Peter Daniel said:
It'll be interesting to find out what happens when he builds it. Since we're summing two signals that drop at 12 db/octave, I'll bet it means he'll get 24. LR with a single gain stage?
OK, Uli, wake up and get building
Perhaps the master can enlighten us without giving away too much of his crossover article.
Everything but the current sink......
"I hope you using are using decent current sources....."
With cheap RCAs? That would be tacky and in bad taste, like wearing white shoes after Labor Day. Chris Sommevigo gave me those RCAs and I don't think they were that cheap ........ He had better send Cardas next time!
Are they current sources or sinks, and just what is the difference exactly?
Fred
Send Cardas when only the very best will do.
"I hope you using are using decent current sources....."
With cheap RCAs? That would be tacky and in bad taste, like wearing white shoes after Labor Day. Chris Sommevigo gave me those RCAs and I don't think they were that cheap ........ He had better send Cardas next time!
Are they current sources or sinks, and just what is the difference exactly?
Fred
Send Cardas when only the very best will do.
What is the sound of one engineer swearing grasshopper?
"It certainly is true that the brighter of our readership can reverse engineer an XVR1 from the owner's manual."
Brighter than me..... I cant seem to figure out the resistor/ frequency steps from the 18K and 4.5 K resistor packs. Knowing just the one of the capacitor values might nudge me toward it but that would be cheating and take all the fun out of it. I seem to remember some thing about changing resititivity mid track to fake a non linear relationship. There is also the law faking resistor with a linear pot.......
Could this be the case of a single word ( or capactance value being worth a thousand pictures and what is that that faint bwaa....haha noise in the distance?
Of course one could save himself the agrevation of pondering and a sporty PCB layout ( Autorouter anyone?) for connecting a series resistor matrix and use an array of resistors than can be selected in parallel combinations.
During the excellent Earth to the Moon series in the episode about building the LEM, one of the engineers worried about the translation of axis during docking when looking out the top window. Tom Kelly told him "The astronauts are smart guys, they will figure it out. Hmm... Several jumpers in a short parallel resistor programming header or one in a long series header (times one, two, or three or four programing headers corresponding to the filter order )?
I wonder if that woud work with the capacitors as well.......
Design of crossover filters requires great peace of mind.
"It certainly is true that the brighter of our readership can reverse engineer an XVR1 from the owner's manual."
Brighter than me..... I cant seem to figure out the resistor/ frequency steps from the 18K and 4.5 K resistor packs. Knowing just the one of the capacitor values might nudge me toward it but that would be cheating and take all the fun out of it. I seem to remember some thing about changing resititivity mid track to fake a non linear relationship. There is also the law faking resistor with a linear pot.......
Could this be the case of a single word ( or capactance value being worth a thousand pictures and what is that that faint bwaa....haha noise in the distance?
Of course one could save himself the agrevation of pondering and a sporty PCB layout ( Autorouter anyone?) for connecting a series resistor matrix and use an array of resistors than can be selected in parallel combinations.
During the excellent Earth to the Moon series in the episode about building the LEM, one of the engineers worried about the translation of axis during docking when looking out the top window. Tom Kelly told him "The astronauts are smart guys, they will figure it out. Hmm... Several jumpers in a short parallel resistor programming header or one in a long series header (times one, two, or three or four programing headers corresponding to the filter order )?
I wonder if that woud work with the capacitors as well.......
Design of crossover filters requires great peace of mind.
I love Uli's innovative design but looking through all the literature it makes me wonder if Mr Pass has already done these hard yards before us.
"It was created over a period of several years for our own loudspeaker design efforts, and it is the design that ultimately performed with the most flexibility and quality, and gave us the most satisfaction with the end result. "
"The basic circuitry is a very simple discrete single-ended Class A stage designed to operated at low gain and feedback levels with high performance and stability."
I am not suggesting an X active filter it is not a brilliant idea on Uli's part. However careful reading of both the Borbely and Pass diy literature regards the classical requirements for active filter buffers and Mr Pass known pre disposition for less is more suggests there may be un realised subjective virtues in the minimalist XVR1 buffer.
It would appear the circuit is rather clever, it uses only two stages, (like the Aleph) has very low noise, very low distortion and low gain with low feeback while the full balanced output topology cancels all common mode noise/distortion.
Surely if full complementry, cascoding or whatever was a subjective a benefit it would have found its way into the final design. The practical implementation is certainly straight forward.
Has anyone considered or trialed this buffer with a breadboard or done any simulations at this stage?
Again, I am not attempting to distract from Uli's endeavour, just interested to hear if any of us has combed over the existing circuit model.
