Special requirement output buffer

I need for a special experiment a very low distortion line level output buffer, powered from differential PSU for DC coupling (+/-15 or 24V). Something in the direction of the Jung buffer.

Unfortunately this is not the only requirement. It should have an output impedance under 1ohm.
The obvious thing that comes to my mind is to do something in the direction of a power amp.

Question is, how would you do that? What would be the best compromise to achieve this? (using power output devices, paralleling output devices, playing with negative feedback, etc)

This low out Z requirement also would exclude any 50ohm resistor added at the output of the buffer to isolate cable capacitance and RFI from the outside. Probably something like the parallel RL network used in the Hardy preamp would solve this (0ohm in audio band, 39ohm at high freq)

I know, too many questions. :grin:
Would be happy to hear you thoughts about this subject.

chrissugar

What kind of power does it have to deliver into what kind of load?

A mini-power amp circuit or an opamp with a discrete transistor output section could do the job, depending on what your requirements are.

It should swing let's say between +/- 15V, and the impedance of the next stage is normal, higher than a couple of kohms.

I need this circuit for some kind of "myth busters" project (something similar to my DC servo, versus non DC servo, versus capacitor coupling experiment), so that is why the less then 1ohm impedance is important.

Of course after I will be ready with my tests, I will share with all of you my results. :grin:

chrissugar

>1 ohm? good luck!

The only thing that comes to mind is a MOSFET output stage..

Solid-state power amps routinely achieve output impedances below 1 Ohm. A similar topology, but scaled down for the lower power requirement, should work.

[quote author="Svart"]>1 ohm? good luck!

The only thing that comes to mind is a MOSFET output stage..[/quote]

Yes, not an easy thing. I consulted the technical specifications for lots of proaudio and audiophile equipment and most of them have line buffers with out Z between 25 and 150ohm (I'm talking about solid state)

[quote author="NewYorkDave"]Solid-state power amps routinely achieve output impedances below 1 Ohm. A similar topology, but scaled down for the lower power requirement, should work.[/quote]

Exactly. It looks like the answer is in the power amp direction. But I would like to avoid to use a complete power amp as a line level buffer. I'm thinking about something like the Borbely super buffer, or an opamp plus some adapted Jung buffer.

chrissugar

Even the Borbely Tube/mosfet headphone amp, with closed loop has 15ohm output Z. :sad:
http://www.borbelyaudio.com/eb804421.asp

OK, getting closer. The Borbely ALL-FET LINE AMP has output Z around 2ohm but it cant work stable without the 47ohm resistor in series.
Page 4:
http://www.borbelyaudio.com/adobe/borb502.pdf

chrissugar

It should have an output impedance under 1 ohm.

Click to expand...

That's trivial with some negative feedback. I'd try a AD797 connected as follower. Distortion is about as low as it get's. What load do you need to drive? If it's well below 600 ohm you might need to include a discrete buffer or some other solution (such as a fast current feedback medium power IC) inside the loop.

Samuel

What range of capacitive loading does it need to be stable into?

It is no big deal to make a circuit with low output impedance. The difficulty arises when connecting to real world loads. Shielded cable is often capacitive so high edge rate signals may draw lots of current, plus phase shift due to capacitive load may destabilize feedback loops.

It is prudent design to incorporate some build out resistance to insure stability and form a simple LPF with any capacitive loading.

To your specific question there are sundry simple opamp + transistor buffers that will more than satisfy your sub 1 ohm target. You may need to remove added build out resistance from those designs, as only high current outputs (like for driving loudspeakers) routinely target such low source impedance.

If you share the object of your myth busting we may be able to give more specific advice.

JR

[quote author="bcarso"]What range of capacitive loading does it need to be stable into?[/quote]
Normal capacitive load would be some hundreds of pF but for any cases 1nF would be more than enough.

