3 Ghz counter and power detector build Part IV Putting the counter into service

After I had soldered the remaining parts like missing passives the lcd display and user buttons the build process was finished.
Before applying power I measured the resistance of the power lines  and some other important points to ensure there are no nasty shorts hiding somewhere.

The big moment applying power.
Nothing except a short blink of the lcd back light happened .
Hmmm
Soon I figured out there is a very nice software power switch  implemented.
You need to press the ok button to switch the unit on.
Very nice, I will implement that logic in my own designs too.
Power switches are so outdated.

So the unit works. As a first test I measured the power and frequency of an fm test generator.
Without calibration the frequency part worked ok but the power measurement was of by 3 db.
To calibrate the unit I need access to the university test lab, which I don't have until September.

First impressions, pros and cons of the unit:

Pros:
Very nice power on design
Usage of qfn components I lover to solder fine pitched smd devices
Cpld design
Open source software for the microchip and the cpld

Cons:
A usb port to export data is missing
The displayed frequency is always drifting
No gate time setting
Silk screen of pcb is horrible , very fuzzy and partly unreadable (the manufacturer promised to take care of this problem though)

In the end I am unfortunately disappointed of the unit.
On the other hand it still will be a help in the spectrum analyzer build process.
I am also thinking about implementing an serial output.
To log the data in a computer could be a big help. You could for example plot the frequency versus power output of a device and visualize the stability of the power in a given range.

Finally the counter in action (Please excuse the flare on the lcd)

Links:
3 Ghz counter and power detector build Part I
3 Ghz counter and power detector build Part II
3 Ghz counter and power detector build Part III
Elektor project page

3 Ghz counter and power detector build Part III Cpld

The counter not only has a Dsp that needs programming, there is also a Cpld.
Cplds (Complex Programmable Logical Devices ) are very interesting parts.
A Cpld gives you the possibility to create your own integrated circuit.
In a very simplified explanation I would say they are made of thousands of gate arrays and configurable wires which connect between the gates and the pins.
Programming a Cpld is more like routing a pcb than programming a microcontroller.
A common language to describe your circuit is VHDL (Verilog Hardware Definition Language).

Cplds and Fpga (Field Programmable Gate Array) are very powerfull devices.
They can process tasks massively parallel are very fast but also expensive and tricky to program.

Programming a Cpld is usually done via Jtag interface.
I downloaded the Altera programmer and used a 5$ Jtag programmer clone of the Altera Usb Blaster.

Everything worked immediately .
The last step of the counter build is to add the rest of the passive components.

Successful programming of the Cpld

Links:
3 Ghz counter and power detector build Part I
3 Ghz counter and power detector build Part II
Elektor project page
Altera Homepage

3 Ghz counter and power detector build Part II Troubles programming the Dsp

When I populate a new pcb I first solder the chips with the smallest footprints.
Next in order are the bigger ics and power supply chips.

Then I make a break and program the controller.
To program a Pic chip you just need the chip, some decoupling capacitors a pullup resistor  on the reset line and the three programming data lines MCLEAR (reset line),PGC,PGD.
Usually pic chips are super easy to program, compared to the Atmel Avr line they even don't need an oscillator.

So with high hopes I started  the Microchip ide Mplab, connected the programmer to the board (using pogo pins to avoid solderin a connector) and clicked the connect menu entry.


Connecting to MPLAB ICD 2
...Connected
Setting Vdd source to target
ICDWarn0020: Invalid target device id (expected=0xF0F, read=0x0)
...Reading ICD Product ID
Running ICD Self Test
... Failed Self Test.  See ICD2 Settings (Programmer->Settings) (status tab) for details.
MPLAB ICD 2 ready for next operation

Hmmmm
At least one head scratching hours later, no difference.
The Dsp chip simply told me in this working environment we ar on strike, go f*** yourself.
Now what to do if a processor is in working refusal ?
The best way is to negotiating with their local labour union boss, the datasheet.
In a good datasheet there is often a minimal working environment section and voila there is a strange pin number 7 called Vcap/Vdcore.

Hmmmm

After further reading I understood the internal core runs on 2.5V and to avoid an extra power line the chi provides an internal regulator which needs a decoupling capacitor.

Next try:

Connecting to MPLAB ICD 2
...Connected
Setting Vdd source to target
Target Device dsPIC33FJ32GP204 found, revision = Rev 0x3006
...Reading ICD Product ID
Running ICD Self Test
...Passed
MPLAB ICD 2 ready for next operation
Programming Target...
...Validating configuration fields
ICDWarn0046:  Because clock switching is enabled, MPLAB ICD 2 requires the user to cycle target power after a program operation.
...Erasing Part
...Programming Program Memory (0x0 - 0x54FF)
Verifying...
...Program Memory
...Verify Succeeded
...Programming Configuration Bits
.. Config Memory
Verifying configuration memory...
...Programming succeeded

BOOOOOM
Success, the chip is now programmed.
Now that was some heavy lifting , usually troubles appear from the most unexpected direction.

