Filter impedance matching

Go To Last Post
20 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I'm making an Arbitrary Waveform Generator and need some advice with impedance matching after the output filter.
Analog stuff really isn't my area. :-)

I have a 7 order low-pass with a terminating impedance of 50 Ohm. Filter is fed by an op-amp.
Simulating shows a nice curve with a cutoff as expected.
Removing the terminating 50R or adding another in parallel obviously make the curve go all bonkers.

The question is now - should I drop the (internal) 50R load and require that the output is connected to nice 50R inputs through nice and shiny 50R cables, or add an internal 50R load, and thereby allowing more freedom for the output connection?

Two factors:
* I've seen that terminating with 25R looks way better than no termination at all.
* As this is not an instrument that mostly will be used in nice labs with nice cables it will more often than not be connected to high impedance inputs.

The above seems to indicate I should terminate internally with 50R.

Obviously, I could also have a switch that connects/disconnect that 50R load res.

Are there other/better ways to handle this?
Any suggestions ?

/Jesper
http://www.yampp.com
The quick black AVR jumped over the lazy PIC.
What boots up, must come down.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Was the filter designed to be driven from 50 Ohm source or the "zero" output impedance of an op-amp? Or is it a 50Ohm line driver?

50 Ohm cable (shiny or not) does not present a 50 Ohm load to anything. What matters is what you have it terminated in at the far end.

Why the low pass filter?

What frequency are you trying to work up to?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

OK, first question, same as davef. What frequency (range)?

Second. Right now, passive (LC) or active (op-amp) filters.?

Third: What amplitude? What load? (I guess you said "mostly" high impedance).

Fourth: is this signal from a "real" DAC? What is the sample rate?

ObservationL up to 100KHz or so corner frequency, you may be better off with a cascade of Sallen-Key 2nd order filters. Then, you would not have to worry about load impedance and they are a lot easier to work with.

Observation: If the source is a real DAC, then you probably only need to be down 30-40db at the sample frequency. How sharp the filter needs to be depends entirely on the ratio of the max signal frequency to the sample frequency.

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!

 

 

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

davef wrote:

50 Ohm cable (shiny or not) does not present a 50 Ohm load to anything. What matters is what you have it terminated in at the far end.

Why not the cable present a load? Not even when the signal has entered the cable but has not exited the other side yet.

If you have a short on the end of the cable, you see the cable impedance until the signal has reached the short circuit, right?

This is why you need same source impedance, transmission line impedance (=cable), and termination impedance. Nobody ever uses 75 ohm cables with systems having 50 ohm source and termination impedance.

Edit:

Normally in signal generators etc filtering is done before output buffering, so therefore it does not matter if you have a high impedance or terminated load, the signal will be correctly filtered.

But in some systems (video, digital audio) the termination is always there, so filters are made to assume terminated environment. For example, a VGA video card has output filters, and it must operate properly when cable connected (terminated properly), but if the VGA cable is not connected, who cares if filters are way off.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Why not the cable present a load? Not even when the signal has entered the cable but has not exited the other side yet.

If you have a short on the end of the cable, you see the cable impedance until the signal has reached the short circuit, right?

Yes, for x nanoseconds (length of cable times velocity factor) the output will "see" the cable impedance from the start of the signal you are trying to propagate down the line. But what happens after that? The signal bounces back (from your short circuit)and if the line is 1/4 wavelength long it looks like it came from an open circuit, doesn't match the source impedance and bounces back again, etc and etc. Standing waves and VSWR are the key words.

In signal generators, well low power ones, you are correct about the position of filtering. In high power RF amplifiers the filtering has to come last.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Filter is a Butterworth, cutoff not decided but about 80 MHz. Filter is fed from an OPA690 opamp (50 R series resistor) following a DAC clocked at 160 MHz. Output 0.5Vpp.
I might rearrange this so the filter goes before the opamp.

/Jesper
http://www.yampp.com
The quick black AVR jumped over the lazy PIC.
What boots up, must come down.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Minor point, but what is the output impedance of the OPA690 at 80MHz? I suspect greater than 0 Ohms. If your filter was designed for 50 Ohms at both ends, subtract the Zout of the OPA690 from your 50 Ohm series resistor.

Is the DAC going to need a buffer to drive this filter? Or, what load can the DAC drive?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The problem is not the feeding of the filter, but how to best handle various loads.
The dac can drive it, it'll do 0.5Vpp on a 50R load.

