DDS Signal Generator - Revisited

[JimB]

A few years ago, in this thread:
https://www.electro-tech-online.com/threads/dds-signal-generator-module.138465/#post-1148288
I described the insides of my own DDS based signal generator as "a bit of an ugly evolving work in progress".

The reason for my being so scathing was that when I was developing it, I paid no attention to good RF technique.
I was focussed on getting the user interface into a form that I liked and as long as something reasonable came out of the BNC connector on the front, everything was good.

The problems became obvious when I tried to use this signal generator to test a low frequency radio receiver.
My usual sig gen (an HP 8640B) will only go down to 500kHz, but I wanted to do some sensitivity measurements at 17kHz.
Using an external attenuator to reduce the test signal from my generator, I connected everything up and found that there was so much stray signal leaking out of the home made generator that meaningful measurements were impossible.

Having thought about mounting the DDS module its self into a diecast box to give good screening, but never quite got around to it, I recently found a source of nice tinplate boxes.
**broken link removed**
Tinplate boxes are fairly cheap and you can solder to them, whereas diecast boxes are expensive and solder does not stick to them.

So I stripped out the insides of the sig gen and re-mounted the PIC control board, removed the RF output amplifier, the AD9851 DDS has plenty of output on its own for what I need, and fitted the DDS module into the tinplate box.

So that no RF leaked out of the tinplate box along the control wires to the PIC, I fitted feedthrough capacitors to pass the signals through the wall of the box.
I was expecting that the capacitors would slow down the risetimes of the pulses and that this could be a problem, but I got a big surprise, one of the signals was inverted.
Instead of a positive going pulses ( 0v - 5v -0v ) I got a negative going pulses (5v - 0v - 5v). Following the very first pulse after the PIC was reset, the digital output settled at 5v for some reason. Out of three control signals, only one of them did this. The problem was all down to the capacitive loading on the PIC digital output.
To overcome this I had to remove my nice feedthrough capacitors and fit simple feedthrough insulators with a lower value of capacitor wired separately. I also added a ferrite bead to give some inductance to help suppress higher frequency leakage from the box.

So here it is in all its glory:


And inside the tinplate box:


I modified the AD9851 module by removing the lowpass filter which was fitted on the board and added my own LPF which can be seen at the lower left of the box.
I also changed the output load resistors to give a 50Ohms output, rather than the 200Ohm output which is configured as standard on these modules.
I was getting some rather strange variations in output from the generator as it was tuned across the range of output frequencies, and the various harmonics and aliases were not as well suppressed as they could have been with the original filter as fitted to the module.

So how does it work and was it a success?
Yes it is a success, there is still some detectable leakage but it is down in the sub microvolt region, I can live with that.

As for the receiver which I was trying to test the sensitivity at 17kHz, it is as deaf as a post, which is what I expected.
But worst of all, my lovely home made VLF converter which enables listening to very low frequencies by converting them up to the 10MHz region, that is not much better.

Ah well, another couple of little jobs to do.
Mañana

JimB

 

[spec]

It would be good, in my opinion, to get some more posts about RF design and techniques on ETO. I used to work with a load of RF men and they taught me a hell of a lot about good layout and getting circuits to operate fast.

One of the RF men made a 10W audio amplifier from a circuit diagram he got from a magazine. I had to laugh because it was built with feed through capacitors, ground planes and screened boxes, but it was a really sweet sounding amplifier.

spec

 

[spec]

In the early days the first 'high speed' system I worked on was an airborne radar processor using the then new and mysterious 74(54) TTL chips from Texas Instruments. The master clock was an unbelievably high 4MHz.

We were having terrible troubles with routing this clock to to the various cards in the case: ringing, edge skew, attenuation etc. But, during a coffee break, one of the RF guys overheard us discussing these problems so he offered to come and have a look. Back at the lab he just managed to hold back a smile.

He said that he could probably sort the problems with the clock distribution. So we took him up on the offer (with funding).

