Author Topic: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.  (Read 3825 times)

0 Members and 1 Guest are viewing this topic.

Offline YansiTopic starter

  • Super Contributor
  • ***
  • Posts: 3893
  • Country: 00
  • STM32, STM8, AVR, 8051
Hi

regarding my last little project for my friend (Short wave AM test transmitter), I am trying to complete a test PCB for the DDS. However I have become quite confused about how to design the output circuitry properly, or at least in an appropriate manner.

So here's the datasheet of AD9834, so you don't need to look it up: AD9834 Datasheet Click Here

I have tried looking at the EVAL board for it to get an idea, however that was of no help, as they just terminate the DAC outputs with a 200ohm resistors and send it down directly to a 50ohm SMA cable. Oh well.

These are the important parameters from the AD9834 datasheet:

Output  compliance voltage: Maximum VDD-1.5V  (haven't they meant VDDA instead, see page 17?)
Maximum DAC output current: 4mA.

So where's the confusion coming from? I see many options how to implement the output circuitry. My requirement is to optimize for maximum signal level (obviously) and end up with a 50ohm output. Also to use a de-symetrization transformer to obtain the most power and best performance.

To maximize output power, the DAC has to be terminated with larger impedance, than 50 ohm. (DAC outputs are high impedance current sources). This probably makes sense.

Datasheet suggests  to terminate with 200 ohm, which gives the maximum voltage (still within the compliance range) at the minimum supply voltage of 2.3V. (I don't care about minimum supply voltage, I will always supply more than 3V. )

So the options now I have are these:

a) Selecting the load impedance for the DAC. Loading 2x200ohm yields 400ohm differentially, resulting in a weird turn ratio down to 50ohm.  Selecting too high impedance makes for a large turn ratio in the transformer, which probably isn't any good either.

b) How to wire up the transformer. There are two possibilities: First one to terminate both DAC outputs directly to gnd with resistors. Second, use a center tapped transformer, with the CT connected to GND and with a single loading resistor on the secondary. Not sure which one is better and/or why.

c) Where to lower the impedance to 50ohm? I can either use the transformer itself for the impedance match (then the filter will be at 50ohm), or make the filter asymmetrically terminated, 200ohm input and 50 ohm output. Or, use another trickery at the output of the filter (another transformer or matching circuit, for example implement a signal buffer with appropriate Zin).

d) Filter itself.  Probably 7th order is "good enough", however shall one prefer a standard Butterworth, or try implementing an Elliptic one? (I think Elliptic may be better option, when pushing the output frequency compared to the MCLK). However implementing Elliptic filter at low impedance seems problematic (too sensitive for component values).

I have also tried surfing the web, what people use, but confusion was just getting bigger, because most circuits I have found were obviously flawed, like for example this one, where there are numerous things just wrong. http://kitsandparts.com/minikits/DDS_1.gif (7mA DAC current, 100 MHz clock, pins 5-6 shorted - wtf? some overclocking hackery?, etc)

What would you suggest as the preffered options and why?

I would like to try an elliptic filter (therefore I'd need higher design impedance of the filter like 200ohm) and then a buffer would be needed to isolate the load and buffer the output to 50ohm.

Thanks for suggestions, I think other may find this topic useful, as it could be applied to other DDS chips too.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14073
  • Country: de
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #1 on: January 16, 2019, 07:12:48 pm »
The transformer for the differential to single ended can also do the 200 Ohms to 50 Ohms step. This would be a 2:1 transformer. With 2 inputs it would be a little different at some 2.8 : 1. One might use a little of the 200 Ohms and thus maybe 3:1 or 5:2.

For the termination it depends on how negative the voltage may go. If less that GND is not allowed the 2 resistors at the primary and than the transformer should be more suitable. If the output may go below GND level the termination at the secondary should be better, as this case there would be less DC voltage lost. With termination at the primary the 2 mA average current would lift the voltage by some 400 mV.

In principle the filter could be at the 200 Ohms or 50 Ohms side. However the 200 Ohms are differential and would thus need a more complicated filter (nearly double filters). As the frequency is not that high, I don't think it would so bad to have the filter at 50 Ohms level.

Looking at other DDS chips might be a good idea for the transformer: The AD9851 suggests the center tapped with termination at the output side.

