Phase Linear Autocorrelator
Phase Linear Autocorrelator (about $150 used). enlarge. This free website's biggest source of support is when you use these links, especially this link directly to the Autocorrelator at eBay (see How to Win at eBay), when you get anything, regardless of the country in which you live. Thanks! Ken.
Rear, Phase Linear Autocorrelator. enlarge.
The Phase Linear Autocorrelator Noise Reduction and Dynamic Range Recovery System is a signal processor that often makes good music sound better. Don't knock it until you've auditioned it; today in 2011, it really does make many good recordings sound even better.
It was wildly expensive when new in the 1970s, about $1,500 corrected for inflation, and has only a subtle but powerful effect. These are for careful listeners, not for impressing people with quick demos in the "retail" stereo stores of old, so few people actually bought them. Today they are a bargain, about $150 used, and if you know How to Win at eBay, I got this one for only $29.
The Phase Linear Model 1000 and Model 4000 are very subtle processors, which is their magic. What they do won't be obvious unless you listen carefully, and if you listen carefully, these can add more to your music than just about anything else.
These brilliant circuits were envisioned and designed by perennial audio genius Bob Carver, founder of Phase Linear in 1970, and then Carver Corporation in 1979, as well as today's Sunfire, founded in 1994.
These processors were included in the Phase Linear Model 4000 preamplifier, and the same were also sold as the stand-alone Model 1000 processor reviewed here.
These processors are:
The Autocorrelator is a single-ended noise reduction processor. It's not really an autocorrelator; "Autocorrelator" was used as a fancy marketing name for a patented and very clever system of noise gates.
The Autocorrelator greatly reduces noise in tapes, FM radio, CDs and SACDs recorded in the 1970s and before. I haven't tried it with LPs, with which it was originally designed to work best.
The Autocorrelator is amazing: it really does reduce noise significantly, doesn't affect the music or frequency response and has no audible artifacts.
It just makes noise go away. It has separate circuits to remove hiss and to remove rumble.
Dynamic Range Recovery System
The Dynamic Range Recovery system is really two processors controlled with one knob, a Peak Unlimiter and Downward Expander.
It has a subtle effect. It improves immediate dynamics, adding life and impact to most sources, without adding any audible artifacts.
It is quite different from the processors sold by dbx, which have a stronger effect overall, but less effect note-to-note. Instead of expanding by a small amount, like 1.5:1 or less, over a large dynamic range, this Phase Linear expands by a larger about, up to 2:1, over only a very narrow dynamic range. This lets this Phase Linear processor add more to most recordings with no artifacts, but it's also so subtle that few people were sophisticated enough to hear it, and thus dbx won out in the market while this excellent processor went out of production.
This Phase Linear system is better for use most of the time, and is better for trying to get some note-to-note life back out of dynamically compressed recordings.
Rated Specifications top
3 VRMS maximum.
3 VRMS or more into 2 kΩ.
20 - 20,000 Hz ±1dB.
Less than 0.25%.
Passive subsonic filter: - 35 dB at 5 Hz.
Low frequency noise reduction starts at 200 Hz and reaches a maximum of 20dB at 20 Hz.
High frequency noise reduction starts at 3 dB at 2kHz, and increases to 10 dB at 4 kHz and above.
Weighted overall noise reduction: 10dB from 20Hz to 20kHz.
Peak Unlimiter and Downward Expander
This processor has two sections that work together; the Peak Unlimiter and the Downward Expander. They are controlled by the same switch and control knob.
Their processing window is 35dB wide. The upper and lower thresholds are simultaneously varied by the front panel Peak Unlimiter Down Expander control.
Maximum gain change: +1.5 dB.
Amplitude-attack threshold (nominal input): 2 volts peak.
Maximum rate of processor change: 0.5 dB/µS.
Maximum gain change: -6 dB in two components: linear and downward.
The linear component is a 1.1:1 linear expansion from 0 dB to -30 dB, or -3 dB total at -30 dB.
The downward expander drops an additional -3 dB starting at -35dB and is fully active at -41dB.
8 integrated circuits ("ICs").
6-foot rubberized cord.
1/4 amp AGC fuse.
9.5 x 5 x 11.8 inches, WHD, without wood case.
6 pounds (2.7 kg).