Within a week he had resolved all the issues, and we had a beautiful clock, which exceeded the requirements.

The first thing he did was to bring a scope from his lab and then replace the coax cable with some RGxx rather than the piece of coax we just found hanging around the lab. He also replaced the BNC connectors with BNC connectors, also from his lab.

Then he got into the detailed stuff; ground planes, ceramic decoupling capacitors, and matching. I remember thinking that this was a touch over the top, but when I saw the radical improvement in the clock, I realized that I had a lot to learn about HF techniques. The initial move was to study some of the books from the RF lab and also the layout application reports from the semiconductor manufacturers: Motorola, Fairchild, Texas Instruments...

Since that time, I have known how important the bits that are not shown on circuit diagrams: decoupling, ground ordering, screening, etc, are, even for simple circuits.

Some months later we returned the favor and gave them a circuit to stabilize the bias current for detector diodes. Just as in our clock case, they were amazed by the performance and couldn't believe how accurate the bias current was even with temperature variations.

spec

 

[Mikebits]

That takes great patience and skill, not to mention hard work. Job well done.

 

[dr pepper]

I heard the filter on the module isnt that great.
I put one if these in a crystal oven of all things, with a 8 pin pic and serial interface, also I borrowed and modded a constant amplitude output buffer from another project I have, output imp is 50 ohm and amplitude is constant to around 40mc, and still fairly flat igher than that, but the pin diode used is now unobtanium.

 

[Mikebits]

As for the receiver which I was trying to test the sensitivity at 17kHz, it is as deaf as a post, which is what I expected.
But worst of all, my lovely home made VLF converter which enables listening to very low frequencies by converting them up to the 10MHz region, that is not much better.

Just curious, if I recall doing some work on submarines, the antennas they used were like a long, very long stretch of wire strung out behind the stern. So what sort of antenna are you using for VLF?

 

[JimB]

It is an active antenna.
A short wire, about 1 metre long, connected to an untuned amplifier with a high impedance input, ie an FET is the first transistor.
The output of the amplifier connects via a coax cable to the input of the VLF to 10MHz converter.

As for the converter being deaf, I found a couple of problems there.

1 The mixer is a Mini-Circuits SBL-1, there was no where near enough drive from the oscillator. I fixed that by replacing the exiting badly designed (by me) amplifier, with the output amplifier which I removed from the DDS Sig Gen.

2 There is a relay which switches between the VLF antenna and converter and a straight through connection for an HF antenna. That relay has some dirty contacts, although it is working now I think I should change that relay.

JimB

 

[spec]

This is an aside, but talking about VLF, the common mistake is to think that because VLF is often at audio frequencies, that it is in fact an audio wave (pressure) rather than electromagnetic (E/H).

It has been the subject of a few heated discussions along the way.

spec

 

[JimB]

This is an aside, but talking about VLF, the common mistake is to think that because VLF is often at audio frequencies, that it is in fact an audio wave (pressure) rather than electromagnetic (E/H).
It has been the subject of a few heated discussions along the way.

Indeed, a point well made. That discussion has taken place here on ETO a few times.

I heard the filter on the module isnt that great.

Here are the plots of the output power vs frequency for the original and modified versions.

You can see that the modified version (red trace) is much flatter.
The output is starting to roll off around 30 to 40MHz, as is to be expected from the sin(x)/x characteristic of the sampled DDS waveform, and at 55MHz the output takes a dive as the LPF enters its stopband region.


JimB

 

[dr pepper]

Ok I have to admit it was better to remove the problem than work around it.

I have played with vlf, but nothing serious and only rx, the 'miniwhip' electrostatic aerial works well for that, I can get russian alpha navigation on a few kc with it, well last time I tried not sure if its still transmitted now.

Just listened to something on 9.5kc using the university of twente's online receiver, it uses a miniwhip:
http://websdr.ewi.utwente.nl:8901/