined at a fairly low frequency,then if the ratio changes,it is indicative of the DSO response.

VK6ZGO

[Mechatrommer]

thanx vk6 you keep the brain running. i should have tried this in the first time. i just solder a 1n4148 diode with 1uF smd on my signal circuit, connect it directly to dso, and walla! better result! wtf, am i just stupid or what? now with some intelligent, i can plot and estimate my signal output and dso reading, not to bad! but still, its not a super detector (0.8v drop). what a waste with hi-speed opamp, at least i learnt and appreciate. thanx.