Home Made Tiny Network Analyzer NWA

[balu]

Hi,

Electronic is just my hobby and not my profession. I started years ago with digital technic and since about four month I engaged in the analog technic stuff. I decide to design a network analyzer to learn more about filter, amplifier, mixer, oscillator etc. In my point of view, it is a good idea to have an objective. Because defining a way to learn stuff is much easier if you know, or have an idea, about the result and what is it good for.

Yes, I know there are a many NWA’s on market – used, new etc. But some time the way is more important than the objective.

First of all I will define the major requirements or my major needs to margin work and efforts.
The network analyzer should be:
- based on modules (easy exchange of function blocks),
- inexpensive (use of standard parts, if possible one or two sources for parts to reduce additional cost. E.g. shipment),
- PC controlled via USB, Bluetooth or RS-232 (no external controls like switches, displays etc.),
- have a frequency range from about 10Hz to max. 50MHz.

I will use the NWA for (see attachment to this post)
- measure signal level changes,
- phase changes and
- reflection/impedance of the DUT.

I will post my ideas, current status, results, questions and more from time to time in this section.

[balu]

NWA – The Basic Concept

The NWA have three building blocks or units based on my concept (see attachment).

The Signal Generator Unit (SGU) will provide the sine signal (from about 10Hz to max. 50MHz) for the test of the DUT. A signal splitter will provide the reference signal for compare. To get the signal reflection from the DUT I will use a RF coupler.

The output of the DUT will deliver a signal to the input of the Signal Detector Unit (SDU). A low pass filter will help to reduce disturbing signals above 50MHz. The final stage of the SDU is the signal detector and ADC.

I will use an Atmel ATmega microcontroller to control the SGU, SDU and is the interface to a PC via RS-232/USB or/and Bluetooth. The microcontroller will have a command interpreter for measurement sequence execution (e.g. for signal level vs. frequency).

[Richard Head]

Balu

Considering that you are not an electronics professional I think that this is quite an advanced project, and I encourage you.
This sort of thing does already exists from a few firms such as MiniVNA and Array solutions. Google them if you aren't aware of them yet.
I would try to get a higher upper frequency limit if possible but maybe 50Mhz is a good start.
Good luck and keep us posted.

[T3sl4co1l]

What kind of options do you expect for the oscillator and ADC?  How much bandwidth, and how does it interface with an AVR processor?

Tim

[balu]

NWA – The Oscillator

Thanks Richard – You know – no risks no fun.

I chose the AD9803 as the heart of the oscillator. Because the interface is easy, the price is low and I have one of these modules from china.

I have made an easy test setup with this module to measure the output signal (level and quality). The result is strange. The first diagram shows the output level vs. frequency without output load. I used a HP54601A oscilloscope for measurement, sure with low precision. But I get the basic information about the signal behavior. The module provides an input (R-I) for amplitude adjustment. I used a MCP4151-10k potentiometer IC to adjust the amplitude depending on the frequency (controlled by an ATmega). The second diagram shows the result with a 200Ohm load at the module output, because the module has an output impedance of 200Ohm. All measurements are done with the HP54601A.  The last attached image shows a screen shot of the HP with a signal of 50MHz.

The blue line in diagram 2 shows the output signal after adjustment and with a 200Ohm load in Vrms. The signal is very low but much flatter than before. The left signal level differences can be compensated with the calibration process (for each measurement with the NWA).

My next step will be the amplifier, filter, splitter and coupler design.

[balu]

Sorry,

And now the diagrams.

[balu]

And the schematic for the oscillator module test with the MCP4151-10k (10kOhm) potentiometer IC.

[balu]

NWA – The Oscillator (2)

The mean value of the oscillator output level is approx. 77,17mVrms (109,1mVp or 218,3mVpp). Regarding the 200Ohm output impedance of the oscillator I need an impedance matching network to adapt the module to 50Ohm input of the splitter, coupler and other.

I chose a LC low pass filter with the following data:
fg = 85Mhz
Zin = 200Ohm
Zout = 50Ohm

The result is the LPF (see attachment) with two 8,2pF capacitors and a 150nH inductor. The loss of the LPF will be approx. 14dB. The output level of the LPF 8 (with the input level approx. 109mVp) will be about 21,8mVp (15,4mVrms). That means I will get a output level (oscillator/LPF) approx. -23dBm.

[dannyf]

I chose the AD9803

What's ad9803?

The result is strange.

Your pot is connected to Rset? You may want to read the datasheet and the schematic of your module to see what requirement there is on the resistance.

To flat-out the output amplitude, you just need to change the pot's value based on the output frequency and your measurements - aka an open loop control.

The last attached image shows a screen shot of the HP with a signal of 50MHz.

At that kind of frequency, you may also want to take a look at the output waveform to see if it is still good enough for your application.

[balu]

Hi Dannyf,

I mean AD9851 - sorry. It is not a good idea to write and read (datasheet of the AD8302) at the same time

There is a resistor (392Ohm between GND and DAC R-SET Input) and jumper on the module to setup the amplitude level. But the level will change with the frequency. I think my solution with the MCU controlled potentiometer IC works fine. A adapt the software routine to adapt the resistor value regarding the frequency.

