And now the 120Hz one. This was trickier, because I haven't worked out the theory for the center tapped controls - so this was done manually. Anyway I now have a manual algorithm that works very efficiently with exceptionally high accuracy (0.01dB error per step).
This one might seem like rather quirky steps of 0.75dB. This is half the 1.5dB steps of the 15Hz and 25kHz designs. Well - I had to pick a number!
This one might seem like rather quirky steps of 0.75dB. This is half the 1.5dB steps of the 15Hz and 25kHz designs. Well - I had to pick a number!
Sawyers,
Just came across the third Palette pre copy I mentioned and supplied a pic of in posting 127. Actually made up three of them. Sold one to Robyn Wyatt and have the second in my main system.
It has the control unit shown with the 30ft umbilical to a separate small box containing the power supply and in/out RCA's.
If interested, could send it to you to play with for a few weeks - and "try out" the concept. Only cost you postage both ways -10lbs or so - I'm in 13820
Charles
Just came across the third Palette pre copy I mentioned and supplied a pic of in posting 127. Actually made up three of them. Sold one to Robyn Wyatt and have the second in my main system.
It has the control unit shown with the 30ft umbilical to a separate small box containing the power supply and in/out RCA's.
If interested, could send it to you to play with for a few weeks - and "try out" the concept. Only cost you postage both ways -10lbs or so - I'm in 13820
Charles
That is a very decent offer Charles! Let me have a look at what it would cost to get it across the pond both ways. Is that 10lbs the unit itself or an estimate of packed weight?
Cheers!
Craig
Cheers!
Craig
I've just thought of a slight problem - my entire system uses balanced. I designed it that way because I have an active crossover with five outputs per channel and 10 separate power amps in 3 chassis. The thought of managing potential hum loops with that drove me to go balanced. That has gone through the whole system now, including a balanced phono stage (both in an out). It made the whole system plug and play, but not interface friendly with anything with RCA connectors.
But male and female RCA to XLR adaptors are easily available - so I think there is a workaround.
Let me further check shipping costs.
Craig
But male and female RCA to XLR adaptors are easily available - so I think there is a workaround.
Let me further check shipping costs.
Craig
I've been doing a look see at different filter topologies.
Original Audio Palette (and Burwen's prototype. Uses a multiple feedback filter in a feedback network.
Palette Preamp. Uses the design from the National Semiconductor Audio Handbook 1977. Big advantage - one opamp per filter.
Others that I've looked at are:
Constant Q. This was described in the Pro Audio company Rane, whose legacy library is located here https://www.ranecommercial.com/legacy/library.html . The relevant note is https://www.ranecommercial.com/legacy/pdf/constanq.pdf . That ends up using the same multiple feedback filter as the original Audio Palette. And his figure 8 (with a center tapped control to ground) with the multiple feedback filter in figure 26 is basically what the 120Hz to 5kHz controls are. And if you recall I commented that those controls in the Audio Palette behaved as constant Q. The outer filters at 15Hz and 25kHz don't have a center tap and are variable Q. Darned subtle, Watson; pass the violin.
State variable. This for some reason is always a bit controversial, for no good reason. It basically insets a state variable filter in place of the multiple feedback filter. You just use the bandpass tap. The rest of the circuit remains pretty much the same. All the other options above need just about every component in the filter block changed to just shift the frequency but leave the Q constant. With a state variable filter block, the frequency and Q can be set independently. So you can shift the frequency without changing the circuit Q. Neat. I guess the downside is that it needs three opamps for the state variable bit, and then one to do the feedback thing to handle the boost and cut control.
Wein network. This basically uses a Wein filter in the feedback loop. The design rules are a bit on the quirky side - if the loop gain exceeds ~3 you get an oscillator. The frequency is determined by the Wein C's and R's, and you can mess with the Q by imbalancing either the R's or C's. It is very difficult to get the correct combination of required boost and Q. The higher the boost the higher the Q I shelved that one.
