Balanced Line Driver and Receiver Schematics

[JohnRoberts]

or add gain control upstream...?

JR

 

[mjrippe]

You are correct. I need to reverse engineer the output stage and see if it can be done there. It looks like an opamp with transistors on the output so it should be possible. Sometimes the attenuator method is nice for driving the preamp hard and pulling levels back at the end.

 

[Bo Deadly]

Since this thread has a lot of folks smarter than me answering questions, here is one regarding THAT balanced out drivers. I have a circuit with a low impedance single ended output, perfect for using the 1646, however there is no output level control. Is it a horrible idea to use, for example, a balanced 600 ohm attenuator after the 1646? If that is a bad idea, how else might one incorporate a trim control (could even be 0-6dB) without adding another opamp?

I have trouble imagining a scenario were you would want to do that. Ideally you should just put a pot in front of your upstream amp. If you really only want 0-6dB (kinda limited) then just run the lower leg of your pot through a resistor of equal value. If it's inverting there's a standard circuit for level adjust there too. Post a schematic.

 

[abbey road d enfer]

Sometimes the attenuator method is nice for driving the preamp hard and pulling levels back at the end.

Unlike a tube amp or an old school SS discrete circuit, the 1646 stays "desperately" clean till it hits the rails, where it just clips. No euphonic saturation.

 

[abbey road d enfer]

While DC from a microphone would be of interest only to a meteorologist, extended low-frequency response is important to music. A typical signal chain for recording and reproduction may contain dozens of high-pass filters, mostly in the form of coupling capacitors - but also those present in the microphone and loudspeaker. Just as in low-pass filters, there is a phase shift in a high-pass filter that's related to its "order." Since most coupling capacitors form single-pole filters, their phase lead approaches 90° below their -3 dB "cutoff" frequency. So the phase response of the signal chain becomes that of a very high order high-pass filter.

That is true, but in a properly designed system, in each link of the chain, the cut-off should be determined by a restricted number of well-defined 1st-order filters, which should put some order in the mess.

And this phase shift affects signal frequencies at least a decade higher than the cutoff frequency of each stage.

No doubt it affects frequencies that are below cut-off, but I would think teh effect on frequencies above to be limited.

Unfortunately, there is no equivalent of a Bessel high-pass filter to bring linear phase response to all this.

Actually there are Bilchikov filters that are the equivalent of Bessel for high-pass. But I agree that in order to implement them, one has to have complete control of the whole signal chain, which is close to impossible.

The only way to undo most of this true phase distortion (or deviation from linear phase) is to move the -3 dB "corner" or "cutoff" frequency down ... way down!

I agree 100%, particularly because it also coincides with the need to reduce distortion due to capacitors (particularly lytics).

Marshall Leach of Georgia Tech wrote a paper about this back in the 1980s. Therefore, sizing coupling capacitors for -3 dB at 0.5 Hz is not unreasonable! It's also why most Jensen transformers have low-frequency response down to well under 1 Hz. Of course, this phase distortion is cumulative - the longer the signal chain, the worse it becomes. Because kick-drums get much of their character from frequencies affected by this time domain distortion, long signal chains often reproduce kick-drums that sound nothing like the real-thing.

I won't discuss the subject of phase audibility, because it's still debated, and there are equally knowledgable and respected opponents and proponents.
I'll speak only for myself. All the tests I have done always demonstrated that whatever was audible was the result of non-linearities reacting to changes in the crest factor.
The cleaner the chain, the less audible phase distortion was audible.
Some attribute the audibility of a high-pass filter to the resulting phase distortion, but I'm convinced what I hear is the difference in frequency content. As mentioned in an other thread "The level of subsonics in a program affects the perception of the rest of the spectrum." Post #15 in Switch for capacitor swap for tonal options

It's also why I've always preferred the sound of a woofer in a sealed box ("acoustic suspension") to one in a vented box.

Couldn't it be attributed to the "barrel effect" (hump at resonance)? Most BR speakers are tuned with a slightly resonant alignment, because it "has more bass". Closed speakers are usually tined with a lower Q, because a larger Q implies a quite large box, which is not a good selling point.

I know that the audition mechanism still holds a number of secrets (How can we detect a binaural difference of a few dozen microseconds when the whole aural perception takes milliseconds to happen?), but so far, I have never been able to produce evidence of phase (or group delay) distortion audibility, except with LF digital "linear-phase" filters.
To me frequency response is so dominant (and physiologically ascertained) I have reservations.

