It's in Fremer's Analog Corner, February 2012. I just read it twice through and saw no mention of any evil opamps
128 bits, looks like they are trying to quantify the position and momentum of each vinyl molecule as it passes the stylus. Don't they know they can't do that?
Wadax looks like an exercise in over-the-topness.
We are saying the same thing but coming at it from 180 degrees.
I don't blindly apply the theory for BSC. It will as you cleary point out, sound like doo-doo. I start there, build the prototype and sit in my room measuring at 1M playing with the parts until I get it resolved. The theory (web spreadsheets) only give the answer for nearfield on an analyzer. Not in the room. Voicing is always finalized by listening.
Most "modern" speakers are made to all look alike and have widths based on nominal 6" drivers, so they tend to have similar points where they shift 2pi-4pi. Mine are not always nominal. We are always guessing, as we almost never know the room. I wind up with about half of the cut of nominal calculated values but pretty close to the frequency.
Someday I am going to try a flush line array and see what a pure 2-pi behaves like. In theory, you can get rid of a lot of refraction issues too.
A tone control could be used to do the same thing IF you could adjust its F3. ( OK, a very few folks were smart enough to do this) You can't, so BSC is not the same as a tone control. The mass market does not care because they use their push-button eq garbage generators and most high end no longer even has tone controls, so it is up to the speaker builder to do his best.
Playing with Spice, dangerous as not real, OK.
Modeled three ccs into the IPS of the Hafler 120. One simple diode biased, one a current mirror, one the feedback version. The 120 has a resistor in the tail so I already had something to measure the current through.
I found the transient response to a square wave input to be the best with the diode biased version. But, it's behavior varied wildly with frequency. The CM behaved not as well with the transients, and even uglier with frequency. The two transistor feedback version seemed to be far better at both. The SPICE model for a ccs plotted perfectly to the transients and to AC analysis.
Then I looked at the FFT for a 20K sine wave at rated power. Distortion was about 6db worse for the feedback pair version. I was not expecting that. I was expecting slightly lower second harmonic. If it did nothing, I would not have been surprised as the rest of the amp is "perfect". Not quite sure what to take away.
I also tried just separating the IPS and VAS ccs's. It seemed to improve the transient response, but nothing showed up on the FFT as different. Changing to a green LED and increasing the emitter resistor also made the simulation prettier, but again, nothing on the FFT.
Modeled three ccs into the IPS of the Hafler 120. One simple diode biased, one a current mirror, one the feedback version. The 120 has a resistor in the tail so I already had something to measure the current through.
I found the transient response to a square wave input to be the best with the diode biased version. But, it's behavior varied wildly with frequency. The CM behaved not as well with the transients, and even uglier with frequency. The two transistor feedback version seemed to be far better at both. The SPICE model for a ccs plotted perfectly to the transients and to AC analysis.
Then I looked at the FFT for a 20K sine wave at rated power. Distortion was about 6db worse for the feedback pair version. I was not expecting that. I was expecting slightly lower second harmonic. If it did nothing, I would not have been surprised as the rest of the amp is "perfect". Not quite sure what to take away.
I also tried just separating the IPS and VAS ccs's. It seemed to improve the transient response, but nothing showed up on the FFT as different. Changing to a green LED and increasing the emitter resistor also made the simulation prettier, but again, nothing on the FFT.
Reminds me of the gravitational-wave researchers, who developed "quantum back-action-avoidance" strategies to get better energy resolution at the expense of poorer time resolution, "cheating" the uncertainty principle.
What astonishes me: who/what billionaire funded this? Who has that sort of money to burn? Custom ICs? 480-pin ones (or BGA more likely)? Measurements of cutter heads for phase response??
Very hard to imagine that the chips are custom.
Note that they somehow divine what the overload requirements will be and set the gain to as high as possible while avoiding overload. Say what? Does it have a precognitive dirt clod module? Although I suppose if you are spending 32.5k US you probably have a pretty decent record-cleaning system.
Last edited:
Try a big bypass cap across the diodes/LED and see what happens. Be sure to let the system settle before the distortion measurement.
Giant FPGA's they have become almost unbelievable in their sophistication (and cost). I've seen some for 10G backplane apps at $2000 a pop (that price did include some IP).
You are right, who would mount such a technological assault on playing LP's while those in the know still seek round stilii.
Last edited:
Hi,
I did not say it will sound like doo-doo. I said it will have a 6dB bass boost in the power response if the anechoic response is made flat at LF.
And I pointed out that the "Baffle Step Correction" makes sense for systems that are routinely (and only) played at SPL's that are lower than those common in recording/mastering environments.