For a much more comprehensive review - more light bedtime reading - https://www.ranecommercial.com/legacy/pdf/ranenotes/Operator_Adjustable_Equalizers_Overview.pdf
Craig
Original Audio Palette (and Burwen's prototype. Uses a multiple feedback filter in a feedback network.
Palette Preamp. Uses the design from the National Semiconductor Audio Handbook 1977. Big advantage - one opamp per filter.
Others that I've looked at are:
Constant Q. This was described in the Pro Audio company Rane, whose legacy library is located here https://www.ranecommercial.com/legacy/library.html . The relevant note is https://www.ranecommercial.com/legacy/pdf/constanq.pdf . That ends up using the same multiple feedback filter as the original Audio Palette. And his figure 8 (with a center tapped control to ground) with the multiple feedback filter in figure 26 is basically what the 120Hz to 5kHz controls are. And if you recall I commented that those controls in the Audio Palette behaved as constant Q. The outer filters at 15Hz and 25kHz don't have a center tap and are variable Q. Darned subtle, Watson; pass the violin.
State variable. This for some reason is always a bit controversial, for no good reason. It basically insets a state variable filter in place of the multiple feedback filter. You just use the bandpass tap. The rest of the circuit remains pretty much the same. All the other options above need just about every component in the filter block changed to just shift the frequency but leave the Q constant. With a state variable filter block, the frequency and Q can be set independently. So you can shift the frequency without changing the circuit Q. Neat. I guess the downside is that it needs three opamps for the state variable bit, and then one to do the feedback thing to handle the boost and cut control.
Wein network. This basically uses a Wein filter in the feedback loop. The design rules are a bit on the quirky side - if the loop gain exceeds ~3 you get an oscillator. The frequency is determined by the Wein C's and R's, and you can mess with the Q by imbalancing either the R's or C's. It is very difficult to get the correct combination of required boost and Q. The higher the boost the higher the Q I shelved that one.
For a much more comprehensive review - more light bedtime reading - https://www.ranecommercial.com/legacy/pdf/ranenotes/Operator_Adjustable_Equalizers_Overview.pdf
Craig
hello Thorsten , 10 years no see in this forum .. what have you been doing for the last 10 years?????
Came across this company. https://www.rme-audio.de/home.html and was looking at the adi-2-pro-fs-be
Comprehensive balanced and single ended IO, and every digital IO going. https://www.rme-audio.de/adi-2-pro-fs-be.html
The filter menus have a fully parametric 5 band EQ with widely adjustable frequency and Q, and shelving filters too, with adjustable turnover frequency and Q. Both channels locked or separate L and R. Phase invert both, or each channel separately. Dual headphone outputs. The spec goes on and on.
It is a pretty phenomenal piece of kit for around $2k (£1,600)
Craig
Comprehensive balanced and single ended IO, and every digital IO going. https://www.rme-audio.de/adi-2-pro-fs-be.html
The filter menus have a fully parametric 5 band EQ with widely adjustable frequency and Q, and shelving filters too, with adjustable turnover frequency and Q. Both channels locked or separate L and R. Phase invert both, or each channel separately. Dual headphone outputs. The spec goes on and on.
It is a pretty phenomenal piece of kit for around $2k (£1,600)
Craig
As far as I know, it is not possible to manipulate DSD data stream.
And therefore RME does the conversion to PCM.
Their absolute great manual gives all the information.
And therefore RME does the conversion to PCM.
Their absolute great manual gives all the information.
Yes, you are right, also once EQ only outputs PCM that's not my cup of tea. But it's a interesting device can record from analog to DSD and other things like measurements,etc. But can't afford 1700€ for RME. Aso nor the discrete or opamp versions to make DRC using Cello Palette clone only 10 Elma 24 step switches attenuators 54€ each so 540€ and you always fixed to some frequencies or make it after you do the room measurement with precise needs so not very flexible.
I think only solution is using HQPlayer to do EQ DRC in DSD domain.....?
I think only solution is using HQPlayer to do EQ DRC in DSD domain.....?
Last edited:
No, there is the alternate solution of using an analogue EQ of your choice in the analogue domain as mentioned before...
