A few years ago, Burr-Brown propagated the use of a GIC low pass filter at a DAC's output or ADC's input instead of the more common Sallen-Key and multiple feedback configurations. Advantages were supposedly lower distortion and noise.
In the newer data sheets, even that of the PCM1704, they have converted to mulitple feedback. Why is that? Lower component count? Were there unknown disadvantages to the GIC topology? Anybody tried it?
Eric
P.S. the attached app note is really pdf format and should be openend with Acrobat, but the forum insisted it does not like pdf file extensions...
In the newer data sheets, even that of the PCM1704, they have converted to mulitple feedback. Why is that? Lower component count? Were there unknown disadvantages to the GIC topology? Anybody tried it?
Eric
P.S. the attached app note is really pdf format and should be openend with Acrobat, but the forum insisted it does not like pdf file extensions...
Jocko,
Since you asked how about a 96k 24 bit D/A with upsampling using PCM 1704's, balanced outputs(true), display of sampling and bit rate, switched phase inversion, selectable filters, four inputs..... have I left anything out. Don't violate any agreements but if a company is out of buisness, are their patents still valid?
You could leave the power supply and I/V conversion to us.
Regards,
Jam
P.S. I am not asking for too much, am I?
Since you asked how about a 96k 24 bit D/A with upsampling using PCM 1704's, balanced outputs(true), display of sampling and bit rate, switched phase inversion, selectable filters, four inputs..... have I left anything out. Don't violate any agreements but if a company is out of buisness, are their patents still valid?
You could leave the power supply and I/V conversion to us.
Regards,
Jam
P.S. I am not asking for too much, am I?
Let's see: 4 inputs? Isn't 2 enough. Hard to get sampling frequency form the Crystal RX, unlees you use the non-96 kHz version. And they aren't a lot of help on getting the de-emphasis flag, either. Switching phase is easy, a pin on the 1704 does it for you. Selectable filters?? Digital or analog. And why?
But since you asked.....I may have something in the works (in the way of a PCB) that I might be able to sell you. Stay tuned.
Nothing I've built is patented. Not so sure about Hairy's stuff. I don't believe in them.
Jocko
But since you asked.....I may have something in the works (in the way of a PCB) that I might be able to sell you. Stay tuned.
Nothing I've built is patented. Not so sure about Hairy's stuff. I don't believe in them.
Jocko
Some further thoughts.........
Lucky you didn't ask about the P/S and I/V stages. Those are the key to making this stuff sound good. The digital stuff can come straight from the data sheet. You can get away with quite a lot here. Provided the grounding and bypassing are good. And I can draw it out, but how you implement it is the trick. The main thing is to keep the current loops small to minimise EMI. The stuff is very forgiving. As long as it gets the right voltage without a lot of pertubations, it works ok.
Same for the front end interface. The schematic is not all there is. I built my first D/A in one night. Spent the next 6 months getting the input circuit right.
As for whom I may have done work for........
This is, say we say, a very egocentric industry. No one wants it to leak out that some goofball in Texas, or North Carolina, designs part, or all, of their stuff. Even if they are out of business now, they won't be happy if I let the cat out of the bag. They will remember that next time they need a "hired gun".
Jocko
Lucky you didn't ask about the P/S and I/V stages. Those are the key to making this stuff sound good. The digital stuff can come straight from the data sheet. You can get away with quite a lot here. Provided the grounding and bypassing are good. And I can draw it out, but how you implement it is the trick. The main thing is to keep the current loops small to minimise EMI. The stuff is very forgiving. As long as it gets the right voltage without a lot of pertubations, it works ok.
Same for the front end interface. The schematic is not all there is. I built my first D/A in one night. Spent the next 6 months getting the input circuit right.
As for whom I may have done work for........
This is, say we say, a very egocentric industry. No one wants it to leak out that some goofball in Texas, or North Carolina, designs part, or all, of their stuff. Even if they are out of business now, they won't be happy if I let the cat out of the bag. They will remember that next time they need a "hired gun".
Jocko
If my memory serves correct, the problem with GIC filters is one of power supply headroom.
I did a GIC notch filter once and discovered that the quality and depth of the notch was dependent upon an equal and opposite signal to "buck" the main signal. Once an opamp hits a power supply rail the quality of the filter goes awry.
If you look at the basic GIC topology it looks almost like a "T" attenuator - much of the signal passes through cleanly right across the top. The opamps in the "leg" counter the incoming signal to remove it (low pass).
Too much feedback for my tastes. It's like a cat continually chasing it's tail. If you can't do it with Ls and Cs, then stick with the basic Sallen-Key or similar.
I did a GIC notch filter once and discovered that the quality and depth of the notch was dependent upon an equal and opposite signal to "buck" the main signal. Once an opamp hits a power supply rail the quality of the filter goes awry.
If you look at the basic GIC topology it looks almost like a "T" attenuator - much of the signal passes through cleanly right across the top. The opamps in the "leg" counter the incoming signal to remove it (low pass).
Too much feedback for my tastes. It's like a cat continually chasing it's tail. If you can't do it with Ls and Cs, then stick with the basic Sallen-Key or similar.
> If my memory serves correct, the problem with GIC filters is one of power supply headroom.
Well, with +/- 15 V, filtering a +/- 3 V peak-peak signal should pose no problem.
> If you look at the basic GIC topology it looks almost like a "T" attenuator - much of the signal passes through cleanly right across the top. The opamps in the "leg" counter the incoming signal to remove it (low pass).