[dannyf]

You probably want to see if your pot is in serial / parallel with Rset.

BTW, 392ohm of Rset is very very low -> Iout is very very high.

[tinhead]

There is a resistor (392Ohm between GND and DAC R-SET Input)

that should be normaly 3.92k

I think my solution with the MCU controlled potentiometer IC works fine. A adapt the software routine to adapt the resistor value regarding the frequency.

sure, that can work, but i would not use 8bit pot to set values for 10bit DAC. There is example circuit in the datasheet, why you not use it? (or at least something similar and not a pot)

[balu]

Thank you for the information.

I have an MCP4822 (simple 12Bit DAC) and will try the solution with a DAC.

[balu]

NWA – SGU Signal Generator Unit

I have made my first concept for the signal generator unit. The picture shows the signal level (dBm) for each stage. As a splitter I chose a simple tee circuit. It is not the best and I have to look for a better solution.

[dannyf]

but i would not use 8bit pot to set values for 10bit DAC

The two (8-bit pot vs. 10-bit dac) are unrelated.

There is example circuit in the datasheet

The two approach are identical - by changing the equivalent resistance on Rset pin.

[PA0PBZ]

I'd put C2 on the other end of L1.

[balu]

Hi PA0PBZ,

C1 and C2 will be one capacitor with 16,4pF. But this value will not be available and I use two 8,2pF in parallel as a reminder.

[balu]

NWA – SGU Signal Generator Unit

Now I have my first and draft version of the schematics for the signal generator unit SGU. Part 1 is the oscillator module with amplitude adjustment, the impedance matching network as a low pass filter and the first RF amplifier stage.

The second schematic is part 2 of the SGU. The input signal to this part comes from the output of the RF amplifier of part 1 and goes into the pi-attenuator (loss: -27dB). The following stages are the splitter, two RF amplifiers and the RF coupler.

[balu]

NWA – SGU Signal Generator Unit

For test purposes and to get information about the possible mechanical dimensions I have made a draft version of the PCB layout for the signal generator unit part 1 and part 2. I will use metal shield housing for each module. My favor electronic supplier offers two types of these housings. A small one with 54x29x16mm (2,12x1,14x0,62”) and the other with the dimension of 54x50x19mm (2,12x1,96x0,74”).

[Berni]

I don't have much experience in terms of RF design but that Impedance matching network seams like a bad idea. You proabobly don't want to put capacitors directly on the DAC output and I don't see how that looks like 50 Ohm except at the just right frequency when that inductance happens to have such a Z. Certainly not over the wide frequency range.

As for those RF amplifier blocks I'd more likely use opamps. These RF amplifier blocks are not meant to do stuff as low as 10Hz (Not saying they don't work so low, but might be quite unpractical). It is not too hard to find 500Mhz GBW opamps to give you plenty of gain at these high freq. As a plus opamps don't care about the exact input and output impedance, so you don't need to worry about them as long as you keep traces between chips short.

[dannyf]

NWA – SGU Signal Generator Unit

You probably want to do some math to see how those RF amps work in your circuit. Most RF amps, like those MMICs, have very low input impedance. Mating them to an electrolytic capacitor (with high ESL) may not produce the desired frequency response for you.

It is better that you do more homework upfront.

[balu]

Hi Berni, hi dannyf,

Thank you for the hints. The RF amp stages are my major pain. I need amplifier with a relative high gain. Opamps with a GBW of 500MHz and more are not really inexpensive and not easy to get (for me). My idea is to use MAR-8A+ amplifier from micro-circuits with the amplification for DC to 1GHz. I have ordered one of these amplifiers and will make tests. I hope, I will get this amp tomorrow.

[Berni]

The passives around the RF amps are going to be very troublesome, that inductor that feeds it power needs to be large enough to not pass 10Hz and such.

Also you if you keep the gain levels low (I don't see a problem with that since the DAC has a large output voltage) then you can get away with like 100MHz opamps and in case of larger gain use 2 or 3 stages to get there. It's very easy to get a super flat frequency response with them too.

You can also try asking TI for a few samples of there fancy high speed opamps (might even have to fill a form since supposedly they can be used to build a guided missile or something ridiculous like that)

[dannyf]

My idea is to use MAR-8A+ amplifier from micro-circuits with the amplification for DC to 1GHz.

I don't know what filters you intend to put in front of the RF stage but you should think hard about the potential of the low input impedance loading down the filter (if you use one). A buffer (like a jfet follower) may be needed here - actually one of the VGA modules by 3G has an interesting output stage utilizing npn+pnp pairs.

Other options are video opamps or CFB opamps (they are typically good for 100Mhz or more).

My own favorite is AD603 - it has the added benefits of allowing each AGC.

[KJDS]

If I get some spare time tomorrow I'll do a proper design of the RF amp using transistors. What you have won't work well as the impedance matching network will only work at one frequency and the inductor for the amplifier will either be too small to block the signal at 10Hz or make it bigger and it will look like a capacitor at higher frequencies.