Driving long cables

[pstamler]

Okay, this is going to brand me as a naif in some ways, but here goes.

I was looking at output stages for mic preamps a couple of days ago, and was thinking about what a preamp should be expected to drive. It doesn't seem unreasonable to me for a professional preamp to be expected to drive 100' (30 meters) of cable. Sometimes we're a good distance away from wherever the recorder is, and in many studios we're looking at running wire through floors, walls, ceilings, etc.. 100' is probably extreme but not unheard of.

Okay, what is the load a 100' cable puts onto an amplifier? Looking at cable specs for something like Gotham GAC-2. It has a conductor-to-conductor capacitance of 38pF/ft (well, less than that, but they don't say how much less, so this is the specified maximum) and a conductor-to-shield capacitance of 58pF/ft (same deal). That adds up to 96pF/ft -- and since this is one of the lower-capacitance pro-audio cables out there, let's round that to 100pF/ft to make the arithmetic easier. (See question #1.) A 100' cable will therefore have about 10,000pF of capacitance, or 10nF.

What kind of signal am I looking at? Let's assume a standard pro-audio output, nominally +4dBu with 20dB of headroom. If this is an electronically-balanced circuit, each leg will be putting out a nominal -2dBu; the maximum output would then be +18dBu on each leg. That works out to about 6.156Vrms, or (assuming a sine-wave) about 8.7Vpk.

According to Walt Jung's concervative rule of thumb, a circuit should have 1V/us of slew rate for every 1V of peak output. So that means our output amplifiers should have a slew rate of 8.7V/us or greater. Not too hard to do with 5534s, OPA604s, etc.. Into a resistive circuit.

But this ain't a resistive circuit. It's got 10nF of capacitance in parallel with the load. So the output amplifier has to drive the capacitance without causing slew rate issues.

There's another rule of thumb for slew rate calculations:

V / us = mA / nF

That means that to get a certain slew rate in a capacitative load, you need a certain amount of current to charge a capacitor. The units compute right if you set the equation up as above. You can also rearrange it, if you want to figure out how many mA of charging current you'll need:

(V / us) * nF = mA

So for a slew rate of 8.7V/us and a capacitance of 10nF you'll need 87mA of current.

It's a bit worse than that, because you'll also need a bit of current to drive the resistive part of the load. Say that's 2.5k (half of a 5k load, not standard but if you check a Studer recorder that's the input impedance). That will take another 3.5mA or so. Figure the output amp has a 10k feedback network, and you need another another 0.9mA. So to drive your Studer recorder back there in the machine room, plus 100' of cable, youl'll need about 91.4mA of current.

Shit.

That's a *lot* of current, guys. Almost nobody these days builds preamps that can put that much out, period, never mind in staying in Class-A or staying clean or anything. It just ain't there. (For comparison, a 5534 opamp maxes out at 38mA.)

So that's what I'm looking at, and I have some thoroughly naive quesiotns:

1) Am I calculating the capacitance properly? Is it reasonable to add the conductor-to-conductor and conductor-to-shield capacitances? Or not?

2) Am I being unreasonable in applying Jung's strictest criterion to this application? Would I be more reasonable applying his looser rule (0.5V/us per peak output volt)?

3) What effect, if any, will there be from the 100-200 ohm resistor usually placed in series with outputs?

4) Have I just re-discovered why real pro equipment is big and heavy and runs hot? And why real pro output stages were made with discrete transistors and stepdown transformers (less voltage traded for more current)?

5) Is there something real obvious that should be staring me in the face but isn't?

6) And if not...what do y'all think is the best strategy for building an output stage that will drive a long cable? All-discrete? Op-amp with power transistors? Op-amp with high-current buffer a la Jung-Martell? BIG vacuum tubes and a whopper transformer?:

Peace,
Paul

PS This train of thought got started because I have a 78 transcription preamp in the living room and the computer is here in the workroom, with a 75' cable between them. The problem is a lot less demanding; my soundcard in this computer is set for -10dBV nominal sensitivity, and the input clips at about +8dBu, so the maximum voltage drive is about 2.75Vpk. Cable capacitance is, I'm guessing, about 75pF/foot (unbalanced), or about 5.6nF, so this would take about 15.4mA to drive to the appropriate slew rate, and even the TL071 in the preamp's output can manage that -- just.