That would be an argument in favor of the GIC topology, wouldn't it? In fact, on of the older PCM data sheets recommends using OPA637s for the I/V and output buffer (the "bar") and OPA2604s in the "leg", as those only worked on the signal to be discarded.
The THD+N vs. frequency plots given in the app note I attached to this thread look as good as one can expect when using a FET-input amp like the 2604 as a unity gain follower (common mode distortion on the relatively high source impedance). I don't see any evidence of extra distortion in the "leg".
> Too much feedback for my tastes. It's like a cat continually chasing it's tail. If you can't do it with Ls and Cs, then stick with the basic Sallen-Key or similar.
I don't see that there is any more or less feedback than in the Sallen-Key or multiple feedback configurations.
Eric
Well, with +/- 15 V, filtering a +/- 3 V peak-peak signal should pose no problem.
> If you look at the basic GIC topology it looks almost like a "T" attenuator - much of the signal passes through cleanly right across the top. The opamps in the "leg" counter the incoming signal to remove it (low pass).
That would be an argument in favor of the GIC topology, wouldn't it? In fact, on of the older PCM data sheets recommends using OPA637s for the I/V and output buffer (the "bar") and OPA2604s in the "leg", as those only worked on the signal to be discarded.
The THD+N vs. frequency plots given in the app note I attached to this thread look as good as one can expect when using a FET-input amp like the 2604 as a unity gain follower (common mode distortion on the relatively high source impedance). I don't see any evidence of extra distortion in the "leg".
> Too much feedback for my tastes. It's like a cat continually chasing it's tail. If you can't do it with Ls and Cs, then stick with the basic Sallen-Key or similar.
I don't see that there is any more or less feedback than in the Sallen-Key or multiple feedback configurations.
Eric
Jocko,
I would be very interested in the PC board that you are designing. What I ment by different filters was the use of different decoding algorithms eg. Wadia.(this might be too much of a hassle).
I agree the front end interface is all imporant, what would be your design criteria for it.
Jam
I would be very interested in the PC board that you are designing. What I ment by different filters was the use of different decoding algorithms eg. Wadia.(this might be too much of a hassle).
I agree the front end interface is all imporant, what would be your design criteria for it.
Jam
WADIA!!!!!!!!!!!
Don't get me started on them.
From what we could tell, their goofy algorithim was just a Bessle done in the digital domain. Let all sorts of ultrasonic crap through.
That, and their "instantaneous JFET converter". All they did was stack up a bunch of B-B DACs in some kludge to get more bits out of 16 bit parts.
Just stick to what the filter spits out.
Design criteria............(Hairy, are you reading this.....?)
LOW REFLECTIONS!
That means you can't use Schmitt trigger inputs without some sort of buffer between it and the line.
It must be 75 ohms, and resistive.
Jocko
Don't get me started on them.
From what we could tell, their goofy algorithim was just a Bessle done in the digital domain. Let all sorts of ultrasonic crap through.
That, and their "instantaneous JFET converter". All they did was stack up a bunch of B-B DACs in some kludge to get more bits out of 16 bit parts.
Just stick to what the filter spits out.
Design criteria............(Hairy, are you reading this.....?)
LOW REFLECTIONS!
That means you can't use Schmitt trigger inputs without some sort of buffer between it and the line.
It must be 75 ohms, and resistive.
Jocko
Jocko,
I did not mean to get your shorts in a twist by referring to Wadia, but I was just throwing out suggestions.
I am sitting at the edge of my seat waiting to see the results of yor labours and Hairy's thoughts on your design.
Jam
P.S. ...meanwhile back in High Point Scott Nixon is drinking mint Julieps and watching the grass grow.
I did not mean to get your shorts in a twist by referring to Wadia, but I was just throwing out suggestions.
I am sitting at the edge of my seat waiting to see the results of yor labours and Hairy's thoughts on your design.
Jam
P.S. ...meanwhile back in High Point Scott Nixon is drinking mint Julieps and watching the grass grow.
>"Well, with +/- 15 V, filtering a +/- 3 V peak-peak signal should pose no problem."
Unfortunately, I don't believe the relationship is quite so simple.
The filter I built using +/-12V supplies couldn't properly handle a 1V signal. It was an extreme case, though (notch). A Bessel or Butterworth type low pass may not be a problem. However, the elliptical or Cauer types with their zeros are certainly a candidate.
One other issue, the closed-loop bandwidth of the "leg" opamps are critical. As they run out of steam, so does your filter. Same goes for phase delay. At some point the filtering becomes meaningless because the opamps are so far behind the error signal that they no longer do their job. That is why I like a solid RC up front outside the feedback (like the Sallen-Key).
Unfortunately, I don't believe the relationship is quite so simple.
The filter I built using +/-12V supplies couldn't properly handle a 1V signal. It was an extreme case, though (notch). A Bessel or Butterworth type low pass may not be a problem. However, the elliptical or Cauer types with their zeros are certainly a candidate.
One other issue, the closed-loop bandwidth of the "leg" opamps are critical. As they run out of steam, so does your filter. Same goes for phase delay. At some point the filtering becomes meaningless because the opamps are so far behind the error signal that they no longer do their job. That is why I like a solid RC up front outside the feedback (like the Sallen-Key).
GIC Notch Filter
Found it. Here's a head start for your simulations...
www.hagtech.com/pdf/notch.pdf
www.hagtech.com/notch.html
jh
Found it. Here's a head start for your simulations...
www.hagtech.com/pdf/notch.pdf
www.hagtech.com/notch.html
jh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.