RF ALC circuits

[Wendymott]

Hi peeps. I need your help please.

In a previous thread I was attempting to utilise a DDS oscillator AD9951 to make an antenna analyser, with the same functions as as the MFJ 269. The idea of using the AD9951 was to allow tuning in one continuous operation rather than switching ranges. This works quite well, until I want to measure the RF Impedance. Unfortunately the DDS output varies about 4 db over the useful range of 1 - 150 Mhz. This compromises the voltage applied to the bridge resistors and thus the measured voltage , giving me unreliable impedance readings.
I could forget about the impedance measurement altogether, some of the earlier MFJ's dont use this function, but it seems a pity not to add it if I can.
Looking at the MFJ circuits they sample the RF out and rectify and feed back to the amplitude control FET in the emitters of the L/C oscillator.
Obviously this works well, I have made a 4 pin diode attenuator with a 40 db operating range 12 to 1V dc control voltage, the RF sampling circuit works ok too. However with the pin diode atten working "backwards" i.e max volts = min attenuation, it creates problems where I have to bias on the pin attenuator to get enough RF to supply the sampling circuit.
I have tried various configurations, but at the moment I am struggling..
I wonder if any of you learned peeps have an "off shelf" schematic that may help. I am trying to keep it simple. I really dont want to use dual power supplies.
I do realise that the 5 pole filter on the output of the AD9951 is a key factor, but I was hoping to "get away" with an ALC instead.
I did look back at previous postings re ALC's but nothing of use pops up.

[Terry_VK5TM]

Wendy, I've been playing with this problem for a while now.

I have had 'some' success with various options like VGA's, none have been really useful over a large frequency span.

Just this last week, I came across an article in VHF Comms magazine (issue 4 2003) for a DDS signal generator that has a feedback mechanism involving the Rset pin of the DDS.

Claimed output variation was 0.3dB from 100kHz to 65MHz.

While this particular article is using the AD9851, I can't see why it shouldn't be adaptable to the '9951.

I had tried this before with not very promising results, but that was using the output of the DDS directly, the above article amplifies the output first.

[MrBungle]

Definitely RSET. See the AN here: http://www.analog.com/media/en/techn...tes/AN-423.pdf

I am looking at the same thing as I need an accurate signal generator and no one is selling a 2019A at a price I can afford. I was thinking of sampling a portion of the output power with a power splitter and AD8307 and feeding that back into RSET via a control loop. I can then amplify the output again and stick a stepped attenuator on the output. Mini 2019A then!

Edit: note I picked the AD9850 for the design I am working on because of that pin.

[Terry_VK5TM]

The majority of AD DDS chips have that pin, it's just called something different in later chips (DAC_Rset in the 9951).

[jimmc101]

This might also be of interest regarding the use of the Rset pin (on AD9850).
I would assume the AD9951 uses a similar arrangement.

I've been trying AM on a Chinese AD9850 module by connecting a second 3k9 resistor to the Rset pin.
With 0v on the other end of this resistor the output is +6dB and with 2Vref (~2.5v) the output is zero.
Up to about 80% mod the device is quite linear, much less than 1% distortion.

Jim

[Wendymott]

Hi peeps... many thanks for the replies... I think the reset pin can also be used for AM Modulaton, using a balanced transformer on the output to get symetrical modulation.
At this time the only "firm" schematic is the Pin diode attenuator... I the other part is the amp / feedback. VR1 sets the centre start point to allow the atten to pass enough RF.#
I will wait for Terry's schematic before I carry on.

[Julesomega]

The other way would be to 'Normalize' the analyser by storing the plot from sweeper to detector and subtracting the test results.

[Radio Wrangler]

You need to do normalisation to account for coupler flatness anyway.

Full monty VNAs have a levelling loop flattening the source, and then they run multiple receivers at the same time. One receiver monitors the drive from the forwards coupler, another measures the reflection from the DUT input, and a third measures the output from the DUT. This will get you S11 and S21 with high precision, then transfer switches are used to reverse the DUT to get S22 and S12.

This seems a lot more elaborate than you'd do for a home unit, but it's useful to know where the road leads.

David

[Argus25]

It is interesting to look at the way Tek did this for their leveled SG503 sine wave generator that goes from about 250kHz to 250MHz. And it is really well leveled.

They sample the output with a peak to detector with two diodes and feed that into one input of an OP amp, but the clever part, into the other OP amp input they feed that with a bias circuit with two extra identical diodes that temperature compensate the two in the detector.

In their case they use the OP amp output to control the emitter current of the oscillator transistor to control the amplitude that way.

(By breaking that loop and driving the OP amp with audio I was able to easily convert an SG503 into a variable frequency pantry transmitter).

If your circuit doesn't have easy amplitude control (but it sounds like it does) you could always pass the signal through a 4 quadrant multiplier IC to make an amplitude modulator and drive that with the OP amp's output.