Three years, parts and labor.
This crazy old piece of gear works great today in 2011 for improving most recordings made before the invention of Dolby A noise reduction in the early 1970s, as well as for FM radio reception.
Additionally, the Dynamic Range Recovery System makes squashed (dynamically compressed) recordings of any vintage sound better.
If you have perfect recordings, this processor won't improve them, but most older and new recordings will benefit. Recordings made before the advent of professional Dolby A noise reduction in the 1970s benefit from the Autocorrelator's hiss removal, and most of today's hottest new music has been so dynamically compressed that the Dynamic Range Recovery System helps breath some life back into squashed recordings.
All the processing is identical between channels; the VCAs are ganged for stereo.
There are almost no operator adjustments. Thank goodness, Bob Carver took the time to design this so the adjustments are already where we need them.
The Autocorrelator's really does just make noise go away.
It has separate circuits to remove hiss and to remove rumble.
The Autocorrelator has a huge effect on noise, and no effect on the music. It's a win-win situation; there are no artifacts.
The Autocorrelator only squelches the highs (and rumble) with no dynamic modifications to the music itself.
Believe it or not, it magically removes noise without affecting the music. This has the wonderful side benefit of sometimes making the music slightly smoother, mellower, warmer and more velvetier than before, all without affecting the music's timber or tonal balance! The background just gets darker and the music sounds more like live music.
It only has 10 dB of noise reduction on purpose. This is perfect to get rid of the noise, and doesn't have enough range to add any audible pumping.
It's great for FM radio, and music recorded before the 1970s, as well as classical music recordings today made with room noise & rumble.
For instance, its effect is astonishing on 1950s Mercury Living Presence CDs. The noise just goes away, and the music is left immaculate.
When recordings are remastered today, the mastering engineer has his choice of many more noise-reduction tools, but often doesn't use them to keep the recording as original as possible. For instance, the 2009 reissue of 1969's Abbey Road sounds great, and what little noise it has is removed by this Autocorrelator without affecting the music.
If you hear any artifacts, it probably was in the recording to begin with. Most people have forgotten what was modulation noise, which is what happens with analog tape. The only time I've been able to imagine even the slightest pumping is on early rock mixes that hard-pan a solo instrument to one side with silence on the other, and since the Autocorrelator's filters are programmed to work identically on both channels at once, the empty channel's noise could follow the other channel — but it still wasn't obvious.
This noise reducer is genius, if you've got an older analog-sourced recording. With good digital recordings, there is no noise to reduce.
Dynamic Range Recovery System
The Dynamic Range Recovery System, which consists of the Peak Unlimiter and Downward Expander together, is very subtle. You won't hear it working; it has a total maximum adjustment range of only -6 dB to +1.5 dB!
This subtlety is deliberate. It is designed this way so that it doesn't tire us with artifacts. Its subtle expansion doesn't do much on the grand scale as do dbx expanders like the dbx 3BX. The Phase Linear expander is superb for restoring note-to-note impact, without needing to adjust our volume controls continuously as the piece continues.
The expander always has flat frequency response; it only modifies the dynamics slightly. It has no large gain shifts, but plenty of instantaneous improvement.
This instantaneous improvement means that drums come forward a bit with just a little more bite, and the the background stays about the same, while softer sections (be they long or short) get just a little bit softer.
This subtle processing is just right; it's always musical and always improves most recordings.
Of course if you've got a perfect recording this won't improve it, but for most commercial recordings of all vintages, as well as radio, it helps.
There is some low-level circuit noise (see measurements) added by the Dynamic Range Recovery System, but it's only audible if you've got a recording with enough dynamic range not to need the processor in the first place.
While most often seen as shown, there was an optional walnut case into which this slid.
The Phase Linear Model 1000 uses all fiberglass epoxy boards. It's loaded with them; unlike modern CD players which are all air inside, the Model 1000 box is loaded with circuit cards, which define the size of the case.
There is a main board along the bottom, and several other boards plug into the main board.
This $29 sample, sold as "found at a garage sale and untested," shipped cross-county improperly packed, and with a Goodwill thrift store sticker from October 2000, had had one of its boards wiggle loose and arrive banging around inside the case.
