We are using the AD604 and the AD603 in our systems. We are getting a lot of signal reflections at higher gain. We do have some filters in between that are separated by amplifiers. I hope it is some simple no-brainer thing that I may have missed. The input impedance to the 604 is a 3K resistor to ground and we have a series resistor (few Ohms) before that. My suspicion is that the series resisitor or the high 3K is causing reflections (our signal is 400 kHz or so) with around 500 Ohm source impedance.
As John Poplish has pointed out 400kHz is a pretty low frequency when t comes to transmission lines - 2.5usec is about 750 metres at the speed of light in vacuum, about 500 metres in common-or-garden dielectrics,
I'd look for ringing in or feedback via the power rails. Have you put in both the ferrite beads (ferrite chips would work just as well) shown in Fig. 42 of the AD604 data sheet?
The Power Supply Rejections Ratio's for the AD603 isn't all that impressive at 40dB (see fig 22 of the data sheet) and isn't specified at all for the AD604, which would worry me.
The AD60x series consists of rather low impedance variable attenuators. It's a switched ladder architecture. AFAIK the 604 is under 200ohms. This one has a built-in preamp that you could use to isolate the younger stages from variations but IIRC the AD603 does not. I have used the AD603 a lot and that one is pretty consistently 100ohms.
Basically, the 604 and the 603 are being used as a 2 stage amplifier, the 603 is used because it has a higher signal output range that we require. Also, there is some curcuitry between these amps that run on low voltages.
I conducted some tests after you mentioned the input impedances, I found that putting a 50 Ohm in series with 100 Ohm to Ground in front of the AD603 reduces the reflections by about 50 %. If I want to exactly match the input resistance to the 603, what should I do? The input from the 603 is coming in from a standard opamp with Av=1. I unfortunately cannot post a schematic because of the proprietary nature of some of the product.
Also, we put ferrites on all the input power lines, that seems to have reduced things a little bit. And we noticed that changing the power supplies (we put an older, inferior unit just to see what it would do) causes increase in these reflections as well as a strange phenomenon: at the switching frequency of the power supply, we see blips in the signal path.Sometimes, reflections also accompany this blip. So, it looks like the 603 is a little sensitive to bad power supplies.
That looks like you have some major instability at your hands with this circuit. Cascading an AD603 and an AD604 provides a whole lot of gain up to tens of MHz. This may sound silly but it's worth a try. Take a piece of metal (tape it up to avoid frying anything) and wave it closely over the amps including the input buffer. Look at the noise levels, whether they change. If they do then the setup is most likely 'over the edge' in terms of stability.
If the AD603 is fed from a fast opamp and they are physically close together there should be no need for matching. Make sure you have excellent bypassing on the supply pins and on the gain controls. However, I have the impression that a noise expert should look at your circuit and most of all at the layout.
If you want to rule out the power supply there is an old trick: Bundle up some beefy batteries such as D-cells and use these to supply power. If you see blips from the switcher this means that the rail is way too noisy. Hook up a spectrum analyzer to the rails and see what's there. But be careful, provide a DC short towards the analyzer and never, never switch the supplies on or off or short out anything while the coax from there is connected to the analyzer. It could fry its input.
Then do the same for the gain control inputs, to see whether the analyzer shows any switcher noise there. The supplies of the DAC or anything else that's in the path to drive these need to be as clean as those of the AD603 and AD604.
I doubt that you are having impedance matching problems at 400 kHz unless you have a 1000s of meters of mighty spiffy cable on the input. The AD600 series of variable gain amplifiers have HORRIBLE power supply rejection. Let me say that again, they have absolutely HORRIBLE PSRR!!!! You need to realize that these parts have significant bandwidth where the power supply rejection is virtually nil. You need to put healthy power supply decoupling on EACH power supply pin. This could include ferrite beads, resistors, and/or tiny linear voltage regulators. Don't forget your bypass capacitors with short leads. A ground plane also helps. The schematics in the data sheet are misleading regarding how to decouple the power supply pins. Use a separate decoupling network for each power supply pin. If you don't, you will have a nice oscillator. Since your at such low frequencies, you can also limit the bandwidth of the amplifier chain by using a RC lowpass filter at the inputs. Highly recommended.
Thank you all. We already put a low pass filter at the output of the
603's and this takes out a lot of high frequency radiation (for this application, high frequency is anything above 10 MHz).
Next, we ran tests at each point in the signal path and we found that at teh input of the 603, there were absolutely no reflections, but at the output, there were a lot - it looks like a million replicas of the signal, (very small). Now the 603 was in the 10 to 50 dB configuration, which means, its output does not have any feedback connection (the other two configurations mentioned in the datasheet have a 2.15K resistor or a short to pin 6 (which is the feedback pin). I am wondering if this could have anythign to do with the problem. Also, this goes to the next stage with only a decoupling capacitor in between. Do we need some sort of a resistive load at the output of the
603 or a series resistor that can absorb these reflections?
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.