Guitar buffer / active DI / rail splitter questions...

I hope this isn't too much to put in one post, but I need a little schooling.  I've been looking through various schematics lately, and I have noticed some similarities, and yet disparities in some designs, and I'm wondering what the purpose is for them.

Firstly, looking at JLM's opamp DI schematic, found at http://www.jlmaudio.com/JLM%20DI%20kit%2048v%20Simple.pdf , it is very similar to the AMZ super buffer, http://www.muzique.com/lab/superbuff.htm .  The super buffer uses four opamps instead of Joe's two, and Joe's provides a balanced, but not differential output (meant to feed a mic input), but otherwise they are quite similar.

What is the difference in performance between using the 100uF bipolar output cap as Joe uses, and the 10uF electro in parallel with the 0.1uF film on the AMZ's output?  Would Joe's have better phase characteristics because of the larger cap?  What are other considerations for determining the value of this cap?

And when I compare Joe's DI design to one by Rod Elliott on the ESP site, http://sound.westhost.com/project35.htm (the last drawing on the page), I have another question.  Rod's design, like Joe's, is phantom powered and uses the TL072, except it uses an inverting stage on the second half of the TL072 to provide a true differential output.  His output bipolars are 22uF, but he adds 'protection' for the opamp in the form of a pair of 1N4148 diodes on the opamp outputs.  He states that these protect the opamps as the output caps (22uF) charge at power up.  Joe has no such protection on his design even though he's using larger caps (100uF), and I've not heard anyone complaining of problems with his design.  This is probably not a big deal, but I'd like to understand better the philosophy behind the use of diodes in this way.

Also, both the JLM and the ESP DI's run on phantom power, so they have the 6.8k resistors at pins 2 and 3 to derive this voltage.  What, if anything, do these resistors do to the output impedance of the circuits?  JLM's have 51R resistors at the outputs, and the ESP 100R, which determine the output impedance of the circits.  Does having 13k6 between these (the series resistance of the two 6k8's) alter this in any way?

Lastly, all of these designs use a resistor divider to create a virtual ground, as they all run opamps off of a single supply rail.  I have recently discovered Texas Instruments' TLE2426 rail splitter, and understand it to be, in theory at least, superior to a standard resistor divider in that it eliminates the need for resistor matching and also provides a lower impedance for the virtual ground.  It is only limited to around 40mA of current max, but in these designs (running off of phantom power) that's not really an issue.  The 8-pin versions of the device allows connection of a capacitor to further lower noise away from the output, but I don't understand how to calculate the value for this capacitor.

So... has anyone had experience with the TLE2426, and if so, can you tell me how to determine the appropriate capacitor for the noise reduction? 

Like I said, that's a lot for one post, but these are things that I don't fully understand, and I'd really like to get a better grasp of them.  With the exception of using TI's rail splitter, I've built all of these circuits, and am satisfied with the performance of them.  In use there is nothing that just jumps out at me as being vastly superior, one to another, but then again I'm using them for different things.  I'm just wondering what the practical impact is of these various design differences.  It would seem to me that these concepts can be (and are) applicable to a variety of other types of circuits, not just DI's, and I really would like to gain a better understanding, particularly of the function of the output capactors' values, and I'm guessing there are others on this group who would benefit from this discussion as well.

Thank you in advance for helping to bring clarity on some of this.



Much peace,

JC

The first, non-tech thought that came to mind here was something in the vein of 'many roads lead to Rome' (I figure there'll be a Stateside-equivalent). Unsurprisingly as in: different people come up with variations on a solution to the same problem and the differences are noticable but often not that significant. And most of them get the job done, but crap out at different stages of the difficult life they may or may not encounter.

Why done this way & not that way ? Reasons may be differing degrees of insight in the matter, but also often historical, having experience with a certain approach etc. I also think to recall the posting from Dan K about the design of a certain DOA; IIRC the outcome was the result of (1) going on on a certain chosen direction + (2) loads of either coffee or beer. As in: had he started on a different day the result would/could have been different. That's at least how I remembered it, and at least it illustrates the thoughts behind more than a few circuits, so just because it happened that way (you could add: songs/books/relations/etc etc. to that, but OK).

Might be pretty unsatisfying in the context of this thread, but so be it.

FWIW, I recall the schematic of a certain bass-amp (from Peavey), with all kind of stuff on board to address certain problem-scenarios. My next amp was from SWR, imho it sounded way better but I was pretty shocked when I saw the schematic; many of the 'prudent stuff' as found in the Peavey was simply not there. Electrically these SWR-amps are still going strong but I felt the people that designed the Peavey-stuff were at least as knowledgeable. Maybe a bit too much even, but I'm no Q&R expert.