It is an absolutely great piece of kit, can't find any fault with it. Of course, it's not 'just' an EQ but contains two stereo DACs of top quality plus ADCs and a host of routing options and two(!) high power headphone amps. But if you want to just sit back and play with the eq while listening, it's not ideal because of the small controls and tiny display, you just need to place it next to your seat. If you can do that (which you also would do with a Palette-style unit I guess), then it'a great option.
Jan
It's a while since I posted here, but the Cello whatever is still alive and kicking - at least in understanding and simulation.
There are two odd features in Burwen's design, at least one of which found itself into the Audio Palette.
The first is the blender. This is an odd beast indeed. It is basically cross coupled bandpass filters. The intention is to connect the two channels together at 20Hz and frequencies close to it. The bandpass filter is tuned to 20Hz with a Q of 0.7. The rationale is attached, both from Burwen's design study, and also from the Palette manual. Note that at 20Hz there is a significant boost of around 4dB with a stereo signal, and the Palette manual says that the boost at 20Hz can be counteracted by tweaking the 15Hz control down by a couple of dB.
Note that there is a mistake in the excerpt from Burwen in the attachment. At one point he says 120Hz, when he actually means 20Hz.
The Blend control was not incorporated into any other Cello product, so its usefulness must have been limited.
The second peculiarity in Burwen's design is the very odd balanced output drive. That has a line that feeds back the -ve balanced signal "to counteract common mode noise with long cables". Nothing I have seen before. I do not know whether this was incorporated into the Audio Palette.
Decades have passed, and the current correct way to do a "blend" is a Douglas Self design. This is ruler flat unity gain for in-phase signals, but has a steep cut and notch for out of phase signals. The intention being to control warps in vinyl records (which are out of phase) from getting to the loudspeakers, while letting music content through. The notch at 8.5Hz is around the arm/cartridge resonance, See http://www.douglas-self.com/ampins/Paris 2016 devinyliser.ppt . I actually have this incorporated into my active crossover, and leave it switched in permanently. Some CD's have out of phase extremely low frequency - possible from studio air conditioning.
Also for balanced output, things have also moved forward in the last 40 years. The THAT 1646 balanced line driver takes quite a bit of beating. Two of these in current boosted mode cheerfully drive many hundreds of feet of balanced cable into a THAT 1246 balanced receiver at over 27dBu. 7ppm distortion etc etc.
There are two odd features in Burwen's design, at least one of which found itself into the Audio Palette.
The first is the blender. This is an odd beast indeed. It is basically cross coupled bandpass filters. The intention is to connect the two channels together at 20Hz and frequencies close to it. The bandpass filter is tuned to 20Hz with a Q of 0.7. The rationale is attached, both from Burwen's design study, and also from the Palette manual. Note that at 20Hz there is a significant boost of around 4dB with a stereo signal, and the Palette manual says that the boost at 20Hz can be counteracted by tweaking the 15Hz control down by a couple of dB.
Note that there is a mistake in the excerpt from Burwen in the attachment. At one point he says 120Hz, when he actually means 20Hz.
The Blend control was not incorporated into any other Cello product, so its usefulness must have been limited.
The second peculiarity in Burwen's design is the very odd balanced output drive. That has a line that feeds back the -ve balanced signal "to counteract common mode noise with long cables". Nothing I have seen before. I do not know whether this was incorporated into the Audio Palette.
Decades have passed, and the current correct way to do a "blend" is a Douglas Self design. This is ruler flat unity gain for in-phase signals, but has a steep cut and notch for out of phase signals. The intention being to control warps in vinyl records (which are out of phase) from getting to the loudspeakers, while letting music content through. The notch at 8.5Hz is around the arm/cartridge resonance, See http://www.douglas-self.com/ampins/Paris 2016 devinyliser.ppt . I actually have this incorporated into my active crossover, and leave it switched in permanently. Some CD's have out of phase extremely low frequency - possible from studio air conditioning.
Also for balanced output, things have also moved forward in the last 40 years. The THAT 1646 balanced line driver takes quite a bit of beating. Two of these in current boosted mode cheerfully drive many hundreds of feet of balanced cable into a THAT 1246 balanced receiver at over 27dBu. 7ppm distortion etc etc.