 

[Samuel Groner]

www.rane.com/note126.html

There's an important difference between slew-rate limit due to internal slewing (the spec you read from the datasheet) and slew-rate limit due to output current limiting. The former causes distortion well below its full power bandwidth limit while the later will usually cause a much more abrupt clipping. Hence you don't need such a large safety margin.

Samuel

 

[PRR]

> Am I calculating the capacitance properly?

I've assumed 300 feet (100ft is domestic toys) and 30pFd/ft, and never ran into an unexpected loss.

300'*30pFd/'= 10,000pFd or 0.01uFd. So we'll take that (or 10 nanos if you prefer) as a number to argue with.

I won't argue Walt, nor the difference between 0.5 and 1.0. As you hint, you are not 2:1 away from heaven, more like 4:1.

> What effect....from the 100-200 ohm resistor usually placed in series with outputs?

You have two limits. Ideally you satisfy both.

The R-C product will give a smooth (undistorted) frequency response error. 220 and 10nFd is -3dB at 72KHz, ~-1dB at 36KHz, ~~-0.5dB at 18KHz, acceptable as a bad-day case.

The Current Limit against the C gives a Slew. Setting Walt's paper aside: most organic acoustic classical music sources DON'T have high slew rates. Power is falling above 500Hz. Phase random coincidences suggest you should plan for full power to 2KHz, but can let it slip above that.

Note too that disk cutters (and slow tape) will NOT handle high power at high rate. Cutter amp power soars, cutters smoke. Tape record EQ rises the highs while tape coating capacity actually falls.

In fact the upper corners of traditional analog recording systems both reflect the maximum slew of the music they were designed for, and limit the possible slew of newer music styles. (One objection to 78s is that they could not slew even the music of their time in the inner diameter, and LP/33 was designed to handle music well.)

So arguably you needn't goose above a few KHz unless you are using/abusing digital recording to carry sounds never needed before. (Some recent CDs, rich in clipped cymbals, do have sounds we never wanted to hear on disc/tape.)

> re-discovered why real pro equipment is big and heavy and runs hot?

One more thing. You should be able to wYe several feeds off one source. Normally we have DAs for that, but sometimes the situation is not normal. Additionally you would really like feed A to stay up when feed B gets shorted. So arguably the driver stage should be able to drive its series resistors to a dead short cleanly.

And for seriously long lines (miles), the series resistor should be under 100 ohms but greater than zero (a zero-source driver can have audible reflections on a very-long line). A value of 60 ohms was proposed because it gives only 1dB error if an old-style 600 ohm load happens, relative to calibration in hi-Z load.

So ~150mA peak current. And if you have that, no capacitive load will ever cause audio slewing before you run into voltage clipping. 2uFd(!) load will cause a 5KHz rolloff but no triangular slew.

> preamp in the living room and the computer is here in the workroom

5532 chip will drive that better/cleaner than your 33rpm records (nevermind your 78s). The only way it matters is if you are gonna throw a Cray at it to decorrelate the noise and scratches; then slew limiting of grit impulses might make the Cray work harder.

How the heck are you gonn drop the needle on a 78 and run 75 feet in time.... oh. Storage is so cheap you are gonna record a whole stack and cut it up later.

 

[JohnRoberts]

Paul: I applaud your comprehensive and conservative approach to this question, and not make light of your effort but if that were an airplane wing, it might have weight issues. :wink:

You have already received several good responses, I would just zero in on a few isolated points to amplify.

Resistive build out impedance will effectively decouple following loads. Cable capacitance is generally a stability issue before rate of change limitation, but decoupling mitigates against both.

