Electronics > RF, Microwave, Ham Radio

Homebrew RF Levelling Amplifier

(1/6) > >>

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.

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.

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.

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.

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



[0] Message Index

[#] Next page

There was an error while thanking
Go to full version