Signal Generator Design

[Chryseus]

I'm currently working on two different signal generator designs, the first is a 30MHz DDS based on the Analog Devices AD9834, one minor problem with it is that the internal comparator for generating a square wave has a high pass filter limiting it to >3MHz, I've been thinking of using a TLV3501 comparator with a DAC101S101 to adjust the duty cycle for the square wave.
The output amplifier is still one thing I'm a little unsure of, I'd like at least 10Vpp in to 50 ohms with a wide bandwidth to ensure at least a rise / fall time of 20ns (preferably better), I'm considering doing a discrete voltage amplifier stage followed by a class AB output to minimize cost which is a major goal with this design.
The AD9834 according to the datasheet has some problems with images as the output approaches half the clock frequency so would it make sense to reduce my maximum output frequency to say 25MHz to avoid this (I'm using a 60MHz clock) and include a low pass filter on the output of the DAC (maybe 150MHz to ensure the triangle is reasonable) ?

My second design is a RF generator using the Analog Devices ADF4351 35MHz - 4.4GHz frequency synthesizer, it has a fairly poor frequency resolution of around 8kHz so I've been thinking would it be possible to simply adjust the clock frequency to fine tune it, maybe using the AD9834 along with a nice stable crystal oscillator ?

Any ideas would be very welcome 

[Mr Simpleton]

Looking at the ADF4351... the resolution is a function several things, and if you reduces the REF freqeuncy, to 10 MHz you can acheive 1 kHz resolution and still have fair phase noise. When going down below 2.2 GHz you do get higher and higher resolution due to the output divider.

A DDS-based reference is neat, but do not forget to filter the DDS-output.
Do you have a specific need, or just doing this for fun?

Good luck with the projects!

[nctnico]

Another drawback of the ADF4351 is that it outputs a square wave. It doesn't say much about that in the datasheet but just look at the numbers for second and third harmonics.

[Chryseus]

Looking at the ADF4351... the resolution is a function several things, and if you reduces the REF freqeuncy, to 10 MHz you can acheive 1 kHz resolution and still have fair phase noise. When going down below 2.2 GHz you do get higher and higher resolution due to the output divider.

A DDS-based reference is neat, but do not forget to filter the DDS-output.
Do you have a specific need, or just doing this for fun?

Good luck with the projects!

1kHz ? I must have screwed up my calculations then but that's still too high for my purpose.
I will of course filter the DDS output, maybe I can find a nice narrowband filter for 20MHz which will be my reference frequency.
I'm mainly doing it for fun and to add a useful bit of equipment to my lab but I may sell it if there is interest which is the reason I'm trying to minimize cost, currently as far as I'm aware there isn't much in the way of budget RF generators, with some effort I could probably get this down to £70 a unit.

Another drawback of the ADF4351 is that it outputs a square wave. It doesn't say much about that in the datasheet but just look at the numbers for second and third harmonics.

It isn't a square wave, more of a poor sine wave but in many application this doesn't matter too much so I think it will work out fine, there isn't many other parts that fit my requirements so I'll have to come up with something if it does cause problems.

[dannyf]

Feed the sine wave output to a digital gate and you get a square wave output.

As to frequency resolution, look for one with longer phase accumulator. 10Mhz oscillator + 28bit accumulator would get you to 10Mhz / 2^28 resolution, for example.

[Mr Simpleton]

There are several companies that offering a simple signal source based on this chip!
Take a look here: http://www.rf-consultant.com/simple-signal-generator/
You may get some good ideas at least!

[Marco]

It isn't a square wave, more of a poor sine wave

The output stage is in the datasheet, a long tailed pair with a current source on the emitter ... this is not a linear amplifier, this is designed to give binary output. The VCO is almost certainly not a resonant tank type VCO, but more akin to an integrator/schmitt trigger type VCO.

Any resemblance to a sine is purely because of limited bandwidth. It isn't a sine, more like a poor square wave.

[Chris Jones]

The VCO will be a LC oscillator, either using spiral inductors or bondwire inductors. To cover the octave range, there are at least 3 VCOs with overlapping ranges, of which one is used at any time. Of course at frequencies where the divider is in use, the output waveform will certainly be more like a square wave, and the output buffer probably squares up the waveform even without the divider in use.

Note that there is now a higher frequnecy variant, ADF5355.

I wish they would integrate the loop filter. This has been standard for many years on cellphone synths.

[Marco]

The VCO will be a LC oscillator

Why do you think that? I can find academic designs for GHz range integrator/comparator VCOs ... seems easier to put on a chip to me.

[edavid]

The VCO will be a LC oscillator

Why do you think that?

Because that's what it shows on the datasheet 

[Marco]

Oops, so it does :/ Ah well, if they can do that on an IC why don't any of them provide a nice wideband sine generator for us?