Secondly, a phono source will not typically exhibit high edge rates. Even the ticks and pops will (should) be LPF by playback EQ which will result in modest final slew rates. RIAA EQ looks like LPF at about 2kHz.

I am comfortable with allocating a healthy bandwidth margin above and beyond immediate signal needs, but when making secondary budgetary analysis (like current to support slew rate) consider the rationale for that extra margin of headroom in the first place. IMO it's not that there's some ocassional outlier signal that fast, but instead the margin insures that behavior well below those limits will be well behaved and linear.

I would approach the cable capacitance as the C in a simple RC to form a LPF well above the audio bandpass, not a C which must be brute forced by drivers.

JR

Note: There are some advanced phono playback processing approaches that prefer to declick/de-pop, before applying playback equalization. Such a raw phono feed could exhibit more demanding edge rates. That's pretty much the point, to not smear or spread out in time, tick energy any longer than it takes to settle. For best results using that approach I'd be tempted to locate the turntable and pre closer to final processing as any LPF at that point is undesirable.

If this "special pre playback EQ" processing is done in the digital domain this is all academic as the A/D convertor will likely LPF the signal lower than any pole formed by cable C. Have fun... I have some old 78s I haven't captured yet.

 

[CJ]

get some rolls of cable and have at it with the dual trace.
Paul use to string Cat 5 for audio, if I remember back when.

 

[pstamler]

[quote author="PRR"]preamp in the living room and the computer is here in the workroom

5532 chip will drive that better/cleaner than your 33rpm records (nevermind your 78s). The only way it matters is if you are gonna throw a Cray at it to decorrelate the noise and scratches; then slew limiting of grit impulses might make the Cray work harder.

How the heck are you gonn drop the needle on a 78 and run 75 feet in time.... oh. Storage is so cheap you are gonna record a whole stack and cut it up later.[/quote]

It doesn't take a Cray; a basic PIII descratches just fine. I use ClickFix, a plug-in for Adobe Audition, plus other software from DC6.

And yes, I start the computer, walk in and start the 78. When I've done both sides I go back into the computer room, stop the recording, highlight the audio from Side A and export it, then highlight the audio from Side B and export that. The rest goes into the trash.

Will think hard about the rest of the post.

Peace,
Paul

 

[pstamler]

[quote author="JohnRoberts"]Paul: I applaud your comprehensive and conservative approach to this question, and not make light of your effort but if that were an airplane wing, it might have weight issues. :wink:

You have already received several good responses, I would just zero in on a few isolated points to amplify.

Resistive build out impedance will effectively decouple following loads. Cable capacitance is generally a stability issue before rate of change limitation, but decoupling mitigates against both.

Secondly, a phono source will not typically exhibit high edge rates. Even the ticks and pops will (should) be LPF by playback EQ which will result in modest final slew rates. RIAA EQ looks like LPF at about 2kHz.

I am comfortable with allocating a healthy bandwidth margin above and beyond immediate signal needs, but when making secondary budgetary analysis (like current to support slew rate) consider the rationale for that extra margin of headroom in the first place. IMO it's not that there's some ocassional outlier signal that fast, but instead the margin insures that behavior well below those limits will be well behaved and linear.

I would approach the cable capacitance as the C in a simple RC to form a LPF well above the audio bandpass, not a C which must be brute forced by drivers. [/quote]

Just to clarify: I started thinking about this in conjunction with phono preamp design, but went on to consider what's going on with a mic preamp. Good transformerless mics have bandwidth out to at least 50kHz (at least, the small-diaphragm ones do), and the Jensen input transformers I usually use have bandwidths out to about 90-120kHz. With close-miked instruments there are some pretty fierce transients. And a mic preamp is typically designed for +4dBu nominal output, as opposed to the approximately -8dBu (-10dBV) used for a phono output. So the challenging case is the mic preamp.