[Terry_VK5TM]

One thing I did find, is that if the output is not well filtered (before being picked off for the levelling circuit), the alias and images in the signal play havoc with the levelling circuit, so this would probably upset an AD8307 based circuit too.

For the AD9850 modules, it might be better to bypass the on board filter and build an outboard one (and use both sine outputs - transformer coupled as in AN 423).

I ordered some 1:1 transformer samples (WBC1-1TLB) from Coilcraft yesterday to try this out.

[Radio Wrangler]

Very little has been written on amplitude control in oscillators, compared to the amount on frequency control and stability. Using a detector and feedback loop to control the quiescent current of the oscillator device has been used quite often in high-class gear.

Simple oscillators rely on the amplitude pushing the device into cutoff over part of the cycle, with inevitable effects on the harmonic content of the output. Wide range oscillators like the Tek constant-amplitude signal generator can't rely on output filtering to clean off harmonics without the complexity of multiband tracking filters, or a bank of sub-octave filters (HP 608 and HP 8640 respectively). Feedback control of the oscillation level allows the oscillator device to stay in class-A to the advantage of the output waveform. That constant amplitude sig gen was heavily used to calibrate scopes (even at HP!), and just like Terry found with the aliases of a DDS, the harmonics would have spoiled the flatness of the levelled output.

To apply AM, you only need a 2-quadrant multiplier, a 4-quadrant one gives DSB unless you bias it away from the centre so it only ever gets into 2 of its quadrants... the offset bias can also be looked upon as the DC component which is mixed upwards to become the carrier.

Aviators still use AM to chat with the ground. Since the changeover from 25kHz to 8.33..kHz channel spacing, the specs for the transmitters have been tightened a lot. The modulated envelope now has to fit inside a spectrum template which allows only little distortion in the modulation process, so transmitters are normally modulated within ALC loops. You also have to be able to do over 70% mod at the same time, which keeps it challenging.

Having duplicate detector circuits just to create a DC reference with the same temperature effects as the main detector is an old and respected technique. The good old 432A/478A thermistor bridge power meter has a complete second thermistor bridge and balancing amplifier to temperature compensate the one measuring the RF power.

David

[Argus25]

Originally Posted by Radio Wrangler

To apply AM, you only need a 2-quadrant multiplier, a 4-quadrant one gives DSB unless you bias it away from the centre

Yes, true. It is interesting though that for their amplitude modulators, Philips in their TV pattern generators used 4 quadrant multiplier IC's with an offset bias, for example in their TV pattern generator VHF/UHF modulators. I often wondered why, perhaps the linearity is is better with these balanced circuits at high modulation levels.

[Radio Wrangler]

I think more development went into the 4-quadrant chips. The extra balance inherent in being doubly balanced is destroyed by the deliberate offset needed for AM.

The full Gilbert cell with its current mode operation and linearising diodes, compared to the MC1496...

David

[Argus25]

Originally Posted by Radio Wrangler

The extra balance inherent in being doubly balanced is destroyed by the deliberate offset needed for AM.

I don't think is destroyed, just not utilized.

One thing about the double balanced configuration (eg MC1496), the modulation (introduced to the low level inputs) is cancelled from the output signal (even with the offset) which means, (compared to a differential amplifier with modulated emitter currents) that no HPF is required at the output and the signal, if required, can be taken from one of the outputs single ended.

Hugo.

[Wendymott]

Hi Peeps. I have obtained some measure of amplitude control with the attached circuit, sent to me by member Terry VK5TM, which was from an article by DJ8ES. The original article showed a 17 pole output filter from the AD9850 DDS Chip, I am sure this has contributed to the claimed near flat response .. I obtained approx 2db overall in the range 2 - 160 Mhz with a 7 pole filter. In my application a 17 pole filter seems a bit "overkill".
I attach the schematic and a pcb screenshot.

[Terry_VK5TM]

I went to Miguels (PY2OH) site today and it looks like he has found the flaw in my original attempt at using Rset for level control (you will need to translate it from Portugese):

http://py2ohh.w2c.com.br/trx/alc4ad9850/alc4ad9850.html

[Wendymott]

I have concluded that my requirements far out reach my ability.... well with this project anyway. I tried the circuit above and found it not suitable...it didnt work.
I have also concluded that a DDS is not suitable to drive an antenna analyser.. reasons ??
I managed to get the ALC to level the output to 1.5db over the 7 - 150 Mhz range..but when connected to a 2 metre Slim Jim aerial, the DDS driver was showing 138 Mhz at maximum dip, compared with my Marconi 2022 and a borrowed MFJ269 which both dipped at 144 Mhz. The amplitude of the dtrive signal in all cases was comparable.
I am going to have a "reclaim" day tomorrow, and look at a L/C oscillator with ALC.