EMRFD Message Archive 8004

Message Date From Subject
8004 2012-12-29 06:44:49 Nick Tsakonas DDS module with buffer / amp

today I faced for once more a problem that I cannot understand and I always solve using trial and error. I would appreciate any help to understand the basics of it.

The case where I stuck is this: For an experiment of mine, I am trying to drive a homemade diode DBM using a cheap AD9851 DDS module from ebay. I built a buffer/amp in order to increase the drive level (the DDS module gives only 250mv peak). Just for the story, the dds module is supposed to have 200 Ohm output impedance and the buffer was taken from here (http://www.pongrance.com/dds-buffer.html). I just replaced the transistors with 2n3904. I simulated the buffer in LTSpice and works excellent with 200 ohm input and 50 Ohm output.

What happens is this:The DDS module has a DC offset of around 0.5V so I added a 100nF blocking capacitor at the buffer input. The DDS signal at the input as well as the output was very distorted.The output was like a rectified signal. After a lot of trial and error I reached a point where I have at the buffer input a 200 Ohm 15db PI attenuator followed by a 10pf blocking capacitor. The output is very clean now. Changing the input capacitor changes the output level proportionally.

The trials to arrive to this result were made on the basis that there must be a mismatch at the input.If this is the case, I unfortunately cannot understand how the input capacitor affects this. Before adding the pi attenuator, changing only the capacitor, created an effect where high capacitance (e.g. 33p) was giving clean output up to 5Mhz and low capacitance (e.g. 10pf) was giving clean output above 10 Mhz. Adding the attenuator stabilized the output.

What is happening underneath? Could you point me to the right direction ? I need to understand this point.

Thanks a lot,
Nick, SV1DJG

[Non-text portions of this message have been removed]
8005 2012-12-29 07:16:24 bobtbobbo Re: DDS module with buffer / amp
Hi Nick: I used that same circuit for the same purpose, although with a different DDS. I also put a 100nf cap at the input. I used PN2222As.
I put a scope on the output and found it distorted at 12 volts. I put a variable supply on it and found increasing gain with linearity up to 9 volts on the circuit. With my DDS at the input I got a clean +10dBM out into 50 ohms with 9 volts
8006 2012-12-29 16:29:17 kb1gmx Re: DDS module with buffer / amp
Used that circuit a lot and the usual areas of trouble are..

It has a lot of gain and too much input will overdrive it.
If the input is sufficiently over driven the amp can stop working.

The other is the input must NOT be a DC connection A capacitor is required (or is part of the prior stages output) there failure to do so will upset the bias and output waveform.

Both PN2222 and 2n3904 work but I find 2n2222A a better device or
even higher FT devices if I'm using that at above 10-20mhz.

In any high gain high frequency amp layout counts, shot leads and
solid ground (deadbug/ugly works well) are a must.


8007 2012-12-29 22:10:55 victor Re: DDS module with buffer / amp
A short analysis of the circuit:
Assuming the amplifier is functional, then Q1 base voltage is 0.6V.
Going back through the R3, R2 voltage divider then at Q2 emitter there is 6.6V.
Q2 base voltage is 0.6V higher so it is 7.2V. Voltage across R1 is then 12-7.2=~5V.
So Q1 current is 5mA. If so its Re (emitter dynamic resistance) is Re=25/5=5 Ohm. Q1 voltage gain is approx its collector load impedance divided by its dynamic emitter resistance hence 1000/5=200.
Q2 voltage gain is approx 1 (emitter follower) so open loop gain of the circuit is around 200.
The circuit input impedance is approx R3 (which is the feedback resistor) divided by the open loop gain: 1000/200=5 Ohm which is a low impedance, certainly much lower than 50 Ohm.
So we have an amplifier with low input and output impedance with closed loop gain of 11 (Which comes from R3/R2 + 1).
This analysis is good for low frequencies. At higher frequency things change because the transistors gain gets lower but it still is a good start to understand the circuit.
Adding at the input a coupling capacitor (a big one: 100nF) in series with a resistor will give you the possibility to control the gain of the circuit. Try putting there a 100 Ohm trimmer in series with the input and see how adjusting it will influence the circuit behaviour.
Victor - 4Z4ME

8008 2012-12-30 04:42:56 Nick Tsakonas Re: DDS module with buffer / amp
Allison, Victor

Thanks for the reply. Victor, your analysis is impressive, the values from the simulation are very close to yours!