In fact, I'm great fan and proponent of correctly applied physiolgical equalisation (aka Loudness). Well, as a German I should, many of the "GöbelsGob" Propaganda Radio's made in Germany in the 30's and 40's had loudness (and anything following later), the faders used for gain-riding during recording had it (calibrated of course) and the studio monitoring level controls had it.
I have only one requirement for a "loudness" equalisation and that is an off-switch, which is not present if it is designed into the speakers frequency response.
Note, I was not referring to tone controls.
You can do BSC using a simple passive network looped into the tape loop (if your pre has such) and you also can construct a physiologically equalised -15dB pad there to correct for appx. 70dB SPL.
But a proper loudness that correctly changes as you adjust the volume is better. BTW, this does imply the presence of "gain" controls and metering circuits to equalise out differences in source levels as well as amplifier/speaker relative gain.
I guess that "best" would be to construct a control centre with the necessary functions and leave the tome and loudness controls there and defeatable, possibly add a few semi-parametric cut only LF EQ's to EQ room modes as well.
Actually modern all digital AV receivers have all this, but sadly often the usability and sound quality is usually beastly.
Ciao T
I did not say it will sound like doo-doo. I said it will have a 6dB bass boost in the power response if the anechoic response is made flat at LF.
And I pointed out that the "Baffle Step Correction" makes sense for systems that are routinely (and only) played at SPL's that are lower than those common in recording/mastering environments.
In fact, I'm great fan and proponent of correctly applied physiolgical equalisation (aka Loudness). Well, as a German I should, many of the "GöbelsGob" Propaganda Radio's made in Germany in the 30's and 40's had loudness (and anything following later), the faders used for gain-riding during recording had it (calibrated of course) and the studio monitoring level controls had it.
I have only one requirement for a "loudness" equalisation and that is an off-switch, which is not present if it is designed into the speakers frequency response.
Note, I was not referring to tone controls.
You can do BSC using a simple passive network looped into the tape loop (if your pre has such) and you also can construct a physiologically equalised -15dB pad there to correct for appx. 70dB SPL.
But a proper loudness that correctly changes as you adjust the volume is better. BTW, this does imply the presence of "gain" controls and metering circuits to equalise out differences in source levels as well as amplifier/speaker relative gain.
I guess that "best" would be to construct a control centre with the necessary functions and leave the tome and loudness controls there and defeatable, possibly add a few semi-parametric cut only LF EQ's to EQ room modes as well.
Actually modern all digital AV receivers have all this, but sadly often the usability and sound quality is usually beastly.
Ciao T
Last edited:
Hi,
As Brad suggested for your Sim (and I earlier for real Amp's) add a big cap across the diodes.
Next replace one of the diodes with a 6.8V Zenner or LM329 and adjust the emitter resistor (keep the big cap) and see what happens.
Finally connect a BF244C (or other low Gm, high pinchoff voltage Fet) as cascode for the BJT...
The input stage is where signal and feedback signal mix. If there is ANY nonlinearity in the tail, you will know about it...
FFT is a steady state analysis... Others can be made.
You may wish to evaluate open-loop linearity with a sawtooth wave...
Ciao T
As Brad suggested for your Sim (and I earlier for real Amp's) add a big cap across the diodes.
Next replace one of the diodes with a 6.8V Zenner or LM329 and adjust the emitter resistor (keep the big cap) and see what happens.
Finally connect a BF244C (or other low Gm, high pinchoff voltage Fet) as cascode for the BJT...
The input stage is where signal and feedback signal mix. If there is ANY nonlinearity in the tail, you will know about it...
FFT is a steady state analysis... Others can be made.
You may wish to evaluate open-loop linearity with a sawtooth wave...
Ciao T
Hi,
Available widely since 1978, available to some large customers (e.g. major labels) before that.
Note, some specialist cutting houses did not use the delay and the cutting guys I knew in east germany NEVER used it (they observed that records cut this way sounded like duh-duh), but listened carefully several times to the whole recording and then controlled levels and groove spacing manually.
But you about right, since the 80's most LP's pressed by majors should have had the "digital" warning label prominently displayed.
BTW, the digital delays are not the only horror story in LP cutting and LP replication. To be brutal, the concept of "fidelity" the way at least some here seem to consider it simply does not apply to LP. They are clearly way below "LoFi". Yet they can sound excellent. Is there a hidden lesson here?
Ciao T
Available widely since 1978, available to some large customers (e.g. major labels) before that.
Note, some specialist cutting houses did not use the delay and the cutting guys I knew in east germany NEVER used it (they observed that records cut this way sounded like duh-duh), but listened carefully several times to the whole recording and then controlled levels and groove spacing manually.