Looking back at the phono issue, for acoustical 78s I'm designing for a 10-20kHz rolloff rather than an RIAA 2.1kHz rolloff. I'll add rolloff later. Anyway, the calculations for the phono circuit came out reasonably manageable. But the mic preamp circuit looks tougher.

Peace,
Paul

 

[pstamler]

Yeah, probably.

I've been doing some reading, including THAT's data-sheet for the OutSmart chips. First thing, I was incorrect in my surmise that one would simply add the conductor-to-conductor capacitances and the conductor-to-shield capacitances. To get the total capacitance a balanced driver sees, you'd add the two conductor-to-shield capacitances in series with each other, then add that total to the conductor-to-conductor capacitance. So with the aforementioned Gotham GAC-2, you'd have 38pF/ft in series with 38pF/ft, or 19pF/ft, adding to 58 pF, for a total of 77pF/ft. Not quite as gruesome a result.

The Rane site suggests that real-world music rolls off at 6dB/octave from 5kHz, so that peak levels at 20kHz would be 12dB less than at 5kHz. Perhaps...but conservative design suggests that one should build something that can drive full level across the entire audio spectrum and then some. (John is probably right that it's a good thing I don't design airplanes.)

Anyway, the THAT chip looks like a dandy device, as does the LT1010 buffer chip. Then of course there are discrete transistors. We'll drive those cables yet!

Peace,
Paul

 

[JohnRoberts]

[quote author="pstamler"]Yeah, probably.



(John is probably right that it's a good thing I don't design airplanes.)



Peace,
Paul[/quote]

I am not opposed to conservative over engineering as long as the cost is not out of whack with the benefit. I have also seen engineers screw up when they assumed the other engineer was conservative so they didn't have to be :roll: . So these are all valid considerations IMO.

I even put discrete transistor drivers on my last phono preamp, but not because it really needed it. In design sometimes you just aim for way beyond "good enough" when the cost is pennies.

Rane is absolutely correct about distribution of energy in acoustic music, but close micing a crash cymbal can easily exceed that in top octave. So perhaps for preamps in a multitrack recording studio, design for higher peaks in top octave. OTOH there is no need to size the PS for every channel to support those edge rates at the same time.

JR

 

[CJ]

leave it coiled! you want the high end roll off.
air core, atually, would mean micro henries.
you need millli henries to do damage.

i do know that if you put a 90 degree in a cat 5 cable, you lose half your bandwidth.

 

[PRR]

> It doesn't take a Cray; a basic PIII descratches just fine.

I'm well aware of that. I was de-noising on 486-50MHz machines. It took a 10-hour run to denoise 11 minutes of camcorder audio. It is asTONishing how fast an AMD 2.2GHz will do the same work. (Much faster than my previous P4-1.6GHz.)

And having been at this too long, I have the habit of rigging the needle in one hand and the REC button in the other hand. I've sweated 37 minute suites on a 400MB drive. In those days, you didn't record walking-time and trash it later. I guess I'm behind the times.

> a phono source will not typically exhibit high edge rates. Even the ticks and pops will (should) be LPF by playback EQ which will result in modest final slew rates. RIAA EQ looks like LPF at about 2kHz.

The complete playback response is the falling preamp and the rising response of the velocity pickup.

If you have a displacement pickup (crystal and more exotic things), the base preamp is flat with a 6dB rise from 500 to 2KHz. Dirt and scratches sound just the same as on a velocity pickup. (Times the nastiness of cheap ceramic needles, but that's cost-cutting, not a fundamental difference.)

Anyway: ideally the stuff coming out of the preamp is "identical" to the stuff on the master tape.

There are no high edge rates because it is tough to generate a high edge rate in air, and it is annoying when you do. Most strings and soundboards and resonant air-pipes and even drumskins are seriously low-passed. To shift the low-pass far up the audio band you'd want smaller instruments, and then it becomes tough to couple player energy into air at sufficient volume for dancing.

> full level across the entire audio spectrum and then some.

