DDS function generator performance

[Radio Wrangler]

Oh, I simplified it a bit. To go fast, it's usual to break the adder into sections and run he lower sections early in a sort of look-ahead scheme because adder logic means a long chain of carries will limit speed, and this is the way round. It's easy because the tuning number doesn't change very often.

David

[Al (astral highway)]

Originally Posted by Radio Wrangler

Phew! both typing fingers want a lie-down right now.

Not surprised, David! Thanks for taking the time and effort to write that up. I find it very interesting and illuminating reference material and will tag it to come back to again.

Cheers

[Al (astral highway)]

I need to attenuate the signal from this in order not to overload some logic circuitry I'm passing the signal through...(destroy, at worst)

I can't, obviously, put a large series R in the output from the DDS side if I want a square wave as I'll end up with a sawtooth.

Apart from taking the output direct from R19 in the ladder, hence bypassing the op-amps, do I have other options?

Relevant circuit fragment attached. It's a part of what jimmc101 referenced in #9

Cheers

[Radio Wrangler]

I don't know. The diagram is too small to read clearly and the identity of the chip is cut off, so I can't tell which it is.

The ladder is a DIY digital to analogue converter. It is not just a simple attenuator. The individual outputs from the chip will be jumping up and down rather erratically all will have the same amplitude, but quite different timing. The resistor network is there to sum these all together in an unequal way... the relative weightings are binary 1 unit, a half, a quarter, an eighth... etc.

IF the chip has a program to make a square wave, it will only be good up to a smallish fraction of the frequency range it can make a sine over. If you try to use this higher in frequency the edges of the 'square' will get very jittery andthemark-space ratio will go all over the shop. The higher the frequency, the worse it gets.

The normal way to get good squares across the frequency range of a DDS is to have it make a sine (and many of the best have no other option) What you get out of the resistor network DAC is a sine which also becomes a sort of erratic lot of big steps as you go faster. BUT and this is the crucial BUT, a relatively simple lowpass filter will remove all the jumps and leave you with a pretty good sinewave, to as high a frequency as the DDS can go. You then use a comparator to square up the filtered sine and you now have a squarewave which can be tuned from zero hertz to about 40% of the clock frequency the DDS is running at, and it stays square across the range.

There are plenty of boards for the Analogue Devices AD9850 chip which is McVities of chocolate biscuits. It has a decent DAC inside it and a comparator. Analog Devices' datasheet shows how to do a filter and how to connect the comparator.

There are digital function generator chips and arbitrary waveform synthesisers, but unless you stump up a lot of cash for a very very fast one you are not going to get a useful squarewave at the frequencies your Tesla coil will need without going the sine, filter, squarer route. There is a mathematical reason why the sine portion gives far higher usable frequency.

David

Originally Posted by astral highway

I can't, obviously, put a large series R in the output from the DDS side if I want a square wave as I'll end up with a sawtooth.

You could use a smaller one, most input circuitry is designed with diodes to the power rails and can take a few mA.

[Al (astral highway)]

Hmmm, but there is one logic device , gate enabler, that is supposed to have max 1uA input!

That 1uA is the maximum current it needs to work so you can design the drive (mind you designing something that gives less than 1uA would be odd), if you go over (or under) the power rails by more than 0.6V (one diode) current is taken limiting the voltage. Most chips have these diodes, you can check by measuring between the pin and the power rails using the diode function of a multimeter.

What is the device?

[Al (astral highway)]

Hi there, the logic devices through which my test signal passes are SNHC14, SNHC74 and then the high and low side drivers.

Under final operating conditions (not applicable yet), there are also current sense and shutdown devices.

They (I think, the part numbers are guessed as 74HC14 and 74HC74) are common CMOS parts and have those diodes fitted, it does vary with manufacturer but 10mA wouldn't be too much.