[quote author="JohnRoberts"]It is prudent design to incorporate some build out resistance to insure stability and form a simple LPF with any capacitive loading. [/quote]
Yes, this is a contardictional situation. If I eliminate the isolation resistor, I expose the circuit to instability, if I try to solve the stability problem I increase the out Z.

[quote author="JohnRoberts"]If you share the object of your myth busting we may be able to give more specific advice. [/quote]
What I want to test is related to the sound differences between interconect cables in audiophilia.
There is some research related to this subject and it looks like the differences people hear are not because of the cables, but because of the interaction between the output buffer, the cable electrical properties and construction, and the input receiver circuit.
Some people tried to corelate the measured and the audible effects, and the results of these researches show that if the output buffer impedance can be lowered to under 1ohm, the cable will not have any effect in the audio range. My own experience with this is very similar, the lower the output Z of the buffer, the smaller the audible differences.

My goal is to practically demonstrate that the right way is not to buy cables that cost hundreds or thousands of euros, but to use some overengineered output buffer that will eliminate the cable from the equation.

One guy who did some research in this domain is a respected person in pro audio and the audiophile world, and you can be sure he is not a lunatic. He is Bruno Putzeys the moderator at PSW, ex researcher at Philips, designer of the popular UCD modules, and also one of the guys involved in designing the top AD converter Grimm AD1.

Another thing that makes me think that the solution is in this direction is a Swiss audiophile preamp, (an extremely expensive one). The manufacturer claims that contrary to most of the other audiophile products, their preamp is totaly imune to cable, and they encourage people to use any type of cable of any length (even a hundred meter) and there is no influence in sound. This preamp has an output stage very similar to a power amp topology.

chrissugar

[quote author="chrissugar"]...it looks like the differences people hear are not because of the cables, but because of the interaction between the output buffer, the cable electrical properties and construction, and the input receiver circuit.[/quote]
YES!!!

THANK YOU!

I've been trying to tell people that for ages.

However, the public doesn't think in those terms. They things distilled to a "four legs good, two legs bad" level. They simply DO NOT want to think about impedances.

People want to think in terms of "carrots are good for me, french fries are bad for me" terms. They don't want to have grey areas, like sunflower oil...

I've gone the OTHER way when attempting to make this point, and switched in series resistances in line with outputs, in order to make the effect MORE noticeable with the same cable. I think it's also how Monster load their smoke-and-mirrors listening demos.

Several domestic CD players have comparatively large value build-out resistances in the outputs, presumably to keep the units more unconditionally stable, or perhaps to protect the unit from the indignities of having the outputs "Y-cabled" together... -We all know that it happens in consumer-land. -Removing -or at the very least REDUCING these build-outs to more sane numbers usually has decreased cable sensitivity, and improved the sound across most cable types.

Well, that's been my experience. -Sometimes (As Monster have learned) it's easier to show people BAD sound than good sound... and let their imagination do the rest.

Keith

Crissugar, your output buffer means overshoot of power amplification that is against audiophile rules.

Up to what frequency do you need the Zout to be <=1Ohm ?

What about a 990 in non-inverting unity gain with a Load Isolator and/or a Zobel ? That should satisfy the Zout and distortion characteristics, no ?

JDB.
[you don't mention output voltage targets. If the down-to-DC spec could be relaxed, I'm sure that a 990 + an appropriate step-down transformer could be found to match the rest of the targets. And yeah, I believe your theory may well be correct]

Ok, there is a topology where you can have stability and reasonably low output Z, at least as long as you open loop gain holds up.

Visualize a simple inverting opamp. Now add a modest build out resistance ( say a couple ohms for your purposes). Instead of connecting the negative feedback resistor from the - opamp input to opamp output, connect the feedback resistor from - input, to the very output after the build out. For stability add a small capacitor connected across the opamp conventionally (- input to opamp output).