Time for some pictures:

Dsp processor

PLL chip very small footprint

 Log Power measurement , Cpld and Pll chips

Links:
3 Ghz counter and power detector build Part I
Elektor project page

3 Ghz counter and power detector build

To measure the output of the phase locked oscillators for the spectrum analyzer I needed a fairly accurate frequency counter.

I test a counter usually against my 10 Mhz rubidium source.
Neither a cheap 1 Ghz counter from China nor a self built counter built around pic micro controller produced good results. They were way of,...

This new project which I found on the Elektor homepage raised my curiosity.
Usually I don't like Eektor projects that much,  they tend to use the most expensive components but then deliver a mediocre circuit that does not make use of all the bells and whistles this components could deliver. On the other hand I like the magazine and always  hope this time they hit gold.

First for pre scaling (dividing the input signal down) it uses a cheap pll chip instead of an expensive dedicated pre scaler. (nice!)
Second as counter it uses a CPLD device and provides the VHDL (Verilog hardware description language) sources.
Third it uses a 0.5 ppm crystal oscillator as reference.
Forth it can measure the power level but unfortunately only within an accuracy of 4 db (if the frequency is modulated). I can live with that, power meters almost always have trouble with modulated sigbak sources .

I ordered an empty pcb, luckily I already ad most of the needed parts in my lab.
In the build process I first populated the most difficult parts like the qfn power detector, the input amplifier and than the rest of the integrated circuits.
Then I soldered the passive components.
Afterwards the big components like the display were soldered.

Before power up I checked for short circuits.

Next step is programming the Cpld and the Dsp about which I will tell you in another blog post.

Counter function blocks

             1) Hf input
             2) Dc input
             3) Battery input
             4) Input splitter (Log detector, Counter)
             5) Log Detctor
             6) Preamp
             7) Pll (prescaler)
             8) Oscillator
             9,10) Linear regulators
             11) Reference
             12) Polyfuse circuit protection
             13) Jtag disable
             14) Jtag (Busblaster programmer input)
             15) Icd (Dsp programmer input)
             16) Cpld
             17) Dsp

Links:
Project link

SVR transfer reference

More than a year ago I bought a 5,5 digit HP 3478A bench multimeter.
Usually professional tools like this are pretty expensive, but if you are lucky they reach their end of life cycle in  big production companies and then they swamp the second hand market.
I bought the HP 3478A at the right time for about 100$.
There is a big problem with this multimeter, the calibration chip is battery buffered.
If you had bad luck you buy one which has lost it's calibration data and then operates way out of spec.

Recently the 6,5 digit workhorses like the Agilent 34401A are becoming more and more affordable.
I could not resist and bought one at a ham radio flee market.

Unfortunately it is hard to test such precise devices and a new calibration would cost twice as much as the device itself.

So I needed a cheap high precision voltage source to test the meter.
I found one from Geller Labs, which is based on the  AD587LQ (cerdip!) or AD587LN from Analog Devices. Since they also sell  bare pcbs I populated mine using a way cheaper (but also more temp. drifting) version  the AD587KNZ.


To get reproduceable results it is advisable to let the meter and voltage source run for at least half an hour.

Btw. I ignore the least significant digit on a display or when important I round it.
If you round the least significant digit all displays would have produced the perfect result of 10V !
That is why I can live with a more drifting version of the Analog chip and drifting it does unfortunately.
I would like to test a calibrated version with the cerdip chip, but only for curiosity.

Bbtw. 8.5 digit meters are becoming more affordable,....

Svr board with Agilent U1272A multimeter

Agilent 34401 6,5 digit multimeter

HP 3478A 5,5 digit multimeter

Links:

Geller Labs Svr transfer reference

Spectrum analyzer Part VI DDS Module(s)

Description (copied from Scotty's web page)

The DDS module is designed and configured with a filter and squaring circuit in the DDS A path.  The filter shown is a 10.7 MHz crystal filter with a 15 KHz bandwidth.  
The squaring circuit of U3 will output a CMOS level, capable of driving a 50 ohm line (J4).  
J3 output is an unfiltered output of the DDS B and will contain all harmonics and aliases of a normal DDS output.  Its output power level is approximately -8 dBm.
For best results, the Clock Input at J1 should be a 5 volt peak to peak square wave, but it will operate at a much lower input.  R3 determines the input impedance of the module.  The input clock frequency must be between 1 MHz and 125 MHz, although the AD9850 is somewhat underrated.

Build process

Soldering the dds chip is somewaht tricky, good magnifying glasses are a must.
Since the DDS chips are serially programmed, I use a different pcb (which I got in a group buy) with only 5 inputs.I think it was routed by Sam Wetterlin but I am not sure.
Studying the schema carefully is highly advisable since there is always a chance that components are placed in a different location.
Because I am building the complete" analyzer, I had to solder two of this modules, the second one is used in the tracking generator.

One of the coming blog posts will be about testing the DDS module and how to shield a Slim module properly.

