Homebrew RF Levelling Amplifier

[G0HZU]

Hi, there has been some interest expressed in my homebrew RF levelling amplifier so I've set up this thread on Eevblog to describe the way it works and what it can be used for. There's also info on how to make one and this info will be expanded on as the thread progresses. I hope you find it interesting or useful.

Going back to the 1960s or even further HP used a source levelling system in their scalar network analysers as in the first image below. This uses what may at first appear to be a flawed form of resistive splitter because it only uses two 50R resistors. This would appear to cause an impedance mismatch. However, because of the AGC/normalisation provided by this system a virtual ground is formed at the feedpoint of the splitter. This means the source impedance fed to the DUT is defined by the quality of the series 50R resistor in the RF splitter. The levelling is achieved because the level detected at the reference channel is used to level the source via normalisation/AGC. See the first image below. For accurate measurements with an old and basic system like this it really is important to have a source with a very low VSWR. It's also important to have a system that can constantly correct any level changes and keep the source power at a consistent level.

It's also possible to use this system with a signal/sweep generator with an RF power meter placed in a feedback path as part of an ALC system. See the second image below. This uses a classic HP 83752A sweep generator, the 11667 resistive splitter and a power meter to make up a the auto levelling system. This 20GHz sweep generator has an external ALC port fitted and this port connects to an internal levelling amplifier that keeps the RF output level levelled via ALC action. This setup ensures that the source impedance of the system has extremely low VSWR and it is defined by the quality of the series 50R resistor in the RF splitter. The accuracy of the levelling is defined by the accuracy of the power sensor used. This forms a very elegant system that minimises measurement uncertainty.

Few signal generators offer external ALC like this so I put together a very crude levelling amplifier that can achieve a similar result using external AM modulation. The signal generator must support external AM modulation and the modulation frequency range must operate down to DC for the system to work. See the third image below. This shows a block diagram of a typical working system using my levelling amplifier. It's worth reading all the comments in this block diagram as this can help understand how to set the system up and how to avoid common mistakes that can lead to a non-working system.

It is also possible to replace the exotic RF power meter with a low barrier Schottky diode detector and I'll give some info and advice about this option later in the thread.

[G0HZU]

See below for an external view of my levelling amplifier. It looks OK on the outside but it is built using the ugly method internally. The second image shows a schematic for it. I've only just created the schematic because up until now it was in my head. This levelling amplifier rarely gets used because I'm lucky to own a HP 83752A 20 GHz sweeper that has an internal levelling amplifier.

I don't think it's easy to make a universal levelling amplifier as the system is prone to instability especially if the power meter has a slow response time. It's really easy to get gain around the loop at 180degrees phase shift if the power meter is slow to read/respond and the system will just go unstable.

I think this is why the old thermistor based HP432A power meter was popular for use here. It responds much faster than a thermocouple power meter and it also has manual range selection. If you do use a thermocouple power meter then make sure it is set to manual range mode and try and avoid using it at power levels below about -12dBm or the meter will be too slow to respond.

It is possible to slug the response time of the levelling amp to compensate for this but it does make the system painfully slow to use. I'll dig out a few sig gen models and try a few power meter types to demonstrate what works and what doesn't. I know the little Marconi 2022C supports external AM down to DC and it works in this system.

The modern Agilent/Keysight E4419 power meters work in this system and so does the older HP432A/478A power meter and also the old Anritsu ML4803A power meter. As long as the power meter has a recorder out BNC connector on the rear panel there is a good chance it will work OK in a system like this.

The levelling amplifier shown below does work well but it probably needs polishing to make it closer to a universal design. In other words, don't rush out and build a copy and expect it all to work with your sig gen and power meter. The power meter might be too slow and this could cause instability and also your sig gen might not support external AM down to DC. Also, make sure there is a some form of current limiting at the op-amp output and that your sig gen can tolerate positive and negative voltages at the external mod input connector. I've limited the op-amp in the levelling amplifier to +/-5V and fitted a series 120R resistor. The op-amp is a JFET type and can probably only produce +/-4V at the output with limited current capability. This should be safe/OK for most sig gen types.

The 11667 RF splitter is really exotic and expensive and works up to 26GHz. Some versions work up to 67GHz! I think the B version with 3.5mm connectors (compatible with SMA) costs about £1500 new.