A buffer / amplifier should be only after at least the first part of the filter. Not sure it is really needed. With a buffer amplifier the transformer could be 1:1 with a 200 Ohms filter and than the amplifier for 200 Ohms in.

The usual filter for sine signals is the elliptic one. When using the triangle waveform one might consider a Bessel.
 

Offline YansiTopic starter

  • Super Contributor
  • ***
  • Posts: 3893
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #2 on: January 16, 2019, 07:29:59 pm »
For the termination it depends on how negative the voltage may go. If less that GND is not allowed the 2 resistors at the primary and than the transformer should be more suitable. If the output may go below GND level the termination at the secondary should be better, as this case there would be less DC voltage lost. With termination at the primary the 2 mA average current would lift the voltage by some 400 mV.

Very good point! I almost forgot about that.  Looking in the datasheet, there is no low limit specifically mention other than in the absolute maximum ratings: Any analog IO can○t go more than 0.3V below GND. So... center tapped transformer with a secondary termination is likely a no go.

Of course, one could make a 400ohm to 50 ohm transformer with that sqrt(400/50) turn ratio, however these are quite difficult to get off the shelf. Hence I would probably go with something like 4:1 impedance ratio (2:1 turn ratio), therefore terminate the DAC with 2x 100ohm resistor. 

I have never toyed with AD9851, but I've done a design with a AD9951 (in fact 9954 which is merely the same). I think I have followed the manufacturer with a primary side termination of 2x 25ohm and a 1:1 transformer (Minicircuits TC1-1T+ was used there).

Yes I meant the buffer after the filter (does not make much sense to amplify the crap before the filter), mainly as the impedance match for the higher impedance filter, to match a 50ohm output.  I will probably need an amplifier anyway, as the DDS output will be quite low.

Okay, so some things may be more clear now.

I have also tried harder with the elliptic filter design and this may be reasonable? What do you think about it? Capacitors seems fine (I have all values available no problem) and so do the inductors.  I may also put this into another simulator such as RFsim99 and try running some component value sensitivity test, to see how will it behave.
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #3 on: January 18, 2019, 01:50:32 pm »
I'd recommend a buffer amplifier (e.g. 2N3904 emitter follower) with the filter on the output of the buffer amplifier.

A DDS will have lots of noise above the output frequency, so choose a filter with minimum or no ripple in the stopband. Personally I'd use a Butterworth for this sort of thing.
 

Offline YansiTopic starter

  • Super Contributor
  • ***
  • Posts: 3893
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #4 on: January 18, 2019, 02:12:40 pm »
Well the ripple is (according ot the simulation with ideal parts) 60dB below, which I find good enough for such plain simple DDS with a just 10bit DAC.

But I get what you mean. The AD9954 I used has had a butterworth filter.

But why the buffer in front of the filter? I think the impedance should be well enough defined from the DDS side?
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #5 on: January 18, 2019, 04:18:09 pm »
The transformer for the differential to single ended can also do the 200 Ohms to 50 Ohms step. This would be a 2:1 transformer. With 2 inputs it would be a little different at some 2.8 : 1. One might use a little of the 200 Ohms and thus maybe 3:1 or 5:2.
I have difficulty understanding this one...can you provide a math how you arrive to 2.8:1 ?

Quote
For the termination it depends on how negative the voltage may go. If less that GND is not allowed the 2 resistors at the primary and than the transformer should be more suitable. If the output may go below GND level the termination at the secondary should be better, as this case there would be less DC voltage lost. With termination at the primary the 2 mA average current would lift the voltage by some 400 mV.
Sorry-can you explain what you mean by "DC loss" and why it is important?
Facebook-free life and Rigol-free shack.
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #6 on: January 18, 2019, 04:41:47 pm »
Very good point! I almost forgot about that.  Looking in the datasheet, there is no low limit specifically mention other than in the absolute maximum ratings: Any analog IO can○t go more than 0.3V below GND. So... center tapped transformer with a secondary termination is likely a no go.
Not at all. Simply connect the center tap to ground through a resistor. The 4mA DAC current will cause a voltage drop across the resistor , biasing the center tap above ground. For example a 100 Ohm resistor will produce a 0.4V drop so your transformer center tap will sit at 0.4V and with the max signal of 0.8V each DAC output will swing from 0V to 0.8V meeting the device compliance voltage requirements.