Which is what? You are over 25, you don't hear 19KHz so well, but that doesn't mean it isn't out there. Probably not a LOT of power at supersonics, because most instruments and microphones have been developed to optimize the stuff we can hear, and along the way lose the stuff we don't hear. But if you put a 1/4" measurement mike on a crystallized cymbal edge (or breaking glass), you may be limited only by the capsule's top corner, 50KHz.

> With close-miked instruments there are some pretty fierce transients.

Ah. If you are not talking "music" but "close-ups for sparkle in the mix", then indeed you can have some zingy high-end energy.

I happened to have Pines Of Rome open. I selected the Big Finish. I applied a filter, flat 40-800Hz, rising 6dB/8ve 800Hz-20KHz. The peak levels hardly changed. For this pretty big orchestral climax, a passive (not feedback) slew-limited system only needs to pass Full Voltage up to 800Hz plus a margin, nowhere near any 15KC, 20KHz, 48KHz "audio band limit".

But that's orchestra in the audience. You get a cymbal fanning your diaphragm, it's different.



> youl'll need about 91.4mA of current.

Note that Jensen 990 audio amp has output current in excess of that. The nominal reason is to drive lo-Z loads such as multiple loads or step-up transformers. But surely to drive C also. And that may be part of the reason the 990 is still a gold standard for audio.


> If you drive a reel (or un-reel box) of cable, aren't we making an air core choke out of the whole mess?

If it is "cable", balanced or unbalanced: you have made a non-inductive resistor. The L going out cancels the L coming back.

Also: the ONE-way inductance of a box of CAT5 (or mike cable) is in the general order of a milliHenry. It is a "speaker choke". It will drop 8 ohms at about 2KHz. So 80 ohms would be 20KHz, 160 ohms about 40KHz, and in the typical case of hi-Z loading (and ignoring the ever-present capacitance) it would droop far above the audio band. Assuming all the L was one lump and all the C was another lump, it's still above the audio band, and it isn't really going to work as two lumps.

And 10X to 100X times higher for "cable" because the inductances cancel.

 

[clintrubber]

OT, but I see you guys are digitizing your LPs in realtime ? :roll:

http://www.cs.huji.ac.il/~springer/ :wink:

OK, back to program.

 

[Freq Band]

I don't know enough about this topic to contribute much (or if this applies at all), but I have one of these driver units and it seems interesting. Can drive up to 2500' of cable.
Unfortunately it does not give any S&N specs, nor distortion.
You can email me for manual w/schemo.

Driver/Receiver:
http://www.modsci.com/products/radioProducts/cld2500/index.asp

Inside:
http://i5.photobucket.com/albums/y177/Midiot/DSCN2125.jpg

I emailed the maker of this part, but they said. "...there are no datasheets available, we sell mostly to the military."
http://i5.photobucket.com/albums/y177/Midiot/DSCN2126.jpg

=FB=

 

[PRR]

> you guys are digitizing your LPs in realtime ?

The optical needle is an old pipe-dream; the scanned groove only somewhat newer.

My pencil says that the finest modulations on an LP are too small to be resolved by incoherent light, and may not be well resoved by an interferometer.

 

[PRR]

> Can drive up to 2500' of cable.

It appears to be a simple double-terminated balanced line. It will put 10V p-p across the 78 ohm termination. 5V peak in 78 ohms is 64mA. It probably terminates the source; ie it has 78 ohm output impedance. Then those hyper-obscure buffers must make 20V p-p across 156 ohms. Still 64mA. Not a tough spec.

What may not be clear: it takes a "composite stereo" signal. This is the "sum" (mono) signal up to 15KC, plus a double sideband supressed carrier "difference" signal centered at 38KC. Nominal 53KC bandwidth. However the nature of DSB means that a mild slope from 23KC to 53KC does little harm. They give a 1-turn trim for that.