It is possible to make a homebrew version using parallel 100R SMD resistors and this can work well to about 6GHz. I have used my homebrew 11667 splitter to level the HP83752A up as high as 18GHz but the source VSWR will not be good up at 18GHz. I can show how I made this but even a basic attempt at making one using SMD resistors on a small PCB should work very well up to 1GHz or so.

[G0HZU]

Here's a screenshot of a recent attempt to measure the frequency response of my Agilent PSA spectrum analyser on band 0 up to 3GHz. I used the levelling amplifier and a decent lab power meter to level a sig gen using the levelling system described in this thread. Note the scaling of the analyser is set to just 1dB/div so this is a very impressive result! it looks like the flatness is close to about +/- 0.1dB across 0-3GHz.

Without the levelling amplifier this would be very difficult to achieve as a regular sig gen will suffer from mismatch uncertainty (imperfect source VSWR) and imperfect RF levelling and also loss and mismatch in any RF cable used to connect it to the analyser.

The other image below shows the typical flatness of this analyser as measured by Agilent across 6 similar analysers. The Agilent PSA analyser was one of the best spectrum analysers available about 15-20 years ago and some aspects of its performance are still incredible today.

The slightly lower level response at lower frequencies in my screenshot are partly due to the fact I didn't correct for the power meter head efficiency (cal factor) and just ran with one fixed cal factor. This spoils the result slightly but it is still very impressive. If I had included cal factor correction for the power head I think I could have matched the flatness of the factory measurement. Almost certainly Agilent will have used a similar setup but they will have used a genuine 11667 RF splitter.

[G0HZU]

Whoops! I just spotted an minor error in my hastily drawn circuit diagram. The top of VR1 should go direct to the 5V supply, not after the 12R resistor fed to the op amp. This probably doesn't matter much but the top of VR1 really should be connected direct to the regulated 5V supply line. I've now managed to upload the corrected schematic to the earlier post so the earlier schematic should be OK now in this respect.

[ahbushnell]

Are you using a TL081 Opamp?  I looked it up and it has a unity gain band width of 5 MHz? 

Thanks
Andy

[G0HZU]

Yes, I used a TL081 as I had one handy at the time. It doesn't need to have much in the way of GBW because it only gets fed the DC (or close to DC)  recorder voltage from the back of the power meter. The recorder output of this type of power meter is often connected to a plotter or a DVM.

The TL081 then only has to provide the DC (or close to DC) output to feed the external AM port of the sig gen. The TL081 is a JFET op-amp and has a high input impedance a fairly low offset voltage so it should work quite well in a circuit like this.

[SilverSolder]

Very cool project.  I guess the idea of using the AM input to level the generator would work at lower frequencies too, if you need an accurate amplitude sweep e.g. at audio frequencies?

[ahbushnell]

Yes, I used a TL081 as I had one handy at the time. It doesn't need to have much in the way of GBW because it only gets fed the DC (or close to DC)  recorder voltage from the back of the power meter. The recorder output of this type of power meter is often connected to a plotter or a DVM.

The TL081 then only has to provide the DC (or close to DC) output to feed the external AM port of the sig gen. The TL081 is a JFET op-amp and has a high input impedance a fairly low offset voltage so it should work quite well in a circuit like this.

I see. Thanks!

[David Hess]

I don't think it's easy to make a universal levelling amplifier as the system is prone to instability especially if the power meter has a slow response time. It's really easy to get gain around the loop at 180degrees phase shift if the power meter is slow to read/respond and the system will just go unstable.

I think this is why the old thermistor based HP432A power meter was popular for use here. It responds much faster than a thermocouple power meter and it also has manual range selection. If you do use a thermocouple power meter then make sure it is set to manual range mode and try and avoid using it at power levels below about -12dBm or the meter will be too slow to respond.

I thought the advantage of a thermister over thermocouple in a power meter is much higher sensitivity.  I think one design I studied from HP uses the thermister as the 50 ohm load, and a DC current is driven into it to maintain the same 50 ohm resistance, which indicates what the power level is.  High performance DC current clamps work like this also.  It has the advantage that the thermister linearity is removed as an error term.

[G0HZU]

I don't think it's easy to make a universal levelling amplifier as the system is prone to instability especially if the power meter has a slow response time. It's really easy to get gain around the loop at 180degrees phase shift if the power meter is slow to read/respond and the system will just go unstable.