Maybe the next version of the XVR1 will be XVR1-Xuli.
macka
"It was created over a period of several years for our own loudspeaker design efforts, and it is the design that ultimately performed with the most flexibility and quality, and gave us the most satisfaction with the end result. "
"The basic circuitry is a very simple discrete single-ended Class A stage designed to operated at low gain and feedback levels with high performance and stability."
I am not suggesting an X active filter it is not a brilliant idea on Uli's part. However careful reading of both the Borbely and Pass diy literature regards the classical requirements for active filter buffers and Mr Pass known pre disposition for less is more suggests there may be un realised subjective virtues in the minimalist XVR1 buffer.
It would appear the circuit is rather clever, it uses only two stages, (like the Aleph) has very low noise, very low distortion and low gain with low feeback while the full balanced output topology cancels all common mode noise/distortion.
Surely if full complementry, cascoding or whatever was a subjective a benefit it would have found its way into the final design. The practical implementation is certainly straight forward.
Has anyone considered or trialed this buffer with a breadboard or done any simulations at this stage?
Again, I am not attempting to distract from Uli's endeavour, just interested to hear if any of us has combed over the existing circuit model.
Maybe the next version of the XVR1 will be XVR1-Xuli.
macka
X simplicity
There are a lot of brilliant things going on in Xed circuits.
First of all the count of active gainstages is exactly the same
as in similar non X circuits.
Example:
In Passlabs XVR you got one input buffer consisting of 2
stages next are 2 filterbuffers consisting of 2 stages each
and finally the outputbuffer consisting of 2 stages.
Total count: 8
X version count: 8
When you X this arrangement the activestagecount stays the
same except the outputs of the stages but this is a matter of
definition whether to count them as extra active device or not.
To Peters hint having a 12dB slope with just one pole.
Mathematically this is not logical so I will try to explain why.
(Please forive my lacking of writing art, Grey is the hero for that.)
Imagine a floating balanced input for example a Xfmr.
It simply doesn´t matter if you couple one of the input nodes via
10µF and the other node directly or couple both of them with two
caps 20µF (!) each as they are in series electrically.
Maybe Peter misinterpreted the resulting one octave higher
hipass frequency as 12dB. This result you get when you couple
two times 10µF instead of 20.
Back to XXover circuitry:
You definitely double the filter components like caps and resistors
but I consider this not as doubling in the signal path since they
work "antiparallel" (out of phase) In implementing such circuitry
you have the problem to intrinsic matching at least the caps in
the circuit whereas the resistors can be chosen 0,1% which is
recommended when using the values shown in an older post.
To be continued.....
Uli
There are a lot of brilliant things going on in Xed circuits.
First of all the count of active gainstages is exactly the same
as in similar non X circuits.
Example:
In Passlabs XVR you got one input buffer consisting of 2
stages next are 2 filterbuffers consisting of 2 stages each
and finally the outputbuffer consisting of 2 stages.
Total count: 8
X version count: 8
When you X this arrangement the activestagecount stays the
same except the outputs of the stages but this is a matter of
definition whether to count them as extra active device or not.
To Peters hint having a 12dB slope with just one pole.
Mathematically this is not logical so I will try to explain why.
(Please forive my lacking of writing art, Grey is the hero for that.)
Imagine a floating balanced input for example a Xfmr.
It simply doesn´t matter if you couple one of the input nodes via
10µF and the other node directly or couple both of them with two
caps 20µF (!) each as they are in series electrically.
Maybe Peter misinterpreted the resulting one octave higher
hipass frequency as 12dB. This result you get when you couple
two times 10µF instead of 20.
Back to XXover circuitry:
You definitely double the filter components like caps and resistors
but I consider this not as doubling in the signal path since they
work "antiparallel" (out of phase) In implementing such circuitry
you have the problem to intrinsic matching at least the caps in
the circuit whereas the resistors can be chosen 0,1% which is
recommended when using the values shown in an older post.
To be continued.....
Uli
Re: X simplicity
I'm almost sure that I didn't misinterpret it. But I was using a balanced input in a single ended fashion, where unused, inverting input was connected to the ground through a coupling cap. This probably made another 6dB pole.
If you'll be using balanced circuit all the way through, I think the poles will behave as expected, but if you start using balanced circuit in a single ended fashion, you might consider getting different results.
uli said:
I'm almost sure that I didn't misinterpret it. But I was using a balanced input in a single ended fashion, where unused, inverting input was connected to the ground through a coupling cap. This probably made another 6dB pole.
If you'll be using balanced circuit all the way through, I think the poles will behave as expected, but if you start using balanced circuit in a single ended fashion, you might consider getting different results.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.