It seems to be driving over 75,000pFd seeing 78/2= 39 ohms. That implies -3dB at 54KHz, which may be how they came to the 2,500 foot spec. However the heap of capacitance is mostly behind the nontrivial inductance of the long line. We have moved from lump parameters to distributed parameters. Performance will be better.

A triangle wave at 5V peak at 53KHz is only 1V/uS. To pass a sine at that level, we need much better than 1V/uS, but not a real challenge. Also, for the same reason we don't need full baseband power much over 1KHz, the composite difference signal is unlikely to be strong past say 40KHz. (Not to mention the obscenities done by modern FM "processors".)

> Unfortunately it does not give any S&N specs, nor distortion.

No; it would just exceed older FCC specs for FM radio. (Apparently today there are no specs except interference; you can transmit as bad a signal as you like.)

At those levels, S/N is surely not limited in the box. In practice you hope that induced noise along the half mile of line is less than the noise at the receiver. Unless you run past a welding plant or parallel to a high voltage line, it probably is. THD is surely far-far-far below the old 1% spec.

Unless you must run a hostile route, a pair of 5534 drivers and a 200 ohm termination driven to 10V p-p will work about the same.

 

[thermionic]

[quote author="PRR"]> you guys are digitizing your LPs in realtime ?

The optical needle is an old pipe-dream; the scanned groove only somewhat newer.

My pencil says that the finest modulations on an LP are too small to be resolved by incoherent light, and may not be well resoved by an interferometer.[/quote]

Have you seen this: http://www.elpj.com/ ?

I got to hear the thing a year or two ago - at a show in a listening room mind you, so it wasn't in a highly critical environment. I wanted to hate the thing, but I've got to admit it sounded pretty good... The most obvious attribute is the lack of surface noise. Separation and general presentation was pretty good.

Funnily enough, I know a chap (ex-Vox and Leak engineer) who was employed by Philips in the sixties to develop an optical LP player. My brain went into overload when I asked him about it as he threw a lot of physics at me, but - apparently - the prototype they had at Philips was totally analogue! I believe they had an optical transducer, much like a cartridge, but substituting an optical mechanism for the cantilever / coil arrangement.

Forgive the OT, Paul.

BTW - don't 78s require a unique RIAA curve? I believe KAB in NY sell a phono preamp that offers a 78 EQ option.


Justin

edit - taken from: http://www.elpj.com/about/how.html

The entire sound reproduction chain is analog

 

[clintrubber]

[quote author="PRR"]> you guys are digitizing your LPs in realtime ?

My pencil says that the finest modulations on an LP are too small to be resolved by incoherent light, and may not be well resoved by an interferometer.[/quote]
I could understand. IIRIC as the story admits already is was just a geeky experiment and for that it deserves applause (but also something that will have learnt him a few things though, that's nice).

Regards,

Peter

 

[PRR]

> don't 78s require a unique RIAA curve?

Early 78s had only mechanical "EQ" which evolved by ear.

Electrical 78s understood the need for EQ, but every company did it different at different times.

NAB proposed a compromise which most organizations leaned to, but Columbia was a hold-out, and some premium european recordings also went their own way.

Not a big deal with the state of home speakers through the 1930s and 1940s.

A semi-decent 78 preamp for the flatter hi-fi speakers of the 1950s will have two knobs each with a half-dozen settings.

I didn't realize the laser stylus had been perfected well enough to stand against needles. Bravo.

 

[pstamler]

[quote author="thermionic"]Forgive the OT, Paul. [/quote]

No problem; it's fun.

BTW - don't 78s require a unique RIAA curve? I believe KAB in NY sell a phono preamp that offers a 78 EQ option.

No -- they require a non-RIAA curve. Which, as PRR notes, is different for every record label, and for that matter is different between decades. What Columbia used in 1929 wasn't what Columbia was using in 1946.

I wrote an article a couple of years ago for audioXpress about the design of filter networks for various curves, RIAA and otherwise. Working my way toward the design of a full preamp; this discussion of driving long cables came, indirectly, from that design project.

Peace,
Paul