I think this is why the old thermistor based HP432A power meter was popular for use here. It responds much faster than a thermocouple power meter and it also has manual range selection. If you do use a thermocouple power meter then make sure it is set to manual range mode and try and avoid using it at power levels below about -12dBm or the meter will be too slow to respond.

I thought the advantage of a thermister over thermocouple in a power meter is much higher sensitivity.  I think one design I studied from HP uses the thermister as the 50 ohm load, and a DC current is driven into it to maintain the same 50 ohm resistance, which indicates what the power level is.  High performance DC current clamps work like this also.  It has the advantage that the thermister linearity is removed as an error term.

In this case it is speed and stability that matters most. If I try and measure a low power level like -23dBm with a thermocouple meter it will take the meter many seconds to stabilise. It could easily take 10 seconds. An old thermistor based meter like the 432A with the 478A sensor settles in a tiny fraction of a second. The levelling amplifier would be painfully slow to use with a modern thermocouple meter at low power levels like this because the ALC loop would have a huge settling time each time a new test frequency or power level change was selected.

[niconiconi]

Your design is a GREAT solution for instrument calibration , and a worth a mention in the Metrology section of the forum as well. Many vintage oscilloscopes could benefit from a complete calibration, however the official procedure requires a signal generator with an amplitude accuracy of 3%-7%. However, leveled sinewave generators are exotic and expensive, and don't have many uses beyond calibration and performance verification. With your design, it's possible to perform the same task with a standard signal generator and RF power meter. For most oscilloscopes, 200 MHz is usually adequate, 1 GHz is basically the highest frequency required, the splitter is easier since it doesn't need to have good microwave performance. I suspect the RF power meter can also be replaced by a cheaper diode detector or log converter, for sinewave leveling you don't need thermal-based RMS measurements, just a peak detector.

[G0HZU]

Hi niconiconi. Thanks for your comments.  It is also possible to lash something up for low frequency use that doesn't require a modern sig gen with external AM capability but the sig gen would have to have low harmonic content if a low cost diode detector was used as the level sensor. See below for my fairly hideous prototype. This uses a chunky PIN diode attenuator that works well even on the LW band as the diode has a very long carrier lifetime. A dual gate MOSFET attenuator could be used here as well.

The setup below levels really well from LF through to about 100MHz. It uses a diode detector with a negative output voltage. The system has a levelling range of about 20dB and this is usually enough for most things. Obviously, it would work best if fitted in a nice enclosure and if SMD parts were used for the RF splitter and diode detector. However, the board below was intended as a hasty prototype to prove I could get something that would level a cheap/vintage sig gen without the system going unstable. There's nothing expensive there apart from the PIN diode and this could be replaced with a different type of attenuator. The attenuator doesn't have to be accurate, it's inaccuracy is corrected by the feedback.

I used two op-amps in series for this so I could break the loop open and plot a Bode plot to look at stability margins. However, I never bothered to try this. It works really well and is really fast because the diode detector response is really fast. With this setup any type of cw sig gen could be used, even a vintage one as long as the generator doesn't produce high harmonic distortion levels.

[G0HZU]

There are ways to improve the frequency response of the diode detector so reasonable performance could also be achieved up into the lower part of UHF using a cheap diode. I used the little PCB above to compare the response of various RF power sensors across 300kHz to 50MHz and I was getting good results down to about +/- 0.02dB. I did use a sig gen with very low harmonic content as I wanted to make very critical comparisons.

[niconiconi]

Thanks for the follow-up. This is my idea as well: feed a CW signal into a variable attenuator, split the signal into two paths, one to the receiver and another to a detector. Finally, control the variable attenuator with the error signal from the detector, close the control loop, and you get leveled sinewave. So I decided to do a quick investigation on its feasibility, and it was how I found your post.

The challenge is whether the design can be accurate enough to match an oscilloscope calibrator. I used the classic Tektronix SG504 as a reference, with the following specs (the original specs are all in voltages, here I converted voltages to power for easy comparison).

Frequency Reference: 50 KHz, 6 MHz
Frequency Range: 245 MHz to 1050 MHz
Leveled Amplitude: 0.5 Vpp (3.9 -2 dBm) to 4 Vpp (5 Vpp, 24 18 dBm @ VSWR 1.2:1, 0-35 C)
Amplitude Accuracy: 3% at reference (0.256 dB)
Amplitude Flatness: 4% to reference (0.340 dB)
    ...so Worst Case Absolute Accuracy is 7.12% (0.597 dB)
Harmonic Content: 25 dBc (2nd), 60 dBc (3rd and more).