With this topology you can make the driver stable with arbitrarily large capacitive loads, and it will use it's negative feedback and loop gain to delive low flat output, at least until the buildout R and cable C becomes troo significant. Then the opamp is no longer a simple inverter, it looks like a 2-pole LP with the output capacitance forming the second pole.

For the purposes of your test you need to dial in values so you can tolerate your worst case capacitve load while still flat up to say 2 octaves above 20 kHz.

You may still need to add a class A or A/B transistor buffer (inside the opamp block) to get an opamp happy driving a ton of capacitance without interacting. My recollection is that topology gave a peak before rolling off with modest values and a big honking load, so you need to make yours from sterner stuff.

Have fun but couldn't you just let your fingers do the walking by looking at cable capacitance specs, preamp source impedance specs, and just do the math?

I have found wrestling with audio fools poor use of my time. In my experience they don't want an arbitrarily clean, flat, audio path... they are on some mystical journey, hoping to discover some "magical" combination of gear and widgets to reveal some new yet undiscovered audio truth.

I wrote a column called Audio Mythology back in '80s and did my best to skewer several popular ones. I usually found an ounce of truth in most, just like you will discover an audible difference in extremis from cable capacitance and drive impedance. It's the unscientific listener that draws some association, with usually the more expensive gear the winner due to halo effect.

I don't like to presume motives about people's behaviors, but if one was trying to compete in a market where all the well designed gear delivers similar performance, and you were looking for ways to sound different (insert your own devious scenario here).

I have for decades argued that if something can be heard reliably, it can be measured. If it can be measured, it can be managed. So much claimed but unmeasured phenomenon are anecdotal claims either innocently misinformed or not so innocently. (IMO flame suit on)



JR

Another thing that makes me think that the solution is in this direction is a Swiss audiophile preamp.

Click to expand...

Some inspiration for that one: class_amp.html

What about a 990 in non-inverting unity gain with a Load Isolator and/or a Zobel? That should satisfy the Zout and distortion characteristics, no?

Click to expand...

I'd go for an inverting configuration as this will provide lower distortion at high frequencies due to absence of common-mode effects. The AD797 (or a modern current feedback opamp) is probably better with respect to distortion though, at least at high levels. I could think of a discrete solution that beats ICs for this application, but that's probably overkill.

Samuel

[quote author="JohnRoberts"]Visualize a simple inverting opamp. Now add a modest build out resistance ( say a couple ohms for your purposes). Instead of connecting the negative feedback resistor from the - opamp input to opamp output, connect the feedback resistor from - input, to the very output after the build out. For stability add a small capacitor connected across the opamp conventionally (- input to opamp output).
...

You may still need to add a class A or A/B transistor buffer (inside the opamp block) to get an opamp happy driving a ton of capacitance without interacting.
[/quote]
The circuit you are describing is exactly the one I was thinking about to try, but I'm a bit worried that the feedback connected direct to the output is exposed to RFI that can ruin the stability of the circuit.
I was also thinking about a high current class A transistor output buffer after the opamp, and the negative feedback network to not include the output buffer, only the opamp. It looks like it is a current trend, even in proaudio (T0NELUX and others).

[quote author="JohnRoberts"]Have fun but couldn't you just let your fingers do the walking by looking at cable capacitance specs, preamp source impedance specs, and just do the math?[/quote]
What I want is not something that will work for a specific situation, but for any possible situation. I know this is a brute force solution, but if I can prove that a very low out Z high current buffer can produce identical results with different cables, for me is a win-win situation.


[quote author="JohnRoberts"]I have found wrestling with audio fools poor use of my time. In my experience they don't want an arbitrarily clean, flat, audio path... they are on some mystical journey, hoping to discover some "magical" combination of gear and widgets to reveal some new yet undiscovered audio truth.[/quote]
You are right, but there are two king of audiophiles.
There are the ones who don't care about how close is the musical presentation to the real musical event, they are just looking for some euphonic pleasure, and they are searching for magical and irational solutions for their pleasure. They use words like natural, warm etc, but there is nothing in common with natural and transparent reproduction.