 

[JohnRoberts]

While DC from a microphone would be of interest only to a meteorologist, extended low-frequency response is important to music. A typical signal chain for recording and reproduction may contain dozens of high-pass filters, mostly in the form of coupling capacitors - but also those present in the microphone and loudspeaker. Just as in low-pass filters, there is a phase shift in a high-pass filter that's related to its "order." Since most coupling capacitors form single-pole filters, their phase lead approaches 90° below their -3 dB "cutoff" frequency. So the phase response of the signal chain becomes that of a very high order high-pass filter. And this phase shift affects signal frequencies at least a decade higher than the cutoff frequency of each stage. Unfortunately, there is no equivalent of a Bessel high-pass filter to bring linear phase response to all this. The only way to undo most of this true phase distortion (or deviation from linear phase) is to move the -3 dB "corner" or "cutoff" frequency down ... way down!

I expect we are all in agreement that DC blocking poles need to be set "way down".

One common DC blocking pole encountered in typical audio paths are the Phantom voltage blocking caps. These are routinely tuned for low single digit cut offs. The luxury of making these another 10x larger is not very practical for mass market designs, and oversizing these particular caps can introduce other problems. Another problematic pole is the capacitor in series with mic preamp gain pot.

My personal preference is to design one dominant real pole using a high quality film capacitor and force lower quality poles "way down".

Marshall Leach of Georgia Tech wrote a paper about this back in the 1980s. Therefore, sizing coupling capacitors for -3 dB at 0.5 Hz is not unreasonable! It's also why most Jensen transformers have low-frequency response down to well under 1 Hz. Of course, this phase distortion is cumulative - the longer the signal chain, the worse it becomes. Because kick-drums get much of their character from frequencies affected by this time domain distortion, long signal chains often reproduce kick-drums that sound nothing like the real-thing.

The weakest link in reproducing natural drum sound is not the line level audio path response, that is arguably the easy part. Hint: LF loudspeaker response is increasingly difficult the lower we go. The energy content from bass/kick drums is lower than vocals but not as low as we might assume. For today's TMI about drums timpani appear to make notes lower than they actually do. The concave sealed back chamber forces a drumhead resonance series containing the first two overtones. The human brain connects the dots and fills in the missing (phantom) fundamental.

It's also why I've always preferred the sound of a woofer in a sealed box ("acoustic suspension") to one in a vented box. The former is a 2nd-order high-pass filter while the latter is a 4th-order. The higher cutoff slope directly translates to increased time-domain distortion. This time-domain distortion at low frequencies is, for me at least, reason enough to use DC coupling when feasible and "over-sized" coupling capacitors (and Jensen or other "over-designed" transformers) in signal paths. And, obviously, the shortest possible signal paths will generally sound better in this regard.

You are in good company. Back in the 70s I did some consulting work for Rudy Bozak (RIP). I asked him why he didn't consider adding a port to his classic and physically large "concert grand" loudspeaker. He said that classic music from a ported cabinet didn't sound natural to him. Since I was only there to consult about electronics I dropped the subject.

I don't want to start on a rant about negative feedback itself because the subject gets really complex really quickly. But I think that extreme open-loop gains (with corresponding extreme feedback factors) to drive THD numbers down is generally a bad idea.

I am inclined to say show me the money (data). NF has been widely attacked for as long as I can remember but I have never seen the proverbial smoking gun. That said I do not endorse pursuing vanishingly low distortion that is inaudible. I question the uber-op amps that can't be tested using conventional bench equipment without tricked up high noise gain. Surely that affects other op amp behaviors.

Better IMHO to have an amplifier with high open loop linearity and keep the feedback factor reasonable - as Deane Jensen did when he designed the 990, which also uses inductors in the input stage emitters to stabilize HF response without paying the usual penalty in slew-rate or equivalent input noise.

The inductor degeneration in his LTP is brilliant and I bet he made a lot of IC design engineers jealous. You can synthesize inductors onto an IC with caps and gain (gyrators), but impractical for use in a low noise IC input stage.

The 990 design was also in keeping with the "spectral contamination" paper that Deane wrote with Gary Sokolich just before Deane died. Less spectral contamination happens in signal chains with higher linearity, lower feedback factors, and perhaps most important, bandwidth limiting. IMHO, it's very misguided to think that bandwidths over 50 kHz have any benefit. Well, I've already gotten deeper into this than I intended!

Back when, a lot of people dismissed HF linearity as unimportant because the out of band HF distortion products didn't measure very bad in typical THD+N measurements. By the late 70s I rolled my own two-tone IMD rig (I modified a SMPTE analyzer to use 19kHz:20kHz). I found it extremely revealing for phono preamp design because the RIAA EQ LPF attenuated harmonic distortion components delivering better THD+N measurements than reality. The HF two-tone distortion product was actually boosted by the RIAA EQ, just like real world IMD from HF music content.