As you can see, distortion shouldn't be a big issue if Tektronix could do it with an external diode detector from a signal source with only 25 dBc harmonic suppression. The main problem is the razor-thin tolerance for both absolute accuracy and flatness. Leveling at a power level as high as 24 dBm 18 dBm can also be problematic.

Being a novice, I have absolutely no idea about what kind of performance is achievable in a discrete design, and even if I do I currently have no equipment to characterize them. But I wanted to start plugging some numbers in and see what I get. So I was looking into RFICs with guaranteed specifications, log amps in particular. I found most log amps are optimized for use in RF receivers, they have the best accuracy and linearity around -20 dBm to 0 dBm. 0.25 dB or higher accuracy is achievable under favorable conditions. Many have RMS measurements as well, so no need to worry about distortion. But above 0 dBm they can't be trusted, there will be significant errors like 3 dB at 10 dBm. So it's necessary to put a 20 dB attenuator before the log amp to knock the power down. Then the attenuator's tolerance also becomes the problem, especially a large one like 20 dB. The best SMD RF attenuator from Mini-Circuits has 0.15 dB flatness under 1 GHz, but the reference accuracy is 0.4 dB max. Finally there's also the tolerance of the variable attenuator, these errors quickly add up. Of course, actual performance is likely better than guaranteed specs. The errors can also be removed with manual calibration. But proper characterization remains a problem.

What kind of performance is achievable in a discrete design? Any input is greatly appreciated.

BTW, the diode detectors used in SG504 and the low-frequency SG503 were all successfully reimplemented. These designs can be used as a starting point.

A replacement levelling head for the Tektronix SG504
http://www.perdrix.co.uk/SG504Head/

Design and Implementation of a Substitute for the SG503 Peak-to-peak Detector (Part no. 155-0107-00)
https://www.eevblog.com/forum/projects/help-needed-calculating-trace-impedance/?action=dlattach;attach=1172968

[G0HZU]

It is possible to use two diodes in the detector and this can help minimise the impact of even order distortion so a second harmonic of -25dBc won't be as significant with a dual diode detector. Otherwise even and odd harmonics can really cause a lot of uncertainty in a single diode design. Odd order harmonics can also upset the accuracy of a typical log detector so I think it is good to try for low harmonic distortion.

Leveling at a power level as high as 24 dBm can also be problematic.

Yes, that is a very high RF level. I think I've only used the little PCB setup at 0dBm output or less.

[niconiconi]

I just found the sinewave leveling problem was also extensively discussed in thread Half-decent diy RF generator utilizing AD9951 - an idea (https://www.eevblog.com/forum/rf-microwave/half-decent-diy-rf-generator-utilizing-ad9951-an-idea/75/ ) Especially page 4 on the selection of a suitable variable attenuator by Yansi.

The technical problems are very similar, with just one important difference. That project is a general-purpose RF signal generator with high dynamic range (-70dBm to 20 dBm), but a leveled sinewave generator only needs a narrow dynamic range, so in this sense our problem is "simpler" - but not really because of the accuracy requirements... The high power level also rules out many selections. Trying to work around the problem with multiple paths will make the tolerance problem worse.

[niconiconi]

I just found the sinewave leveling problem was also extensively discussed in thread Half-decent diy RF generator utilizing AD9951 - an idea (https://www.eevblog.com/forum/rf-microwave/half-decent-diy-rf-generator-utilizing-ad9951-an-idea/75/ ) Especially page 4 on the selection of a suitable variable attenuator by Yansi.

The technical problems are very similar, with just one important difference. That project is a general-purpose RF signal generator with high dynamic range (-70dBm to 20 dBm), but a leveled sinewave generator only needs a narrow dynamic range, so in this sense our problem is "simpler" - but not really because of the accuracy requirements... The high power level also rules out many selections. Trying to work around the problem with multiple paths will make the tolerance problem worse.

First project milestone. I made a leveling head using a two-resistor splitter, one output goes to an RF connector, another goes to a AD8361 logarithmic amplifier (padded by a Mini-Circuits SMD attenuator, since these log amps have significant error at high input power). The AD8361 at the reference channel is used to read an RF voltage identical to the output channel.

