I have now almost finished building a differential active crossover, with solid-state circuits for the low-pass leg, and the SLCF buffers from Allen's RTP5 preamp for the high-pass leg.
I have just done some frequency response measurements, and although the high-pass and low-pass filters appear to be doing their job well around the crossover frequency (550Hz), I was surprised to see a large peak in the high-pass response at about 80Hz.
I can't see any obvious passive time constants as high as this (2ms). After some thought, the only way that I can imagine such a low-frequency resonance could arise is from the biasing circuit for the buffers, which keep the grids at 150V (half the 300V B+). Here is the circuit:
The circuit here is fed from another SLCF buffer. The second-order high-pass filter (which I have marked in red) is made up of the 20.5K and 41K resistors and the 10n capacitors, and the 41K resistors meet at a virtual earth which is held at a DC potential of 150V.
Can anyone see what might be going on here, and is there a better way to bias the grids?
Thanks!
Alex
I have just done some frequency response measurements, and although the high-pass and low-pass filters appear to be doing their job well around the crossover frequency (550Hz), I was surprised to see a large peak in the high-pass response at about 80Hz.
I can't see any obvious passive time constants as high as this (2ms). After some thought, the only way that I can imagine such a low-frequency resonance could arise is from the biasing circuit for the buffers, which keep the grids at 150V (half the 300V B+). Here is the circuit:
An externally hosted image should be here but it was not working when we last tested it.
The circuit here is fed from another SLCF buffer. The second-order high-pass filter (which I have marked in red) is made up of the 20.5K and 41K resistors and the 10n capacitors, and the 41K resistors meet at a virtual earth which is held at a DC potential of 150V.
Can anyone see what might be going on here, and is there a better way to bias the grids?
Thanks!
Alex
Further tests reveal that the resonance at 80Hz is actually common-mode (it disappears when I measure across hot and cold outputs, while the signal doubles). I find that very puzzling.
Here is a frequency response plot for one channel of the crossover:
The low-pass response is near perfect; high-pass shows the peak and also a difference between hot and cold which may connected with the phase changes (though I would have expected a sharp null somewhere...).
Alex
Here is a frequency response plot for one channel of the crossover:
An externally hosted image should be here but it was not working when we last tested it.
The low-pass response is near perfect; high-pass shows the peak and also a difference between hot and cold which may connected with the phase changes (though I would have expected a sharp null somewhere...).
Alex
Differential filter: problem solved
I've now fixed the problem, simply by shorting across the 1M0 resistors between the virtual earth point and the 150V bias divider, thereby removing the differential action of the filter circuits (although there is still a LTP at the input to ensure that a differential signal is passed to the filters).
I completely fail to understand why such a differential filter should be unstable, however. On paper it should work fine. Might it be that the filter requires a higher degree of balanced antiphase inputs than I am supplying? I still can't imagine where the 80Hz time constant would come from.
Alex
I've now fixed the problem, simply by shorting across the 1M0 resistors between the virtual earth point and the 150V bias divider, thereby removing the differential action of the filter circuits (although there is still a LTP at the input to ensure that a differential signal is passed to the filters).
I completely fail to understand why such a differential filter should be unstable, however. On paper it should work fine. Might it be that the filter requires a higher degree of balanced antiphase inputs than I am supplying? I still can't imagine where the 80Hz time constant would come from.
Alex
Hi Alex,
That all sounds familiar.
I worked on a differential active filter about a year ago and ultimately
decided that the component matching problem would be too much
trouble. I asked myself why it needed to be differential or balanced
and couldn't justify it. I'll use transformer coupling for input and output
anyway which will work fine with my balanced XLR studio interconnects.
I decided to work on SE circuits to reduce the noise and linearize the
transfer function and came up with a SE augmented cathode follower.
The PSRR is improved by the MOSFET bootstraps and CCS bias. For
my purposes at least, this gets me to my goal of a great active line
level crossover sooner with less risk.
I have another version with cascoded MOSFETS everywhere but in
the end decided to simplify it to this for the first try.
Cheers,
Michael
That all sounds familiar.
I worked on a differential active filter about a year ago and ultimately
decided that the component matching problem would be too much
trouble. I asked myself why it needed to be differential or balanced
and couldn't justify it. I'll use transformer coupling for input and output
anyway which will work fine with my balanced XLR studio interconnects.
I decided to work on SE circuits to reduce the noise and linearize the
transfer function and came up with a SE augmented cathode follower.
The PSRR is improved by the MOSFET bootstraps and CCS bias. For
my purposes at least, this gets me to my goal of a great active line
level crossover sooner with less risk.
I have another version with cascoded MOSFETS everywhere but in
the end decided to simplify it to this for the first try.
Cheers,
Michael
Attachments
Michael,
Thanks for your response. The primary reason for going for a differential crossover was because I already use, and am very happy with, the Vacuum State RTP3 preamp I built two or three years ago, and my next project will be a 300B amp with a differential input.
Your buffer circuit looks rather like mine, although you have clearly put effort into optimising the current sink and bootstrap elements. Transformer coupling is an elegant solution. The one thing that has kept me away from using transformers in my system, apart from the cost and real estate, is the thought of all that copper. I have had one short experience with an OTL amplifier, next to a similar amp with an output transformer from the same designer, and heard a huge gain in transparency. All the same, I understand that there are plenty of people out there who would rather use a whole pile of transformers than a single capacitor.
My crossover, while on my workbench, is being fed from a single-ended signal generator, and I suspect that the differential input stage isn't quite perfoming as it should: the plot I posted has all the traces normalised by the passband amplitude, and in reality my circuit has a few dB of imbalance which I need to sort out. I still don't see why two single-ended filter stages which are working correctly should have such a response anomaly when they are used differentially.
Now I am running the filters single-ended, their responses seem to be matched to within a fraction of a dB (and also very close to the target LR4 curves), which is very pleasing. I matched components to within 1%, which seems to be enough in this case.
Alex
Thanks for your response. The primary reason for going for a differential crossover was because I already use, and am very happy with, the Vacuum State RTP3 preamp I built two or three years ago, and my next project will be a 300B amp with a differential input.
Your buffer circuit looks rather like mine, although you have clearly put effort into optimising the current sink and bootstrap elements. Transformer coupling is an elegant solution. The one thing that has kept me away from using transformers in my system, apart from the cost and real estate, is the thought of all that copper. I have had one short experience with an OTL amplifier, next to a similar amp with an output transformer from the same designer, and heard a huge gain in transparency. All the same, I understand that there are plenty of people out there who would rather use a whole pile of transformers than a single capacitor.
My crossover, while on my workbench, is being fed from a single-ended signal generator, and I suspect that the differential input stage isn't quite perfoming as it should: the plot I posted has all the traces normalised by the passband amplitude, and in reality my circuit has a few dB of imbalance which I need to sort out. I still don't see why two single-ended filter stages which are working correctly should have such a response anomaly when they are used differentially.
Now I am running the filters single-ended, their responses seem to be matched to within a fraction of a dB (and also very close to the target LR4 curves), which is very pleasing. I matched components to within 1%, which seems to be enough in this case.
Alex
- Status
- Not open for further replies.