The problem with this design is there there's no serious board support, and it expects the solder joints to keep the boards vertical. Over the decades some of these solder joints can crack from the flexing, which are easy for a licensed technician to resolder.
The Power switch is on the wrong side, at the far right of the lower pushbuttons.
There is a teflon buffer strip to keep the pushbuttons from scraping on the punch-out in the aluminum face.
The LEDs are driven from DC, so they stay lit a second or two after the power is removed.
As a musical signal processor, it's supposed to add distortion when tested with sine waves as shown here. The gain-changing circuitry which is optimized for processing music changes gain as well on pure sine waves, and this gain-changing appears to be distortion when tested in a laboratory. What appears as distortion here is often the exact euphonic distortion this device is designed to add.
All in all, tubes amplifiers and LPs distort the music even more than this processor, and audiophiles pee in their pants with pleasure at these distortions.
I used a state-of-the-art Rohde &Schwarz UPL for these measurements.
The traces are color coded for the Left Channel and the Right Channel.
Gain vs frequency, Phase Linear 1000, processing bypassed, 200mV sine wave input. (measurement: Rohde & Schwarz UPL.)
I didn't separate the traces for clarity; there really is an 0.2 dB imbalance.
THD, Phase Linear 1000, processors bypassed, 200mV sine wave input. (measurement: Rohde & Schwarz UPL.)
All these measurements are made with the Correlation Threshold and Low Frequency Calibration knobs (on the right) at midpoint.
THD, Phase Linear 1000 Autocorrelator, 200 mV swept sine wave input. (measurement: Rohde & Schwarz UPL.)
This is as expected; the Autocorrelator works in several bands, and two bands come together at around 200-300 Hz. We won't necessarily see this with music, but with a sine wave, it can cause the unit to wiggle gains and appear to add THD.
Likewise, the worst way to measure frequency response is with a swept sine wave, as it triggers the noise reduction filters in weird ways. With this caveat, it looks great at 1 V RMS:
Phase Linear 1000 Autocorrelator frequency response, 1V RMS swept sine wave input. (measurement: Rohde & Schwarz UPL.)
The kink at 300 Hz won't be seen with music; it's a artifact introduced by testing with sine waves.
Let's reduce the input level to 200 mV RMS:
Phase Linear 1000 Autocorrelator frequency response, 200mV RMS sine wave input. (measurement: Rohde & Schwarz UPL.)
As we reduced the input level to 200 mV RMS, more of the bands are getting mistriggered by the sine waves, leading to what looks like a rougher response. The lowest frequencies are also starting to get reduced. This isn't representative of how it will sound with music; it's more of an educational exercise to help us see what's going on inside:
Phase Linear 1000 Autocorrelator frequency response, 20mV RMS sine wave input. (measurement: Rohde & Schwarz UPL.) 2 mV curve is the same.
With a lower-level 20 mV RMS sine wave, we can see the low-frequency noise circuit kick in, reducing noise and rumble below 200 Hz.
Let's now see what happens with a noise input, which better simulates performance with music:
Phase Linear 1000 output spectrum, 10 mV RMS noise input. (measurement: Rohde & Schwarz UPL.)
Here we've confirmed that the high frequency response really is flat with musical signals. The 10 mV RMS input noise, which is -40 dBV, is enough to open the gates for the high frequencies, but little enough to close the low-frequency gate.
Let's turn the threshold control completely clockwise and see if anything changes. It shouldn't:
Phase Linear 1000 output spectrum, autocorrelator maximum CW, 10 mV noise input. (measurement: Rohde & Schwarz UPL.)
As expected, since the high frequency gates were open at midpoint, turning the threshold control clockwise didn't change anything.
Now lets turn the threshold control fully counterclockwise, which is supposed to force the high frequency gates to squelch the signal:
Phase Linear 1000 output spectrum, autocorrelator maximum CCW, 10 mV noise input. (measurement: Rohde & Schwarz UPL.)
Aha! This is the frequency response of the autocorrelator at the instants it's eliminating noise.
Exactly as specified, there is about 10 dB of noise reduction above 4 kHz.
Peak Unlimiter & Downward Expander
The Peak Unlimiter and Downward Expander are two circuits, but since they are controlled by the same switch and knob, we'll treat them as one.