I ran the simulation for the input/output impedance of the circuit and somehow I thought I got input impedance of 200 Ohm (probably wrong). I re-ran it today and it is indeed in the range of 5 - 25 Ohm on the HF range.

It also seems that the input should be less than 50mV Peak to Peak to avoid distorted output.
I noticed on the simulation that when I add a 100n capacitor at the input, even with 50mv drive the input and output starts to become distorted (On higher levels I get exactly the same distortion I see on my scope).

The input capacitor seems to modify the input impedance of the circuit thus giving better or worse match to the DDS filter, and also affects the actual  drive level, right? Would you say that what I was experiencing was mainly a mismatch case with all the collateral effects?


[Non-text portions of this message have been removed]
8009 2012-12-30 05:51:34 victor Re: DDS module with buffer / amp
I don't think that there is a case for distortion caused by impedance mismatch.
First, I think that using the input coupling capacitor as a output signal level control is not good because when you use a small capacitor you might attenuate the fundamental frequency but the higher harmonics of the input signal pass with lower attenuation and so help to distort the output signal. Using a resistor in series with a 100nF capacitor is much better.
Second, it is more reasonably to guess that the signal distorti
8010 2012-12-30 06:22:24 kb1gmx Re: DDS module with buffer / amp
There is a small argument for mismatch at the input as the
DDS output filter expects a certain load impedance, if not met the filter shape and the output levels would be different.

So matching impedance and not over driving the input are both


8029 2013-01-03 23:01:50 w5jhjerry Re: DDS module with buffer / amp
Hi Nick,
I know this is a late reply but I just joined the group recently. I added a buffer to my 9834 DDS VFO. I chose a 50 ohm input/output broadband amp originally designed by Wes Hayward, W7ZOI. I used a 2N5109 as the transistor to give me a very broad response. With a 2N5109, the amp should work well from 2-50MHZ. The circuit is shown on page 2.20 of my very old copy of EMRFD. This amp will produce 20-21db gain. On the input, I used a 2:1 turns ratio transformer to match the 200 ohms to 50 ohms. The transformer can be the bifilar type or just a standard transformer wound on a FT50-43 or FT50-75 core (I used a 12 turn primary and a 6 turn secondary on a FT50-43 core). I have also used this same amp
8030 2013-01-04 03:07:37 Nick Tsakonas Re: DDS module with buffer / amp
Hi Jerry,

thanks for the reply. It is not at all late reply... I am still working on it, I just hadn't got the time to give a feedback. I tried Victor's  recipe: I lowered the emiter resistor to 220 and then to 150 Ohm and added a 100n capacitor with a trimmer at the input. Indeed, I can control the output better and I can get (undistorted)  around 500mv Peak on 50Ohm load. If I adjust the output for at the lowest frequency (around 4Mhz) it is also OK for the upper end (around 25Mhz), the reverse is not true.

However, last night I had exactly the same idea, to try the "W7ZOI amp"...I also have an unbuilt kit from "kits and parts" of what I think is a similar amp which I wanted to try. And I will, since I would like to reach 1V peak for the range 4-20Mhz.

I will try it and report back soon.


8031 2013-01-04 21:30:53 Gene Dorcas Re: DDS module with buffer / amp
Here’s what I use. The gain is a little less than the bipolar transistor version I’ve seen used but the input impedance is set with a single resistor so is easily adjusted.


Notice the input goes thru the coupling capacitor to the input of one of the two op-amps. A 51 ohm resistor is connected to that point and its other end is connected to ground for RF thru a 0.1uf cap. The high input impedance of the op-amp might bring the 51 ohms down just a little.

The point is that you can change the input impedance by simply changing that resistor.

The bad news is that most distributors don’t have any Lt1253 in the 8pin dip in stock and Linear Tech isn’t going to make any more ‘til next Summer. You can get it in the SOIC pkg from most distributors. If anyone wants to try it out I can solder up some SOIC (SMD) LT1253 on a 8pin dip header so it is installed just like it was an 8-pin DIP. I’m set up here for SMD with a hot-air/soldering iron SMD station along with a stereo lab microscope.