For the test, I used an RF power meter (HP 436A meter + HP 8484A sensor + Anritsu 40 dB calibrated attenuator) to monitor the RF output, and a voltmeter (Fluke 8840A) to monitor the DC output. Then I started turning up the frequency on my signal generator. At each frequency, I adjusted my RF power to maintain the same DC voltage, and finally read the deviation from the RF power meter. The test was performed at a power level of 11 dBm input, 5 dBm output, and my result suggests my leveling head has probably reached the 0.3 dB flatness design goal from 50 MHz to 1 GHz. Flatness below 200 MHz is even better, no more than 0.1 dB! 

Nevertheless, several challenges remain. As demonstrated, the two-resistor splitter is wideband and highly accurate, but it imposes a 6 dB power loss penalty. To reach 5 Vpp output (18 dBm), the signal generator must output at least 24 dBm, plus an attenuation budget around 3 dB. And because the AD8361 is in fact a True-RMS device, which acts like an average-responding RF voltmeter, this signal must have low harmonic distortion, otherwise there will be a Vrms to Vpp conversion error. Thus, now the problem is making a low distortion 500 mW RF power amplifier, which is not easy. An possible alternative solution is using an envelope-detecting log amp, they act like peak-responding RF voltmeters and will have relaxed distortion requirements.

BTW, I'll perform more tests to characterize the performance of my circuit. The entire setup is supposed to be automated (thus a fancy Fluke 8840A), and I already used a partial automation in an earlier test run. However, now I have encountered a GPIB mechanical incompatibility problem. My GPIB cable has a thick shield, which prevented it to be plugged into HP 436A's GPIB port  , a screw at the port is blocking it, and the screw is too rusty to remove. It's just super annoying. I need to find a better cable...

[niconiconi]

I still haven't solved the GPIB mechanical incompatibility (the alternative cable I ordered hasn't shipped yet), but I was able to work around the problem and started the automatic test by using HP 436A's analog recorder output and digitizing it with my oscilloscope.

Here's some initial test data, using the HP 8484A microwave power sensor to level the RF output at 3 dBm, and comparing readings from AD8361 at the reference output. The leveling detector circuit still needs improvements, there's peaking below 50 MHz, which is not acceptable since 50 kHz is used for the flatness reference. There's also some drooping above 700 MHz.

But still, overall, the amplitude flatness is a very reasonable [-4%, 2.5%] around the 50 MHz reference.

[niconiconi]

AD8361 frequency flatness data for all output levels has been measured at more than 1000 data points using my automatic test setup.

This time, I was using an RF power amplifier, so data at high levels can be obtained. The signal distortion is much worse thus I expect a slight decrease of measurement quality, but since both the AD8361 RMS-to-DC converter and the HP power meters are True RMS devices, probably not too much.

 0 dBm: [-4.11% @ 1000 MHz, +2.51% @  10 MHz]
 1 dBm: [-3.71% @  980 MHz, +2.72% @  10 MHz]
 2 dBm: [-3.64% @  990 MHz, +2.71% @  10 MHz]
 3 dBm: [-3.46% @ 1000 MHz, +2.36% @  10 MHz]
 4 dBm: [-3.17% @  970 MHz, +3.26% @  10 MHz]
 5 dBm: [-3.69% @ 1000 MHz, +2.87% @  10 MHz]
 6 dBm: [-3.60% @ 1000 MHz, +2.81% @  10 MHz]
 7 dBm: [-3.52% @ 1000 MHz, +2.78% @  10 MHz]
 8 dBm: [-3.29% @ 1000 MHz, +3.01% @  10 MHz]
 9 dBm: [-3.55% @ 1000 MHz, +2.59% @  10 MHz]
10 dBm: [-3.87% @  990 MHz, +2.57% @ 430 MHz]
11 dBm: [-4.56% @ 1000 MHz, +1.97% @  10 MHz]
12 dBm: [-3.50% @ 1000 MHz, +2.76% @ 610 MHz]
13 dBm: [-3.53% @  980 MHz, +2.42% @ 610 MHz]
14 dBm: [-3.20% @  990 MHz, +2.76% @  10 MHz]
15 dBm: [-4.71% @ 1000 MHz, +3.38% @  10 MHz]
16 dBm: [-4.78% @ 1000 MHz, +2.55% @  10 MHz]
17 dBm: [-4.21% @  990 MHz, +2.67% @  10 MHz]
18 dBm: [-4.27% @  990 MHz, +2.44% @  10 MHz]
19 dBm: [-4.28% @  990 MHz, +3.19% @  10 MHz]
20 dBm: [-4.85% @  940 MHz, +2.78% @  10 MHz]

As you can see, performance is pretty consistent, around [-4%, +3%] around all power levels.