There is also another kind of audiophile (like Keith Johnson for example), who is looking for transparency and to stay away as much as possible in the way of the original acoustic event. For me this is the kind of audiophile that is the reasonable one.

Probably I should make clear my position.
I think all the cable business is crap and paying hundreds or thousands of $ for interconects is absurd, but I have to say that there is a good reason for this thing to happen.
I can detect significant differences between various cables, but I think the problem should not be solved at the cable level but at the driver (and receiver).
Just to quote Bruno Putzeys:
"people who claim that cables do not make a difference are plainly deluding themselves. On the other hand, those that say that cables should not make a difference, are dead right."

chrissugar

I was also thinking about a high current class A transistor output buffer after the opamp, and the negative feedback network to not include the output buffer, only the opamp. It looks like it is a current trend, even in proaudio (TONELUX and others).

Click to expand...

Well a "current trend" is hardly an engineering justification, right? :grin: Excluding the buffer from negative FB is for sure not something you want for this application. You'll never get really low distortion and output impedance will be much too high as well.

The circuit you are describing is exactly the one I was thinking about to try, but I'm a bit worried that the feedback connected direct to the output is exposed to RFI that can ruin the stability of the circuit.

Click to expand...

In fact this approach is not that different from using a discrete LR output isolator. You just synthesise the L using the opamp feedback.

Samuel

[quote author="chrissugar"]
...
The circuit you are describing is exactly the one I was thinking about to try, but I'm a bit worried that the feedback connected direct to the output is exposed to RFI that can ruin the stability of the circuit.
I was also thinking about a high current class A transistor output buffer after the opamp, and the negative feedback network to not include the output buffer, only the opamp. It looks like it is a current trend, even in proaudio (T0NELUX and others). [/quote]

Any RF or phase lag at very output is dealt with by the cap from opamp output to - input, so the circuit is stable. Power amps routinely use output inductors in series (after feedback loop) to isolate capacitive loads. I would just target flat response a couple of octaves above passband.

Putting a class A buffer "after" the feedback loop is just adding a source of distortion. While the religion of Class A preaches zero "crossover" distortion, you can't ignore the sin of a changing Vb-e drop with varying signal. This may be nice sounding distortion but far from signal purity. A class A buffer inside the loop will be quite linear. You could even "Wavebourne" the opamp into class A by putting a resistor from base-emitter of the class A follower, which will look like a constant current sink to the opamp, pushing it's output crossover to a higher voltage.

What I want is not something that will work for a specific situation, but for any possible situation. I know this is a brute force solution, but if I can prove that a very low out Z high current buffer can produce identical results with different cables, for me is a win-win situation.

Click to expand...

Not exactly. The reason expensive high capacitance cables have gained some traction is precisely because they sound different and some people apparently like that extra roll off. You will just ruin it for them.

I see this a little like trying to convince kids that the earth is round.. They can tell from looking out the window that you must be crazy.

You are right, but there are two kind of audiophiles.
chrissugar

Click to expand...

Only two? The niche has actually quieted down some since dentists and lawyers with too much disposable income are impressing their neighbors with home theaters instead of mega bux record players these days but there is still a general distrust of science among the audio faithful. To some extent this is fomented by the knee jerk reaction to audiophiles from engineer types who say they can't possibly hear what they say they do. I believe they may indeed be hearing conventional stuff like good old frequency response which is precisely what your experiment is looking to parse out.

I don't wish to discourage you but there is a whole industry who profits from the concept that cables sound different and promotes mythology over science in defense of that. They have invested huge sums in advertisements to that end. Consumers who have purchased expensive cables will not be receptive to learning they spent their money foolishly. An informed consumer would never support such products.

Good luck. You are absolutely correct but IMO this falls more into the category of a fool with too much money being separated from some of that excess, and a fringe market/activity that has little to do with mainstream science or technology.

JR