I’m interested in seeing how everyone is using the DDS in various apps. I have a N3ZI DDS2 installed in an enclosure with a 2n3866 power amp driving the 50ohm output. It’s convenient to have such a lab instrument. With the 50ohm output I can, of course, use my 50ohm step attenuator to get accurate low-level signals that I need for building receivers.

Well, I’m starting to ramble so I’d better close this and get back to work.

Gene, W5DOR

8032 2013-01-05 07:10:34 Todd F. Carney Re: DDS module with buffer / amp
Gene and all,

the LT1253CN 8-pin dip version is available here:



K7TFC / Medford, Oregon, USA / CN82ni / UTC-8
QRP (CW & SSB) / EmComm / SOTA / Homebrew / Design

8038 2013-01-05 12:29:06 Gene Dorcas Re: DDS module with buffer / amp
Thanks Todd. I immediately ordered a few.


8039 2013-01-05 12:29:12 Jim Re: DDS module with buffer / amp
I have to express my reluctance to buy stuff like this from China, without any provenance whatsoever.
If the vendor were ISO-certified and could show chain-of-custody from the manufacturer, it might be different ... but I have yet to see Digi-Key or Mouser open a shipping warehouse in Shenzhen.
Meantime, a better solution may be to populate the currently available SOIC package onto a DIP carrier, or get something like a Schmartboard for it.
At my age, life is too short to deal with defective or counterfeit semiconductors from China, bought off of ebay.

8041 2013-01-05 18:23:06 Paul Re: DDS module with buffer / amp
I admire your attitude.

I am repairing a radio and find the driver transistors are defective. From the radio manufacturer they are $115 each and I need 2.

China via eBay they are $25 a pair – shipped.

Decisions, decisions.

de Paul, W8AEF

8042 2013-01-05 18:23:10 Todd F. Carney Re: DDS module with buffer / amp
8043 2013-01-05 18:23:12 w5jhjerry Re: DDS module with buffer / amp
Hi Nick,
Another amp you may want to investigate is the VFO buffer designed by W7EL and used in his "The Qptimized QRP Transceiver for 7 MHz." The amp uses two transistors the first one being an emitter follower and the second is a common emitter circuit. The circuit is a bit different in that the output transformer has to have a high impedance otherwise you get distortion. The primary side of the transformer has an impedance of 6300 ohms at 7MHz and the secondary has an impedance of 250 ohms. A transformer made from a FT50-75 core could be used to achieve the high impedance. The high impedance transformer is discussed in EMRFD.
Jerry W5JH
"building something without experimenting is just solder practice"

8044 2013-01-05 18:23:16 Gene Dorcas Re: DDS module with buffer / amp
I thought of that. I'll see how they work and let the group know. I
already have the LT1253 in the SOIC pkg and mounted on a 8-pin dip header so
it plugs into my 8-pin dip socket just like it was actually a p-dip device.

I'll try 'em both out and let ya know. I know some of the Chinese products
are not to be trusted for quality.

Oh well, Onward and upward. We'll see how it goes.


Gene, W5DOR

8049 2013-01-06 08:23:33 John Levreault Breadboarding with SM packages
8706 2013-06-01 03:17:34 drmail377 Re: DDS module with buffer / amp

I just brought-up one of the $4.50 ebay Chinese AD9850 boards with the Parallax Propeller. It works fine and has the right output filter cutoff frequency. But unfortunately the schematic the seller provided was wrong. I had to reverse-engineer the board a bit to fully sort out some of the pins.

I was looking around for an output buffer so I can use the 1Vpp sine output to drive a Level-7 DBM and ran across this thread.

Have a look at the new (April 2013) LTC6957 Low Phase Noise, Dual Output Buffer/Driver/Logic Converter from Linear Technology.

The LTC6957 is specifically optimized for conversion of sine waves to logic levels with minimal additive phase noise.

If I'm not mistaken, ideally we want to drive a diode DBM with a 50% duty-cycle square wave. The LTC6957 has very low additive jitter: 45fsrms over 12KHz to 20MHz and 123fsrms at 10Hz to 150MHz for the LVPECL version. The -3 (in-phase output) and -4 (complementary output) CMOS versions have around -159dBc/Hz additive noise density level at 10KHz, -155dBc/Hz at 1KHz. The LVPECL version is even lower. Unless you're throwing a lot of time and/or money at the AD985X reference oscillator, this level of additive phase noise is likely not going to matter.