If its performance is stable and repeatable (yet to be proved), it can be calibrated and functions as a reasonably accurate power leveling detector. Thus, I made the following calibration table.

  10 MHz, 102.5%
  20 MHz, 100.5%
 600 MHz, 102.0%
 700 MHz, 101.0%
 850 MHz,  98.5%
1000 MHz,  95.5%

(Also note the response from 20 MHz to 500 MHz or so is essentially flat, as my previous chart in the last reply already showed).

After calibration, the accuracy is significantly improved. 

 0 dBm: [-1.24% @  520 MHz, +1.83% @  900 MHz]
 1 dBm: [-0.91% @   90 MHz, +2.26% @  830 MHz]
 2 dBm: [-1.05% @  560 MHz, +1.74% @  810 MHz]
 3 dBm: [-0.71% @  100 MHz, +1.75% @  820 MHz]
 4 dBm: [-0.24% @  190 MHz, +2.03% @ 1000 MHz]
 5 dBm: [-0.23% @  310 MHz, +1.94% @  810 MHz]
 6 dBm: [-0.51% @  140 MHz, +1.76% @  830 MHz]
 7 dBm: [-0.40% @  140 MHz, +1.80% @  790 MHz]
 8 dBm: [-0.25% @  120 MHz, +1.99% @  750 MHz]
 9 dBm: [-0.59% @  140 MHz, +1.87% @  800 MHz]
10 dBm: [-0.62% @  600 MHz, +2.14% @  800 MHz]
11 dBm: [-1.23% @  550 MHz, +0.54% @  820 MHz]
12 dBm: [-0.71% @  100 MHz, +1.82% @  840 MHz]
13 dBm: [-0.59% @  100 MHz, +1.47% @  840 MHz]
14 dBm: [-0.41% @  620 MHz, +1.74% @ 1000 MHz]
15 dBm: [-0.58% @  540 MHz, +1.29% @  790 MHz]
16 dBm: [-1.03% @  360 MHz, +0.84% @  740 MHz]
17 dBm: [-1.22% @  290 MHz, +0.65% @  700 MHz]
18 dBm: [-1.12% @  190 MHz, +0.52% @  740 MHz]
19 dBm: [-1.58% @  530 MHz, +0.76% @  740 MHz]
20 dBm: [-1.72% @  540 MHz, +0.59% @  690 MHz]

[niconiconi]

Ah, today I found Rhode & Schwartz actually has a commercial product based on this idea. It's known as the "R&S NRP-Z28 / R&S NRP-Z98 Level Control Sensors". R&S advertises it as

Two new sensors place Rohde & Schwarz once again in the lead of power meter technology. The sensors convert any signal source into a highly accurate power reference.

https://cdn.rohde-schwarz.com/pws/dl_downloads/dl_common_library/dl_news_from_rs/195/n195_NRP-Z28_NRP-Z98_e.pdf

[e.sotillet]

Today I found a levelled amplifier containing this usually circuit, but it call my attention the fact that uses a diode in parallel with the integrating capacitor. Any thoughts about it?


Best regards

[David Hess]

Today I found a levelled amplifier containing this usually circuit, but it call my attention the fact that uses a diode in parallel with the integrating capacitor. Any thoughts about it?

It is not temperature compensated.

[e.sotillet]

Hi David, do you mean that the diode is used for temperature compensation?

[David Hess]

Hi David, do you mean that the diode is used for temperature compensation?

I mean that the diode's variation in forward voltage drop with temperature is not compensated.  RF leveling amplifiers commonly use two diodes with one compensating the other.

[e.sotillet]

Oh! Yes, a good example is the circuit of the Tektronix SG503 leveled signal generator.

Regading to the extraing diode, I think it is used in case of Vrf is too high (to speed up the opamp output voltage??).

Best regards