Last edited:
Aha - the peculiar output drive in Burwen's design is called a Ground Cancelling output. According to Self (Small Signal Audio Design, Ed 2, 539-544) this was introduced in mixing consoles in the early 80's. Essentially they are only useful when a balanced output drives an unbalanced input at the remote end, and the ground cancelling arrangement dramatically improves hum issues due to dissimilar chassis hum potentials.
Last edited:
Further update - ground cancelled or ground compensated outputs can equally be used with balanced inputs at the remote end. See part way down this https://sound-au.com/articles/balanced-io.htm
I compared various order filters in this, subjectively. I disagree that steep slopes are desirable. In various Phono-Stages I included such a "blend filter" (defeatable naturally), first order. It is very effective in stopping woofer wobble and is otherwise inaudible. Try a 1st order slope with ~ 80Hz for fun. This is for LP of course, not Seedee.
It does? In what way?
And that makes it sound good?
Thor
Thorsten - try open baffle speakers before you make comments regarding woofer wobble. 6dB/octave might OK for sealed box speakers - which have a second order 12dB/octave fall in any case - so you end up with 18dB/octave by default with a 6dB/octave filter.
The only thing that the 1646 fails with is noise. At -101dBu at the output it essentially works out at an input referred noise of 44nV/rootHz - which is pretty lousy. But for professional applications where driving 1000 feet of screened twisted pair is typical, noise is not such a great disadvantage. If noise was a lot lower I'd use them in a heartbeat.
Personally I use vanilla 5532's to drive the short balanced runs in my system, and more of them in low noise balanced receivers. But in anything I am designing more recently I'm using OPA828 and OPA2828. With those in a balanced driver the output referred noise is 1.5uV - 2uV in 20kHz BW (-118.7dBu to -116.2dBu) and 24dBu into 600 ohm load with vanishingly small distortion. With the correct overall design that is.
Next question?
The only thing that the 1646 fails with is noise. At -101dBu at the output it essentially works out at an input referred noise of 44nV/rootHz - which is pretty lousy. But for professional applications where driving 1000 feet of screened twisted pair is typical, noise is not such a great disadvantage. If noise was a lot lower I'd use them in a heartbeat.
Personally I use vanilla 5532's to drive the short balanced runs in my system, and more of them in low noise balanced receivers. But in anything I am designing more recently I'm using OPA828 and OPA2828. With those in a balanced driver the output referred noise is 1.5uV - 2uV in 20kHz BW (-118.7dBu to -116.2dBu) and 24dBu into 600 ohm load with vanishingly small distortion. With the correct overall design that is.
Next question?
You mean speakers like this:
https://www.diyaudio.com/community/threads/are-you-open-baffled-yet.19883
?
Incidentally, below resonance reflex speakers are effectively "open baffle".
As for "true balanced" circuits done electronically, they are rarely beneficial in home audio.
Even in pro audio I personally would bet on an SE driver with an impedance only balanced output and a suitable well designed input.
They can be done discrete with noise and distortion levels the canned IC driver/receiver combo's cannot dream of at a small fraction of the cost of IC's, if we are willing to invest +/- 24V @ 25mA in the power supply (driver).
If driving really long lines, the isolation and EMI resistance afforded by real transformers becomes necessary anyway.
Thor
https://www.diyaudio.com/community/threads/are-you-open-baffled-yet.19883
?
Incidentally, below resonance reflex speakers are effectively "open baffle".
As for "true balanced" circuits done electronically, they are rarely beneficial in home audio.
Even in pro audio I personally would bet on an SE driver with an impedance only balanced output and a suitable well designed input.
They can be done discrete with noise and distortion levels the canned IC driver/receiver combo's cannot dream of at a small fraction of the cost of IC's, if we are willing to invest +/- 24V @ 25mA in the power supply (driver).
If driving really long lines, the isolation and EMI resistance afforded by real transformers becomes necessary anyway.
Thor