For comparison, if you used the AD9850's on-chip comparator to convert the output sine wave to a square wave, the AD9850 comparator's additive jitter is a whopping 80ps typical (but do note, the bandwidth is not specified in the data sheet). Therefore, the AD9850's on-die comparator is worse than the LTC6957 by several orders of magnitude!

So, I guess this is why nobody is using the AD9850/51's on-chip comparator to square-up the output(?)

* Let's look at the input...

The LTC6957 CMOS output version maximum input frequency is 300MHz. There are four user pin selectable input filters with -3dB frequencies of 50, 160, 500, and 1200 MHz. The selection of the filter bandwidth minimizes additive jitter (phase noise) depending on the characteristics of the input signal. (See the data sheet for more on this.)

The single and double-ended input range is 0.2V minimum, 0.8V typical, 2.0V maximum, so it should work with the eBay AD9850/1 Chinese boards which typically have 1Vpp sine out (at Vcc=5.0V) after the 7th order Cauer (elliptical) filter. But 1Vpp is pretty close to the typical 0.8V specification. The data sheet shows examples where the input is +7dBm at 50Ohms input, that wold be 1.42Vpp. The input resistance is 2KOhm typical. Consider the Cauer Nyquist filter on the Chinese AD98XX boards have 200 Ohm termination resistors on the inputs and outputs.

* What about the output?

The LTC6957-4 has dual CMOS complementary outputs, each capable of driving +/- 30mA. The supply range is 3.15-3.45VDC. The CMOS output swings within 0.1-0.2V of the rails. Assuming we're running at Vdd=3.3V and the CMOS version is driving a 50 Ohm Level-7 (+7dBm L.O. input level) diode DBM, a +7dBm sine wave into 50 Ohms is 1.42Vpp, so the output of around 3Vpp will be no problem. At +7dBm into 50 Ohms a sine wave supplies 10mArms. Considering the RMS voltage of a square wave is its peak voltage and looking at the output current capability of the LTC6957-4, it looks like the CMOS outputs can handle the current without breaking a sweat.

Since the AD9850/51 DDS's don't have a quadrature output, the dual outputs of LTC6957 may be useful if one adds a 90 or +/-45 degree phase shift to one or both of the outputs for quadrature testing over a relatively narrow frequency range.

* Can you buy this part?

The LTC6957 (all versions) are in stock at Digi-Key (Digi-Key only as far as Octopart knows), but it is $7.69 USD each (ouch) in unit quantity.

There are Linear demo boards available for this part. The demo board for the LTC6957-4 is Linear P/N DC1766A-B. Digi-Key has them in-stock for $75.38 USD each (ouch again!) But the good thing is that the demo board documents are free to download so you can look at the reference layout.

* Do I work for LTC

No; no affiliation whatsoever...

* Concluding Remarks...

As for external components required. It looks like the LTC6957 needs nothing more than your usual bypass and coupling cap(s) and/or transformer(s) - depending on your application of-course. For the AD9850/51, one or two input coupling caps, an output coupling cap and a bypass cap looks like all that's needed.

I don't see a simulation model available for the LTC6957 on the Linear site, or in my copy of LTSpice-IV after a sync update today.

Did I miss anything? Anyone see something wrong with this part (other than the fixed 3.3V supply and the price) in this AD95XX buffer application?

If you made it this far, thanks for taking the time to read this post.

Best Regards, David
8708 2013-06-03 11:57:48 Gary, WB9JPS Re: DDS module with buffer / amp
We normally back-terminate the driver to absorb back-propagating RF and IF port energy, which may otherwise re-mix. So then voltage swing at the mixer is divided by two, but apparently still ok with the 3.3V CMOS driver.

My favorite DDS output amp: LMH6703. Good for all kinds of higher-level RF work, makes a great little low-distortion gain block. $3.25. Eval boards from TI are $10.


