Integrated AGC 100kHz-10MHz

Hammy a écrit :

Use a better output opamp (yours must be... well...) and adjust the circuit response so that you're within the specs. Oh, is your DAC up to the task too?

--
Thanks,
Fred.

On a sunny day (Sun, 29 Nov 2009 08:56:06 -0500) it happened Hammy wrote in :

If the DAC is specified for that frequency, then, if you have control over the digital part, could be that soft needs improving? If the DAC is not specified for at least 10MHz get a better DAC. AGC circuits, the very simple ones with a couple of FETs and or opamp / diodes and a capacitor to get rid of the DC bias, seem a much more complicated solution, introduces distortion too.

Fred Bartoli a écrit :

Also check your sampling frequency (DAC sinc response). But the 500mV offset surely indicates your opamp choice is bad.

--
Thanks,
Fred.

It's not an op-amp it's the DAC on an AD9834 DDS IC. I want to feed the signal to an LPF then to an AD8045 for final amplification. I want to use the AD8045 for variable gain selectable by a POT in the feedback loop. This is easier done using a consistent input signal. I think I may have found something AD8330ACPZ.

I'm using the 50MHz part when I use the 75MHz part it probably wont be as bad.

I cant find any specs in the datasheet about offset vs frequency. It is possible I damaged the part inadvertently while testing on the breadbaord;)

Anybody use AD's DDS IC's is this typical?

It seems that this is normal for these parts. This guy is using an opamp (IC6 OPA843) to adjust the output of the DDS before the LPF.

SEE

I want to be able to use an AGC Wideband Op-Amp there instead.

How accurate does it have to be? You can use AC coupling to get rid of the offset, and a 1 of 8 analog multiplexer with some resistor divider to split the range into 8 gain stages. Control the mux (74HC5051 ?) from the same micro that you use to control the DDS.

2 muxes ->16 ranges.... snall part count.

Yes, but it's expensive:

If on a budget you could build one around a uA733 or similar chip, add a simple detector and a FET as the gain control element. If your budget is super-tight use a dual-gate MOSFET which allows the gain control range you are looking for. One gate gets the signal, the other the control voltage.

Here is another trick to save a buck: Characterize the signal drop versus frequency. Then use a simple gain control stage and place a 2nd DAC. This 2nd DAC is used to set the gain control input and is updated whenever you update the frequency.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Years ago, I used a temperature-driven look-up table, controlling a varicap, to make crystal oscillators as stable as oven-controlled... without the power requirements of an oven. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

BTDT. The problem is there are many more of unstable parts in the control loop, then in case of simple OCXO. Even if the compensation is initially perfect, it will degrade with aging.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

If sounds like you have a cos/sin of your output shifting amplitude. This would thus be the same as changing the duty cycle and causing the offset you're seeing. I wonder if you have a unity stage for the DAC to prevent excess loading, do to load capacitance ?

This all maybe a shot in the dark but it's my shot :)

With a 50MHz clock the signal would normally drop to maybe 500mV at 10MHz but that's assuming you're measuring using a 10:1 scope probe with say 13pF tip capacitance. The 'FSadjust' and 'Rset' can be used for AGC. Put a DC 'level set' voltage on the Rser resistor end that normally connects to 0V. (OV=max out). Idea is to throttle the output swing to (say) 1/2 of normal, which then allows a

+50% increase to compensate for losses.

What isn't ;)

I was thinking of putting a digital pot in the AD8045 feedback loop. But by the time all is said and done 6 Bucks for the AD8637 looks pretty good.Newark sent me out a 25% off flyer so it's a good time to buy parts.

I'm using an AD9834 DDS IC to build my own signal generator controlled from the serial port. The parts pretty impressive for $7 sine, square ,triangle output can perform various modulation with minimal parts ,ability to sweep frequency range defined step size and rate and a

75MHz clock. I know AD make much faster ones but a 10MHz output max is good for me.

Trying to keep it in the $35 to $50 range. I was going to buy a cheap $150 - $250 Chinese one but decided to build my own.

Thanks John.

I'll try playing around with that.

You did WHAT? A LUT? Meaning digital stuff?

Now go, stand in the corner, and be ashamed. Tsk, tsk, tsk ...

--
SCNR, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

The client, Saunders and Associates (builders of crystal testing equipment) did the LUT. I did the analog and A/D/A interfaces. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

Good. Your are redeemed. Welcome back to the analog club :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Careful with digital pots. They are essentially like flash drives. After a certain amount of write cycles they could become unreliable. Just keep that in mind when you are thinking about doing sweeps and stuff.

Hmm, 25% off, why don't they carry good beer?

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

the

to

10MHz=20 13pF=20

voltage=20

Idea is=20

a=20

With all that has been said, i suspect a miswire or some circuit problem. Both the DSS and the first OPA should be up to the task (without the stated aberrations). Though it could very well be a layout issue as well. Since the DSS is current output are you using an appropriate termination and coupling to the OPA?

Actually I'm embarrassed to say I had my probe on 1x (I know stupid). I really have to stop doing that it's getting ridiculous.

I'm just working on the power supply now.

On the plus side I learnt a lot about AGC ccts and what's out there for future reference.

Thanks to all who helped.