As usual Deane (RIP) was at the cutting edge of high performance audio. You are lucky to have worked with him and known him so well.

JR

 

[Newmarket]

One common DC blocking pole encountered in typical audio paths are the Phantom voltage blocking caps. These are routinely tuned for low single digit cut offs. The luxury of making these another 10x larger is not very practical for mass market designs, and oversizing these particular caps can introduce other problems.

Could you expand on this. TMI welcome

 

[JohnRoberts]

The 48 Volt Phantom Menace Returns (1,480k) phantom menace paper
By Rosalfonso Bortoni and Wayne Kirkwood, THAT Corporation
AES Convention Paper 7909, 127th Convention, October, 2009

In 2001, Hebert and Thomas presented a paper at the 110th AES Convention which described the “phantom menace” phenomenon wherein microphone phantom power faults can damage audio input circuitry. This paper offers new information about the phantom menace fault mechanisms, analyzes common protection circuits, and introduces a new protection scheme that is more robust.

===

The 48 Volt Phantom Menace (153k) link to paper
By Gary K. Hebert and Frank Thomas, THAT Corporation
AES Convention Paper 5335, 110th Convention, May, 2001

The authors encountered anecdotal evidence suggesting that field failures of existing line driver and microphone preamplifier integrated circuits (ICs) were correlated with accidental connections between line outputs and microphone inputs with phantom power applied. Analysis showed that the most probable mechanism was large currents flowing as a result of rapid discharge of the high-valued ac-coupling capacitors. Commonly used protection schemes are measured, analyzed, and shown to be lacking. More robust schemes that address these shortcomings are presented. It is concluded that the small additional cost of these more robust protection schemes is likely outweighed by the reduction in field failures and their associated repair cost.

===
lots of good reference material on That corp website.

JR

 

[FIX]

Adding an attenuator to the output of a 1646 pretty much negates it's balancing feature. The best way is to put the gain before the amp driving the 1646.

 

[mjrippe]

Unlike a tube amp or an old school SS discrete circuit, the 1646 stays "desperately" clean till it hits the rails, where it just clips. No euphonic saturation.

I do not mean overdriving the 1646, I was referring to driving the input stages hard (which could be tube, transistor, DOA, etc.) and then reducing the output. In any case, it is clear that the best practice is to have the level control before the balanced driver IC.

 

[mjrippe]

Adding an attenuator to the output of a 1646 pretty much negates it's balancing feature. The best way is to put the gain before the amp driving the 1646.

I was referring specifically to a balanced 600 ohm attenuator, which should not. Someone mentioned they are expensive, but it doesn't have to be a Daven. Member AVDO sells modern balanced 600 ohm attenuators for under 10USD. More than a pot would cost, but it does more too!

 

[FIX]

Yes we've been through that rant. Strangely a surprising number of the higher level folks here have expressed that high levels of feedback are bad in one way or another. Now you and FIX too?

For the record: Hogwash I say! There has been no reasonable explanation for high levels of feedback being bad other than proclamations about how it "sounds". Null testing is hard to argue with.

If you inquire within the professional audio engineers and producers, they will tell you one thing about my design of the Tonelux op-amp, is that it sounds more open at any gain. With the way I biased it, I increased it until it improved the distortion and when that leveled off, I stopped, then figured out a way to make that fixed. So the actual bias is what I would call a Class AAB design. The open loop gain is only 75dB, and like I said, the first pole is at 10KHz at open loop gain. The 990 is somewhere around 250-2KHz, and even with the low pass chokes on the input stage, when you run it at unity, is sounds a little pinched. Go to 6 or 12, and it sounds a lot better. I never understood why op-amp designers designed around unity gain stability, which effected the rest of the amp's characteristics so much. For me, I went for a design that was inherently stable without that kind of internal compensation. To do that, the gain had to be reasonable. Otherwise, it's like pressing the gas and the brakes at the same time. The TX op-amp still had distortion well below .01% worst case, and was usually around .005%.

 

[abbey road d enfer]

it sounds more open at any gain.
at unity, is sounds a little pinched.

Define "open" and "pinched". Someone has to invent an openness meter and a pinch meter.

I never understood why op-amp designers designed around unity gain stability, which effected the rest of the amp's characteristics so much.

Several have undertaken the issue, and came out with current feedback opamps.
Some are quite remarkable, e.g. LT1795 or ADEL2020. They are not cheap though. They are not plagued with the G.BW limitation. For some reason audio designers have neglected them, so the only available are not really suitable for audio; they often lack voltage headroom (due to low rails) and are optimized for high output current, which is not always necessary.