-Gary, WB9JPS

> * What about the output?
> The LTC6957-4 has dual CMOS complementary outputs, each capable of driving +/- 30mA. The supply range is 3.15-3.45VDC. The CMOS output swings within 0.1-0.2V of the rails. Assuming we're running at Vdd=3.3V and the CMOS version is driving a 50 Ohm Level-7 (+7dBm L.O. input level) diode DBM, a +7dBm sine wave into 50 Ohms is 1.42Vpp, so the output of around 3Vpp will be no problem. At +7dBm into 50 Ohms a sine wave supplies 10mArms. Considering the RMS voltage of a square wave is its peak voltage and looking at the output current capability of the LTC6957-4, it looks like the CMOS outputs can handle the current without breaking a sweat.
8746 2013-06-19 06:34:25 drmail377 Re: DDS module with buffer / amp
Hi Gary,

Good point on the 'back' or 'double' termination point. Thanks.

I've been looking at the LMH6703, it is a nice part from the spec sheet. Also, at first glance it seems TI has done a decent job of migrating the IP (Intellectual Property, e.g. data-sheets, app-notes, sim models, etc.) over from National since TI took over National Semiconductor.

Anyway; I now have a custom LMH6703 LTSpice-IV easy to use .asy symbol with the TI PSPice models for this part working as per the paper you referenced.

But as I expected - in quick simulations moving this part from a dual-rail example as in IQPRo paper you referenced to single rail is problematic, especially for the single-rail non-inverting configuration shown in the IQPro paper.

That's not surprising - in my (limited) experience these current-mode parts require care when feeding. But simulation and realiztion mate up fairly well at HF. But a working example can save a lot of time...

So - can you point me to a working example schematic of the LMH6703 buffering an AD9850/51 or similar DDS? Especially when taken to single rail supply and driving at least a level-7 mixer? Nothing special about the signal chain. Straight-forward single conversion (DC) using a DBM; no NE602's or the like. Just a first simple realizati
8748 2013-06-19 13:03:43 Gary, WB9JPS Re: DDS module with buffer / amp
I would only trust the model for low-frequency behavior. But then, I'm a crusty old protoge of Bob Pease, a fellow analog engineer who only reluctantly trusts Spice...

The schematic in my report referred to in this thread is exactly what I used as the buffer at the output of my IQPro, after the output balun. DO use the balun (transformer) output for lowest distortion. Gain can be tweaked, of course, by changing the resistor to ground. I had bipolar supplies available, so that makes things simpler. Use the highest supply voltages you can to maximize output swing and minimize distortion. The eval boards work perfectly and are cheap and easy to hack. If you're doing your own boards, try to clone that design.

As for single-supply operation, it certainly adds components and a small degree of risk. In my report, the LT1253 was configured that way and worked well enough. I don't think the 6703 eval board supported single-supply configurations, but I'm not sure. Anyway, a good 12 V supply will yield a driver with back-terminati
8750 2013-06-19 22:20:24 Dana Myers SPICE (was Re: [emrfd] Re: DDS module with buffer / amp)
8751 2013-06-19 22:20:32 augustinetez Re: DDS module with buffer / amp
Based on Gary's work, I have an LMH6703 amp design running on single supply. I'm currently working to finish the ALC part of my design but can get +10dBm from 2MHz to around 40MHz at this stage.

There are several things I found when working on this. Keep the ground plane away from the LMH6703 signal paths (even small amounts of capacitive coupling can upset the signal) and don't coat the board with anything (solder mask, kept clear of the LMH6703 should be OK).

You can see the beginnings of it on my site at here:


under the entry for 19 November, about half way down the page.

I can send you more up to date details if you are interested, some component values have changed and I am using two chips in my design.

Terry VK5TM

8758 2013-06-25 20:13:58 augustinetez Re: DDS module with buffer / amp
In several posts on this site and various other places, I have seen people stating the output impedance of the cheap Chinese DDS
modules as 200 Ohms.

They are in fact 100 Ohms. There is a 200 Ohm resistor directly at the output of the DDS chip and another at the output of the filter. The two in parallel set the output impedance to 100 Ohms.

You will notice that the complementary resistor on the other Sine output is 100 Ohms.

I hope this helps those people having problems interfacing buffer amps to these modules, as it caused me some problems before I had a proper look at the module schematic.

Terry VK5TM
8763 2013-06-27 09:31:34 drmail377 Re: DDS module with buffer / amp
Hi Terry, I sent you an Email. I'd like to see your latest developments. And thanks for the follow-