But you about right, since the 80's most LP's pressed by majors should have had the "digital" warning label prominently displayed.
BTW, the digital delays are not the only horror story in LP cutting and LP replication. To be brutal, the concept of "fidelity" the way at least some here seem to consider it simply does not apply to LP. They are clearly way below "LoFi". Yet they can sound excellent. Is there a hidden lesson here?
Ciao T
I wonder if the digital delay was actually in the audio chain? Perhaps it was fed audio to adjust the servo for moving the cutting head, with the acutal cutter head fed seperately. I am not an analogue mastering engineer but perhaps we have a few on the board who would know if this split audio path was the case. Regards
Hi,
Not great on cause and effect, are we?
You must delay the audio fed to the cutterhead, so that actua audio (not delayed) can be analysed for level and frequency content and then the groove spacing can be adjusted automatically so that the groves ALREADY have more spacing when the signal at the cutterhead demands higher excursions...
Ciao T
Not great on cause and effect, are we?
You must delay the audio fed to the cutterhead, so that actua audio (not delayed) can be analysed for level and frequency content and then the groove spacing can be adjusted automatically so that the groves ALREADY have more spacing when the signal at the cutterhead demands higher excursions...
Ciao T
Hi,
This is how it was done BDD (before digital delays), actually one tape but two heads. You needed a fair bit of tape between the two heads though (several seconds worth, which 15ips gets long quick). This is a fairly non-trivial affair and the works often jammed up (or so I was told).
The best solution is still a human who can read sheet music and follow along making manual adjustments. Stan Ricker still cut's in "fully manual" mode, or so I am told.
The automated systems introduced by Neumann are a bit like the compressors everyone seems to insist upon putting on the final mix (instead of gain-riding the mix to domestically usable dynamics). It means fairly unskilled people can operate the system and get decent results, but there always is a price to pay...
Ciao T
This is how it was done BDD (before digital delays), actually one tape but two heads. You needed a fair bit of tape between the two heads though (several seconds worth, which 15ips gets long quick). This is a fairly non-trivial affair and the works often jammed up (or so I was told).
The best solution is still a human who can read sheet music and follow along making manual adjustments. Stan Ricker still cut's in "fully manual" mode, or so I am told.
The automated systems introduced by Neumann are a bit like the compressors everyone seems to insist upon putting on the final mix (instead of gain-riding the mix to domestically usable dynamics). It means fairly unskilled people can operate the system and get decent results, but there always is a price to pay...
Ciao T
Will do. I had thought about that but my first attempts I was fooling myself by not looking at the right place. I also found a plug in to LTSpice that plots distortion across frequency. Easier than three of four separate plots. Still need to pick several levels.
I figure if I concentrate on such a "small" part while I get a better understanding of transistors, it will help much more further on. Spice has a lot of capabilities not obvious to an amateur and it just scratches the surface of reality. Lack of models and their completeness is one problem. I bet LT has a much better set that applies to on-chip parts with a lot more precision.
Got to do something waiting on my parts to complete a passive crossover, with BSC.
Oh, the comment about doo-doo. That is for those who build speakers out of perfect spreadsheet and web "design your crossover here" sites. Many do and and find out that is not all there is to it. For instance, the wider a baffle, the less deep I tend to make the BSC, not just the frequency. The pair I am doing right now is only 6 inches wide, and it took a lot of tries to get it happy. This particular design had a fixed aesthetic requirement so I could not play with the width and crossover points to achieve the same thing without the extra parts.
I am sure we agree, a studio monitor has a different job to do than my main listening speakers. The environment is different, solution will be different.
I figure if I concentrate on such a "small" part while I get a better understanding of transistors, it will help much more further on. Spice has a lot of capabilities not obvious to an amateur and it just scratches the surface of reality. Lack of models and their completeness is one problem. I bet LT has a much better set that applies to on-chip parts with a lot more precision.
Got to do something waiting on my parts to complete a passive crossover, with BSC.
Oh, the comment about doo-doo. That is for those who build speakers out of perfect spreadsheet and web "design your crossover here" sites. Many do and and find out that is not all there is to it. For instance, the wider a baffle, the less deep I tend to make the BSC, not just the frequency. The pair I am doing right now is only 6 inches wide, and it took a lot of tries to get it happy. This particular design had a fixed aesthetic requirement so I could not play with the width and crossover points to achieve the same thing without the extra parts.
I am sure we agree, a studio monitor has a different job to do than my main listening speakers. The environment is different, solution will be different.
The best way is to use digital delay once, recording resulting sidechain, then apply this sidechain to direct actual recording without any delays. Just 2-step procedure: adjust well then shoot once.
- Status
- Not open for further replies.