My Unlimit Threshold light was always lit, suggesting that my Goodwill > garage-sale > eBay sample isn't peak unlimiting. As shown below, it peak unlimits as triggered by level, but I wasn't able to trigger the unlimiter with impulses. I don't know if just my light is defective, or if my peak unlimiter is only half working. It works great in any case, and works over its fully-specified -6 dB to +1.5 dB range.
These are measured with their threshold control (the left knob) at midpoint unless otherwise noted.
Phase Linear 1000 Unlimiter gain versus frequency, 200mV RMS sine wave input, 5 Hz - 22 kHz. (measurement: Rohde & Schwarz UPL.) It measures the same at levels from 20mV to 1V RMS.
The subsonic rolloff is deliberate.
Phase Linear 1000 Unlimiter THD versus frequency, 1V RMS sine wave input, 40 Hz - 22 kHz. (measurement: Rohde & Schwarz UPL.)
This is better than specified.
Let's plot its gain changes versus input level. In these curves, the X-axis (lower scale) is the input level in V RMS, and the Y-axis (left scale) is the processor's gain. The input is a 440 Hz (musical note "A") sinewave. In these tests, I swept the signal up in amplitude, but not down:
Phase Linear Model 1000 Peak Unlimiter and Downward Expander gain versus 440 Hz sinewave input level in dBV, control at minimum (CCW). (measurement: Rohde & Schwarz UPL.)
Set to minimum, we never get to the threshold for any gain increase. All the processor does is lower gain by 6 dB.
Phase Linear Model 1000 Peak Unlimiter and Downward Expander gain versus 440 Hz sinewave input level in dBV, control at 9 o'clock. (measurement: Rohde & Schwarz UPL.)
With the Threshold control turned up a little, we can now see that the gain starts to increase at -3 dBV.
Phase Linear Model 1000 Peak Unlimiter and Downward Expander gain versus 440 Hz sinewave input level in dBV, control at 10:30 o'clock. (measurement: Rohde & Schwarz UPL.)
Turned up a little more, we now see that the expander increases its gain as the input signal increases from -14 dBV to -8 dBV. In other words, the expander is doing its best work at this setting if your input is around these levels.
Phase Linear Model 1000 Peak Unlimiter and Downward Expander gain versus 440 Hz sinewave input level in dBV, control at midpoint. (measurement: Rohde & Schwarz UPL.)
With the threshold control at its midpoint, the expander is most active between -20 dBV and -18 dBV.
Phase Linear Model 1000 Peak Unlimiter and Downward Expander gain versus 440 Hz sinewave input level in dBV, control at 1:30 o'clock. (measurement: Rohde & Schwarz UPL.)
With the Threshold control advanced past 12 o'clock, we see the first instance of the Peak Unlimiter increasing gain for input levels above -12 dBV, while the expander is most active between -25 dBV and -18 dBV.
Phase Linear Model 1000 Peak Unlimiter and Downward Expander gain versus 440 Hz sinewave input level in dBV, control at 3 o'clock. (measurement: Rohde & Schwarz UPL.)
Cranked most of the way up, the peak unlimiter is active above -17 dBV, and the expander is most active between -32 dBV and -24 dBV.
Phase Linear Model 1000 Peak Unlimiter and Downward Expander gain versus 440 Hz sinewave input level in dBV, control at maximum (full CW). (measurement: Rohde & Schwarz UPL.)
Cranked to the max, the peak unlimiter is active above -20 dBV, and the expander is most active between -40 dBV and -27 dBV.
All Processors Active
Phase Linear 1000 THD, unlimiter and autocorrelator together, 1V RMS sine wave input. (measurement: Rohde & Schwarz UPL.)
Frequency response looks funny as explained above because the autocorrelator is getting mistriggered because this test is with sine waves, not music.
Phase Linear 1000 THD, unlimiter and autocorrelator, 1V sine wave input. (measurement: Rohde & Schwarz UPL.)
This looks like nasty THD, but in fact, it's not audible with music. It's mostly artifacts of hitting a music processor with sine waves instead of music. It reduces at lower levels:
Phase Linear 1000 THD, unlimiter and autocorrelator, 200mV sine wave input. (measurement: Rohde & Schwarz UPL.)
