The Results Are In

 . . . . I think.

Last night I took my Huff-n-Puff driven Master Clock for the Propeller Microprocessor to Jack's Homebrew Amateur Radio Meeting, to compare it with his GPS Disciplined 10MHz Standard (see previous post).

Jack had problem getting his equipment ready, the 10MHz Standard was questionable and he had not built the Phase Comparator/Detector as of yet. We proceeded anyway to look at his and my 10MHz signal on his Dual Trace Scope.

I had previously miss-adjusted my TCVCXO to a slightly lower frequency, so that we could watch the Huff-n-Puff circuit in action. Jack's 10MHz signal was connected to one Scope Channel (and driving the Sync), my signal was on the second channel.

With the Huff-n-Puff still turned OFF, it was obvious that my signal was  marching across the Scope screen, right to left (indicating a lower freq). We waited a short period to allow my GPS to be acquire it's satellites, and then I turned ON the Huff-n-Puff function.

Like magic my Scope signal slowed down its march across the screen as the Huff-n-Puff derived TCVCXO correction voltage was being applied. The I2C POT displayed value went from 0 (wiper center) to +12 in about 45 seconds. After a little more settling, it stopped on +14, the two 10MHz signals were almost the same, But not perfect.

At this point Jack's Standard was still suspect, because we could not verify that it was working with the internal GPS receiver and in Sync with the satellites. We were not familiar with the computer program that provided information and therefore unsure. We were just going to assume that it was working.

Jack did not have his detector built, so we did the next best thing. We put the the scope in "Signal Add" mode, so I could count the "in and out" of phase condition by observing the screen.

With Jack's stop-watch, I counted 50 nulls in 141 seconds:

50 / 141 / 10M = 0.0000000354 => 35 ppb

At his point, the best that we could do was; state that we were close in Frequency, by 35 ppb (parts per billion) or 35 x 10^-9 (I think I have the math correct).

This morning I received an email from Jack, stating that he has learned more about his Standard and its operation, and now he thinks his 10MHz Standard was probably working with the GPS and therefore we can assume it to be correct, or within reported/stated accuracy In either case, we will do the experiment again, maybe next month.

If our initial results are correct, then my Huff-n-Puff has Out Performed my expectation by 15 ppb, as suggested on my previous post were I had calculated 50 ppb.

I am a happy camper :-)

I really should rename this circuit, because it is NOT a traditional active/passive Huff-n-Puff circuit, it is mostly in Software (with only a small computer controlled POT). Maybe, I should  call it a: Software Defined Huff-n-Puff - or SDHnP !

With this configuration I should be able to provide very accurate WSPR and QRSS signals - That is the GOAL.

---

UPDATE
The method used to compare the standard with my TCVCXO did not take into account the direction (sign) of the drift between the two, and therefore the "absolute" frequency may be better than the reported 35 ppb. About half of the time my TCVCXO signal appeared to change directions relative to Jack's Standard. More experiments and some research is necessary to understand what was observed.

--

More Huff-n-Puff

I have been working several days, trying to get my Huff-n-Puff circuit to properly control the TCVCXO Oscillator, which is used as the Master Clock replacement for my Propellers.  When I started the experiment, I thought for sure it would be an easy task.

The plan was, to use a one second square wave from a GPS to compare with timing from the Propeller, the output would drive an Double Gang RC circuit, and that would provide a DC voltage to steer the TCVCXO to make the output Frequency as accurate as the received GPS. My goal was to use only passive components.

It worked, but much less than optimal.

For the Passive RC circuit to work correctly, components would have to be optimized:

• for RC filtering
• for low input impedance for the Charging Circuit
• for high output impedance to avoid Discharge
• yet, for low output impedance to provide drive to the TCVCXO

There were just too many things to consider while trying to make them all work during Startup and Stable Run conditions. I got just enough working to validate the Huff-n-Puff goal, but this circuit was not going to be successful.

To continue this approach, active components would be needed.

My new plan was to introduce two OpAmps, one to sample-n-hold the output from the comparator, a second to provide drive for the TCVCXO. This circuit could be contained within one chip, but I really did not want to build it.

Yet, another approach.

While working with another project (my UI) I had planned to provide control of the Backlight and Beeper Volume via a I2C POT. A simple I2C command sets the POT wiper value - sweet !

The I2C POT that I planned on using was a MCP4018. Late in the UI design and after ordering some parts. I noticed that you can NOT have more than one MCP4018 on a single I2C Bus (What?? only one I2C address??). I had to change the design to Quad I2C POT MCP4441, which contains 4 POTs, plus the chip provides address pins so as many as eight can be used on a single I2C circuit - very nice. My UI circuit design was modified.

Now, while thinking about my Huff-n-Puff problem, it came to me that I could replace all of the above planed active OpAmp circuits with just one MCP4018 I2C POT, that would be controlled by the Propeller (I only need one POT for this test). The POT has wiper value storage and the proper output impedance to drive the TCVCXO, via a simple RC and resister divider circuit (similar to the original passive circuit, but smaller values). The comparing, filtering and tracking will all be done in software - easy to do.

CircuitLab

The new I2C POT Huff-n-Puff was installed, and checked with the Bus Pirate. A little change to Propeller SPIN code (and I2C driver) was all that was needed to make the circuit work! The I2C driver needed to be modified because, the MCP4018 does not use internal Register addresses, only one address byte followed by one data byte.

Success

It is now fun to watch the I2C POT adjust the voltage on the TCVCXO to correct the Frequency of the Propeller's Master Clock, and then continue to stay locked onto the GPS. The I2C POT wiper's numerical position is displayed on the LCD.

Tomorrow night I plan to take this new circuit to Jack's Amateur Radio Homebrew Meeting where a Frequency Standard is available and can be compared. I hope this all works as planned. If my calculations are correct, my Propeller Master Clock should now be within +/-50 ppb of GPS Standard Frequency.

I do not expect this effort will do anything to correct inherent PLL jitter, it will only improve the Output Frequency Accuracy.

--

TCVCXO Six Hour Stats

In preperation for finish building Huff-n-Puff Circuit, and then running the Huff-n-Puff TCVCXO Spin Code on the Propeller, I have been running some Master Clock Frequency Stability and Accuracy data collection code in monitor mode and displaying the statics.

The displayed Frequency Error is computed as; the clock count form the desired 80M clocks, per 1 second pulse from a GPS. The percentage reflects; the direction, and magnitude of over a sample period. For this test the sample period is 6 hours.

Note: the resolution from the clock counter within the Propeller when compared to the GPS second pulse is 16Hz, this maybe a function of my particular GPS or the clock count function of the Prop. Regardless, the error is recorded as; ERROR/16.

After about 12 hours of data collection, the TCVCXO stats indicates about +6% X 16 cycles per 80M clock cycles, or an error of;

16 x .06 / 80M => 12 x 10^-9 

or, 12 parts per billion

The intent is to install the Huff-n-Puff circuit and then note the same stats.


Is this necessary?      No, . . but it is fun to see what frequency accuracy and stability can be obtained with simple hardware!  :-)

--