 

[JohnRoberts]

IC design for unity gain stability was done to support inexperienced design engineers (the larger market). Many early op amps required external compensation caps. The first unity gain stable IC op amp, the ua741 was an industry game changer, despite not being a very remarkable performer for modern audio metrics.

Multiple later op amps were available in unity gain and decompensated versions. The iconic NE5534 will remind you of that if you attempt less than 10 dB of closed loop gain, while the more popular(?) NE5532 is unity gain stable.

JR

 

[swpaskett]

Many years ago I was doing sound reinforcement in churches where they insisted I drive the installed system because "speaker boxes are ugly," I learned the transformer was my friend. If you don't know what you are driving, a good bridging transformer and a handful of pads are hard to beat -- it almost always works -- even when the output driver is poorly designed.
A question: Has anyone ever met a purist who has actually worked in the industry? I haven't...

 

[abbey road d enfer]

A question: Has anyone ever met a purist who has actually worked in the industry? I haven't...

A question: What is a "purist" in this context?

 

[abbey road d enfer]

There has been no reasonable explanation for high levels of feedback being bad other than proclamations about how it "sounds". Null testing is hard to argue with.

The subject has been discussed times and times again. It has long been known that NFB substitutes an "infinite" series of harmonics to the "simple" open-loop distortion that is supposedly constituted of 2nd and 3rd. Particularly Peter J. Baxendall, in 1978, in his 5th chapter, where he invites the reader to analyse the issue under the light of shape instead of harmonic content.
The subject was reopened at about the same time, by a bunch of "Young-Turks" who wanted to show their elders what they were made of, and came with the rotten concept of TIM, which is based on the paradox of Zenon of Elea. The correction signal provided by NFB can never arrive when it's needed, just like the arrow cannot reach the turtle (that's where science meets philosophy). Which led to assaults of imagination for getting rid of global FB.
Recently, a distinguished member of the LTspice group tried to demonstrate that NFB resulted in a vastly increased harmonic spectrum (which is mathematically true) but did not succed in demonstrating that is applied to real circuits, simply because no active element acts like a pure single-harmonic generator. Exponentials are transcendent.

 

[Matt Syson]

and oversizing these particular caps can introduce other problems.
The question posed earlier:
Well a couple of issues I can think of are when an input gets shorted to ground accidentally, faulty cables or patching for example a 'massive' amount of discharge current will get shoved into the first transistors of the input stage, and secondly, air conditioning 'rumble' can be present through the audio chain until it reaches an effective filter. This can go 'undetected' for a fair way in a mixing desk for example and switches may then appear to 'click' as they cut a significant signal at a couple of Hz.
Thus the 'high and low pass filtering of the input, as electrically close as possible to the input connector may be sensible.
Matt S

 

[abbey road d enfer]

I can't speak for John, but large capacitors are efficient antennae, which could create problems of HF rectification and unstability due to unwelcome parasitic capacitance.

 

[Newmarket]

The 48 Volt Phantom Menace Returns (1,480k) phantom menace paper
By Rosalfonso Bortoni and Wayne Kirkwood, THAT Corporation
AES Convention Paper 7909, 127th Convention, October, 2009

In 2001, Hebert and Thomas presented a paper at the 110th AES Convention which described the “phantom menace” phenomenon wherein microphone phantom power faults can damage audio input circuitry. This paper offers new information about the phantom menace fault mechanisms, analyzes common protection circuits, and introduces a new protection scheme that is more robust.

===

The 48 Volt Phantom Menace (153k) link to paper
By Gary K. Hebert and Frank Thomas, THAT Corporation
AES Convention Paper 5335, 110th Convention, May, 2001

The authors encountered anecdotal evidence suggesting that field failures of existing line driver and microphone preamplifier integrated circuits (ICs) were correlated with accidental connections between line outputs and microphone inputs with phantom power applied. Analysis showed that the most probable mechanism was large currents flowing as a result of rapid discharge of the high-valued ac-coupling capacitors. Commonly used protection schemes are measured, analyzed, and shown to be lacking. More robust schemes that address these shortcomings are presented. It is concluded that the small additional cost of these more robust protection schemes is likely outweighed by the reduction in field failures and their associated repair cost.

===
lots of good reference material on That corp website.

JR

Thanks. I read the 2001 paper previously but not the 2009.
So - yes - 'energy storage'/phantom issues. I was wondering about more 'audio' aspects due to the nature/size of larger electrolytics I might be missing if that makes sense.
I see others have chipped in wrt rf rectification considerations.