Hum and Noise
60 Hz hum can be a large component, at about -75 dBV unweighted, if the unpolarized plug is inserted the wrong way into a power socket. Reverse the plug to minimize it back to -100 dBV.
Noise is inaudible with the processors bypassed.
With a 16-bit 1V RMS FS source muted and playback gain increased, there is just a little bit of extra noise audible when the Autocorrelator switched in, and plenty of extra noise audible when the Peak Unlimiter is switched in. In actual use the Autocorrelator noise is never audible, and while the Peak Unlimiter noise can be audible with very quiet recordings, recordings that good don't need the Peak Unlimiter. In other words, the only audible noise comes from the Peak Unlimiter when used with exceptionally good recordings, and those recordings don't need the Peak Unlimiter.
The RMS noise levels are:
For reference, an iPod's maximum full-scale output is 1 VRMS (0 dBV), and a CD player's full-scale output is 2 VRMS (+6 dBV). One LSB at 16 bits is therefore about -96 dBV from an iPod, and about -90 dBV from a CD player.
Here are 8k-point FFTs of the A-weighted noise spectra. Roll your mouse over the processor curves to compare against the noise in bypass mode:
Phase Linear 1000 output noise spectrum, autocorrelator and unlimiter bypassed, no input.
Phase Linear 1000 output noise spectrum, autocorrelator only at midpoint, no input. (mouse-over to compare to bypass)
Phase Linear 1000 output noise spectrum, expander only at midpoint, no input. (mouse-over to compare to bypass)
Phase Linear 1000 output noise spectrum, expander and autocorrelator both at midpoints, no input. (mouse-over to compare to bypass)
Dynamic Range versus Data Compression
Dynamic range compression is when the differences between loud and soft are reduced. After dynamic range compression, usually just called "compression" or "limiting," the peaks and valleys of louds and softs that give music its life get squashed into sounding all the same.
The Phase Linear Model 1000 and Model 4000 work to correct the degradation added from deliberate dynamic range reduction, or compression. This is added deliberately in the process of recording and mastering to make CDs sound louder, to make it easier on the mixer as the vocalist moves around, and to help smooth out sloppy orchestrations. Radio has been adding even more dynamic range compression as each station tries to make itself the loudest on the dial.
Data compression is completely different. Data compression is done in computers to be able to store or transmit more digital music in the same amount of memory or transmission bandwidth. Data compression is what's done to take uncompressed AIFF audio files and convert them into smaller AAC or MP3 files. The Phase Linear 1000 or 4000 have nothing to do with data compression, or reducing its effects.
I don't know of any consumer product that does as good a job, or even comes close, to how well this Phase Linear reduces noise without harming the music.
This is a single-ended system that works with all recordings. It it not a double-ended system like the Dolby or dbx tape-noise reduction systems, which process a signal before it's recorded on tape, and then again on playback, to eliminate tape hiss.
Today there should be many stand-alone or Pro-Tools plug-in digital processors to do the same thing, but better. I haven't researched them.
Compared to a dbx 3BX, the effect of the Phase LInear expander is very subtle. It therefore also doesn't add any artifacts, which means that you'll still be enjoying this unit long after you put the dbx away in the closet.
I prefer this Phase Linear for peak expansion over the Impact Restoration of the dbx 3BX III. The 3BX III's Impact Restoration never seems to do much except to add audible artifacts to modern squashed recordings that so dearly need impact restoration.
Most expanders add a lot of gain change over a broad range of levels. By comparison, the Phase Linear adds what it adds only over only a very narrow dynamic range. This is how it can add so much back into so many recordings without any artifacts.
The Phase Linear's peak expander is subtle, and always improves the music.
Phase Linear Autocorrelator. enlarge.
When both processors are bypassed, this Phase Linear Model 1000 buffers the signal, so if power is off, the signal won't go though the unit.
Put on your recording, kick-in the Autocorrelator, turn the Auto Correlator knob all the way clockwise, and listen for the noise. Turn up the treble temporarily if it helps you hear the noise.
As you turn the control counterclockwise, you'll hear a point at which the noise suddenly vanishes, but the music is untouched. Turn it farther, and the music suddenly will sound duller.
Leave the knob halfway between these points.