I guess it's for the benefit of this thread if additional reactions that more directly target your issues show up, but nevertheless let these more vague ramblings be said...  8)

Regards,

  Peter
 

Hello,
Found that:
http://tangentsoft.net/elec/vgrounds.html
There's a TLE2426 circuit with noise reduction cap at about 1/3 the page. Hope that can help...

Laurent.

Thanks for the comments.

Yes, Laurent, I found that article too, but it left many questions.  Thank you for looking it up though.  I'll talk with TI to see if I can get some answers regarding the 2426.

JC

The idea of paralleling opamp sections has been around for many years and is a common way to provide more output drive.

The AMZ Super Buffer is designed for use with guitars and guitar pedals, and that accounts for the differences in component values, including the output capacitors.  You're not driving 600 ohm inputs with it so the 10uF is plenty on the output.

The simple resistor divider is more than adequate when the current draw is low, and an effective solution especially in a pedal where space is limited.

regards, Jack

> something in the vein of 'many roads lead to Rome' (I figure there'll be a Stateside-equivalent).

"All roads lead to Rome." Except around here, most roads will get you to NYC.

I don't see a significant difference between the JLM and AMZ plans. A lot of "pick a number from 10K to 10Meg... what do I have too-many of in my parts-bin?"

A single TL071 can drive 2K well, 600R straining. So JLM's plan will pull 1K to 275, AMZ's plan will pull 500R to 100. The vast majority of loads are over 2K resistive. Cable capacitance sucks. AMZ's plan will whack a half-mile of cable to 20KHz, JLM's "only" a quarter mile.

A TLE2426 is -not- needed because all we need is sub-picoAmp leakage into TL07x inputs, plus the leakage of the filter cap. Since the power supply "should" be clean (9V batt or Phantom), it should not need huge filtering. JLM's values roll-off above 31Hz, killing hiss; if hum is a problem you might increase the 0.1u (or use cleaner power). Note that when you have a pickup plugged-in (un-plugged inputs are not interesting), any trash here is further clobbered by 2.2Meg:5K. Cleanliness is a good habit but not an issue here. AMZ used 10uFd which rolls-off at 0.3Hz, and would clobber hum; but again we have 2.2Meg to the jack so any interesting source ought to be unaffected by bias-divider trash. I think here we have a designer who buys 10uFd caps by the bucket-load. It is not the lowest-value which will "work good", but the price is very-very low. OTOH, JLM may be over-stocked on 0.1u caps.

The output cap: if you actually drove 100 ohm resistive load, 10uFd will shave your bass. However if load is >2K the bass is fine. AMZ's goal seems to be long-line capacitance, though many-load is mentioned. I think it would be hard to find a modern real-world case where 10uFd hurts bass; if you wanna over-kill, pay the dime extra for 100uFd.

Electrolytics have faults. Every recording you own has been through a pile of electros, but it has become common practice to throw a film-cap across electros to carry the midrange around the rusty aluminum. It is unclear if JML disagrees, or knew his 100u caps sound fine, or if his solder-hand throws film-caps on by habit without telling his drawing-hand. Get some good beer, spend a night with a handful of various caps, and listen for yourself.

JLM used TL072 because that is about all the current you can pull from Phantom. AMZ is not showing a Phantom version, and I suspect he has a bucketload of the '074 chips. TL074 seems useful, but in anything more complicated than 4-buffer, the four sets of external components get crowded; I started a design with '074 and re-drew it with twice as many '072 so the other-junk fit better. All us old guys have a tube or two of quad-amps we never find enough use for. This 4-buffer is a perfect way to use them up.

> 6.8k resistors ..  What, if anything, do these resistors do to the output impedance of the circuits?

"Nothing".

How fussy are you? The exact Zout of a TL071 is uncertain but likely <10R. We find another 50 or 100 added to each one, because a very-low Zout can cause troubles. In fact both plans give 25-30R Zout before we add the 6.8K.

Assume the raw amp is 30.00000 ohms output. And the raw amp sees only one 6K8 resistor. Gain is 6K8/(30+6K8) or 0.9956, Zout is 30||6K8 or 29.868R. The uncertainty in raw amp Zout is far greater than 30.000:29.868. The necessary accuracy of Zout is very low-- we generally don't care zero or 200 ohms. And in this case, you are going to pluck a fiddle and adjust the knob on the recorder for happy peaks... gain accuracy is not an issue like it might be for telco, network, or copy work.