DDS module rev D with serial input only

 Links:
Spectrum analyzer Part I Controller board
Spectrum analyzer Part II Phase Detector
Spectrum analyzer Part III ADC 16
Spectrum analyzer Part IV Logarithmic Detector
Spectrum analyzer Part V Master Oscillator

DDS Module

Spectrum analyzer Part V Master Oszillator

Description (copied from Scotty's web page)

The Master Oscillator, contains a 64 MHz oscillator and 3 buffered line drivers.
Each output is 5 volt CMOS that can drive a 50 ohm line that is terminated with either, a high impedance load, or 50 ohms.  A 33 ohm resistor is shown as a series element in each output.
The frequency will drift with temperature. I was able to test only one sample. It had a positive frequency/temperature coefficient of .15 parts per million per degree F (9.6 Hz / 1 deg F)

The build was fairly easy you just have to take care of possible cold solder joints around the crystal oscillator. This kind of packaging easily produces cold solder joints.

Since the unit is already tested I already shielded it.

Master Oscillator rev B

MO back side

Side view of shielded oscillator

 Links:
Spectrum analyzer Part I Controller board Spectrum analyzer Part II Phase Detector   Spectrum analyzer Part III ADC 16 
Spectrum analyzer Part IV Logarithmic Detector 

Master Oscillator 

Spectrum analyzer Part IV Logarithmic Detector

Description (copied from Scotty's web page)

The 8306 Log Detector Module has a dual function.  It is used as a detector to convert RF power to DC voltage (RSSI).  And, it is used as a high gain, RF limited amplifier.
The module has an input impedance of 50 ohms (J1) and a bandwidth of  3 MHz to 160 MHz.  
The RSSI dynamic range is -90 dBm to +10 dBm, with a DC output of +0.4 volts to +2.4 volts, on J2, "MAGVOLTS".  The Limited I.F. Output (J3) is a 50 ohm source with 50 mv peak to peak output.  
The limiter input dynamic range is from -77 dBm to +10 dBm.

Log Detector Rev 0

Build process

So far the Log Detector was one of the easiest modules to solder.

The only problem was to get hold of T1 but fortunately Coilcraft was kind enough to send me two pieces.

Once the module is thoroughly tested and confirmed working it is very important to shield this kind of units properly. 

 Links:
Spectrum analyzer Part I Controller board
Spectrum analyzer Part II Phase Detector  
Spectrum analyzer Part III ADC 16

Log Detector Module
 

Spectrum analyzer Part III ADC 16

There are two Slim modules for the ADC section.
Their main difference is resolution one delivers 16 bit the other module 12 bit.
Although the 12 bit version is cheaper and has a way easier to solder footprint, I choose the 16 bit version.
 
 Description

The ADC-16 is a dual 16 bit, serial, analog to digital converter, using two AD7685's.  
There is no manual adjustment to set the A to D range.  It is not needed to obtain excellent resolution in the MSA and VNA systems.  Each ADC will digitize its input of 0 to 5 volts to a bit value of 0 to  65535 bits.  This equates to 76.3 uv per bit.
Both A/D's will capture, and clock out their data simultaneously. 

ADC 16 rev A

Spectrum analyzer Part II Phase Detector Module

The  Phase Detector Module is a 360 degree, Phase to Voltage Converter.  It is specifically designed to operate at 10.7 MHz, but will operate in the KHz range up to 30 MHz.  
 The J1 and J2 inputs can be sine or square wave .  The min / max sine input is -20 dBm to +18 dBm.  The min / max square input is 10 mvpp to 5vpp.  Input impedance is nominally 50 ohms, but can be changed to any input impedance.
The output, at J3, is a DC voltage that is proportional to the differential phase of the input signals at J1 and J2.  The 0v to +5v output should be loaded with 100 K or higher.  A realistic phase range is from 20 degrees to 340 degrees with error less than .1 degrees.
Power requirement for the is +7 volts to +15 volts at 50 ma.

To build the RevC module I had to made some component changes and add some capacitors.
The most visible change is the added jumper wire in the bottom right.
It is not easy to solder a jumper to an ic with such a small footprint, but it is manageable as long as you have a good magnifying glass as guidance.

Phase Detector rev C

Links:
Spectrum analyzer Part I Controller board build notes

Phase Detector Slim Module

Spectrum Analyzer Build

Two weeks ago I started to build a 0-3 Ghz Spectrum Analyser based on Scotty's design.

For more information you can visit his build page.


I first started with the rev. c controller module.

From left to right: Usb2Lpt Gender changer Controller board rev C

Since the controller uses an parallel board which modern computers dont't have anymore, I used an Usb2Lpt adapter to talk to the board.
Unfortunately the controller and the adapter are routed for a female plug , so their pins are mirrored.
I could easily fix that problem by soldering a quick and dirty gender changer (the board in the middle).

First tests showed the board is working.

Links:
Controller Page
Usb2Lpt



The next step to a spectrum analyser is the  phase detector module.