The Low Frequency Calibration control adjusts the rumble reducer. Unless you're still playing records, leave this turned completely clockwise.
To reduce turntable rumble (or room noise on classical music recordings), turn the Low Frequency control more counterclockwise. Just like the Threshold control, it should suck out the rumble and leave the bass untouched. If you turn it too far counterclockwise, it will suck-out your bass.
Use this for older recordings or radio that needs it. For playing digitally recorded (DDD) CDs and modern recordings that don't have any noise, leave it off.
Dynamic Range Recovery System
The Dynamic Range Recovery System is activated with the "PEAK UNLIM." button, and controlled only by the PEAK UNLIMITER DOWN EXPANDER UNLIMIT THRESHOLD control.
The control seems to do almost nothing as rotated. All it does is change the input level range over which the processor is most effective, which means that all it seems to do is make a very slight change in level.
Phase Linear says to rotate the control until the Unlimit Threshold light blinks with the music. My light is stuck, so I rotate it so that the overall level is about the same with the Peak Unlimiter button IN or OUT.
The audible effect of this knob is that it gets slightly louder when rotated clockwise, and slightly softer when rotated counterclockwise. The key to setting the expander's control is to set it to a point at which the processor will be varying its gain to expand the music both up and down, and not simply crank it so that everything is the same, but just a little louder.
Hey, for $150 or so, the Phase Linear Model 1000 adds a lot to many recordings, and never adds any artifacts.
It's most obvious for careful listening. For casual listening, many people won't hear anything at all. For careful listeners, this is the key to pulling the music out of the noise, and restoring some of music's life back after its been squashed by the recording and broadcasting process.
If you find all the effort I've put into this report helpful, this free website's biggest source of support is when you use these links, especially this link directly to the Autocorrelator at eBay (see How to Win at eBay), when you get anything, regardless of the country in which you live. Thanks! Ken.
Phase Linear Autocorrelator Model 1000 Series Two. enlarge.
Rear, Phase Linear Autocorrelator Model 1000 Series Two. enlarge.
Top, Phase Linear Autocorrelator Model 1000 Series Two. enlarge.
Inside of Model 1000 Series Two after 120V conversion. enlarge.
Much quieter than Model 1000:
Te noise levels oddle don't vary with settings.
Noise Levels, A-weighted:
Expander: -93.3 dBV.
Autocorrelator: -97.6 dBV.
Both: -93.0 dBV.
Neither: -119.2 dBV (residual of analyzer).
Bypass is a hard bypass; works even with power off.
The 240V Series II model has a dual-primary power transformer.
As shipped for 240V, the blue wire is fed in a loom towards the power switch, and connects to the brown/yellow wire near the power switch. This connects the two primaries in series.
For 120V operation:
1.) Cut the solid blue wire where it's connected to the brown/yellow wire in the loom near the power switch.
2.) Cut to length, and connect this brown/yellow wire to the same point on the power switch as the blue/yellow wire.
3.) Pull the solid blue wire out of the harness that goes to the power switch, cut it to a shorter length and instead connect it to the same fuse terminal as the solid brown wire. You'll have a few inches of solid blue wire left over.
Replace the 250V AGC 1/16A fuse with a 250V AGC 1/8A fuse.
Now it draws 65 mA at 120V, and you're done.
In summary, for 240V, the line is applied to the blue/yellow and solid brown wires, while the solid blue and brown/yellow wires are connected to each other. For 120V operation, the two solid wires are connected to one side of the line, and the two yellow-striped wires are connected to the other side of the line.
Help me help you top
I support my growing family through this website, as crazy as it might seem.
The biggest help is when you use any of these links to Adorama, Amazon, eBay, B&H, Ritz, Calumet, J&R and ScanCafe when you get anything, regardless of the country in which you live. It costs you nothing, and is this site's, and thus my family's, biggest source of support. These places have the best prices and service, which is why I've used them since before this website existed. I recommend them all personally.
If you find this page as helpful as a book you might have had to buy or a workshop you may have had to take, feel free to help me continue helping everyone.
As this page is copyrighted and formally registered, it is unlawful to make copies, especially in the form of printouts for personal use. If you wish to make a printout for personal use, you are granted one-time permission only if you PayPal me $5.00 per printout or part thereof. Thank you!
Thanks for reading!