/Jesper
http://www.yampp.com
The quick black AVR jumped over the lazy PIC.
What boots up, must come down.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Commercial DDS function generators like this one:

http://www.tti-test.com/products-tti/text-pages/gen-tg4001.htm

have a 50 ohm output, but can obviously be used into a high impedance. I think it has a broadband amplifier following the filter, which is how I'd do it.

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Filter - line driver - overvoltage, short circuit protection - attenuator - output

Stealing Proteus doesn't make you an engineer.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

An attenuator isn't needed, the amplitude can be varied in the software.

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jepael wrote:

This is why you need same source impedance, transmission line impedance (=cable), and termination impedance. Nobody ever uses 75 ohm cables with systems having 50 ohm source and termination impedance.

Actually, they do.

Losses in 75 ohm cable are lower than 50 ohm. So if the loss due to mismatch is less than the gain in lowered feedline loss, it makes sense to do this.
Also, 75 ohm cable can be dramatically less expensive than 50 ohm cable. The reflections can be managed, and even made to work for you as stub filters.

In general, higher impedance cables are less lossy, since the current goes down, and I^2R losses drop dramatically. The reason low Z cables exist, is that they avoid the high voltages required for pushing large power in the higher impedance cables. At 1W it's irrelevant, at 1kW it matters.

Everything is a tradeoff..

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

jesper wrote:

Two factors:
* I've seen that terminating with 25R looks way better than no termination at all.
* As this is not an instrument that mostly will be used in nice labs with nice cables it will more often than not be connected to high impedance inputs.

Definitely true on #1, a 2-1 VSWR isn't that big of a deal. For a lab instrument, at low-ish frequencies, you might just have the output be a unity gain op-amp, and not worry about termination. Terminate your filter in the impedance it expects, which feeds the summing junction of the opamp, and let the opamp worry about keeping the output level constant over whatever impedance is presented.

A socketed opamp can be replaced easily if it couldn't quite take that big of a joke.

If the frequency range works for you, the LM-12 is a pretty butt-kicking output stage. :) Obsolete though.. :( LMH6644 looks nice through VHF

On the antenna end of things, people WAY over-worry on VSWR, and under-worry on resonance. A proper antenna design is made resonant first, then matched to the feedline. Works MUCH better that way.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

leon_heller wrote:
An attenuator isn't needed, the amplitude can be varied in the software.
Yes, and when you are fiddling with an amplitude close to the LSB of your DAC you wish you had a simple fixed attenuator you could switch into the output.

Stealing Proteus doesn't make you an engineer.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Attenuators: When you have a 10dB attenuator in the output (almost all commercial function generators I have seen have this option, and on the good HP ones there's one they don't tell you about you can't disable), the output stage inside will see about 60 ohm with the output unloaded. If only a 6dB attenuator is used, 83 ohm is seen. The attenuator also serves to protect the output stage from reverse power (not just reflection, also accidentally connecting the generator to the wrong port on the amplifier)

When you're using the generator at >5MHz you will need correct cabling. Otherwise you run into resonance in cabling itself, and exceed legal emission limits.

/Kasper

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Jesper isn't using an attenuator, AFAIK. All he needs is a 50 ohm broadband amp after the filter

Leon

Leon Heller G1HSM

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I'll move things around to have the op-amp at the output. That'll make the impedance matching so much simpler.

Attenuation is a completely different chapter.
While I of course can vary the output by manipulating DAC data, this does, as mentioned above, get tricky at low amplitudes.

I would prefer to have a variable output voltage, under CPU control, obviously, but haven't found the "right" way to do this yet at these frequencies.
Suggestions welcome ;-)

/Jesper
http://www.yampp.com
The quick black AVR jumped over the lazy PIC.
What boots up, must come down.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Put some kind of selectable gain to opamp circuit. Digital potentiometre or some analog multiplexers to select feedback values. Maybe even relays?

Another trick is to have large enough gain on the opamp amplifier, and have some sort of attennuation network after it to select proper level. Relays/muxes.

I don't know how you would manage the 50 ohm output, except if you have one buffer opamp after the DAC filter, then the attennuation network and another buffer opamp right before output connector.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Relays are out, and digital pots and muxes don't like 100MHz much.

/Jesper
http://www.yampp.com
The quick black AVR jumped over the lazy PIC.
What boots up, must come down.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

How about putting a "variable gain amplifier" in the middle of the circuit somewhere.

This page: Analog Devices Variable Gain Amplifiers has a table which lists a bunch of them. Analog and Digital control, bandwidth from 10's of MHz up through several GHz, and a range of gains.

JC