Quote
Of course, one could make a 400ohm to 50 ohm transformer with that sqrt(400/50) turn ratio, however these are quite difficult to get off the shelf. Hence I would probably go with something like 4:1 impedance ratio (2:1 turn ratio), therefore terminate the DAC with 2x 100ohm resistor. 
You should decide if matching is a requirement or not. Typically you cant have a match to 50 Ohm and at the same time maximize the output voltage swing. What you may consider is 100 Ohm load impedance on the transformer secondary and 4:1 transformer. This will produce 200 Ohm equivalent impedance on each DAC output. With 4mA DAC current this will maximize output voltage of 0.8V at each DAC output specified in the datasheet. That will produce 1.6V p-p accross the transformer primary on the DDS side , and 0.8V across the 100 Ohm load on the transformer secondary. To further match to 50 Ohm you can either use a buffer with appropriate input and output impedances  or design a filter with impedance transformation.
Facebook-free life and Rigol-free shack.
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #7 on: January 18, 2019, 05:36:50 pm »
Google for slaa399 application note from TI, it should answer most of your questions.
Facebook-free life and Rigol-free shack.
 

Offline YansiTopic starter

  • Super Contributor
  • ***
  • Posts: 3893
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #8 on: January 18, 2019, 06:54:10 pm »
The transformer for the differential to single ended can also do the 200 Ohms to 50 Ohms step. This would be a 2:1 transformer. With 2 inputs it would be a little different at some 2.8 : 1. One might use a little of the 200 Ohms and thus maybe 3:1 or 5:2.
I have difficulty understanding this one...can you provide a math how you arrive to 2.8:1 ?

I think that it is more than obvious, that  2.8 = SQRT(200/50), i.e. the turns ratio of the transformer.
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #9 on: January 18, 2019, 07:03:59 pm »
Ok, the usual confusion when it is not mentioned if talking about impedance ratio or turns ratio.
Facebook-free life and Rigol-free shack.
 

Offline YansiTopic starter

  • Super Contributor
  • ***
  • Posts: 3893
  • Country: 00
  • STM32, STM8, AVR, 8051
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #10 on: January 18, 2019, 08:10:58 pm »
Well I always default to the voltage ratio, as that is the typical figure when talking transformers and their transfer/transformation ratio. (Unless otherwise noted, of course).

Thanks for mentioning that application note, not yet read it carefully, unfortunately seems not much new in there. Hopefully they at least discuss the advantages and disadvanteges of various solutions, there might (not) be so obvious.
 

Offline Bud

  • Super Contributor
  • ***
  • Posts: 6877
  • Country: ca
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #11 on: January 18, 2019, 08:35:26 pm »
I designed and implemented lumped components elliptic 50 Ohm filters up to 500MHz . But this gets to accounting for components PCB pads capacitance (i used smallest pads possible and 0603 components) and i have a VNA for final tweaking on the PCB.
Facebook-free life and Rigol-free shack.
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: AD9834 DDS Output Filter Circuit Design. I am confused how to do it.
« Reply #12 on: January 18, 2019, 09:44:08 pm »
Well the ripple is (according ot the simulation with ideal parts) 60dB below, which I find good enough for such plain simple DDS with a just 10bit DAC.

But I get what you mean. The AD9954 I used has had a butterworth filter.

But why the buffer in front of the filter? I think the impedance should be well enough defined from the DDS side?

The concern about the ripple is spurs.  Those won't be 60 dB down on a 10 bit ADC.  I've not looked at the datasheet for that particular part but they will specify spur free dynamic range somewhere.

The buffer amp is more "belt and suspenders" than anything else.  It precludes not having quite enough power from the DACs to drive a long line and it gives you more control over the impedance matching.  Really depends on whether you're building a radio or a lab signal generator.  My comment was with reference to the latter case.

My design experience to date with generating RF has been analog where they are essential. 

But if you're designing a module rather than a system, the load side becomes ambiguous.  Someone correct me if I'm wrong, but my understanding is that a mismatch on the output of a passive impedance matching network will produce a mismatch at the input also and that using a buffer amp as part of the impedance matching is therefore more robust.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf