TLE2426 vs resistive divider

[Ben N]

I'm building DeadEndFX's  Redstone boost (Hey, Dino!) and it calls for a Tle2426. I don't have one, and I just placed a Mouser order (including the 1590bbs for this beast). Is there any reason I can't get by with the usual resistive divider?
Or would something like this on a little daughterboard be preferable:
(Taken from [ur]lhttps://tangentsoft.net/elec/vgrounds.html[/url].)

[ElectricDruid]

I don't see anything that looks particularly critical, so I'd be inclined to at least give it a try with the usual resistive divider.

The "discrete rail splitter" you posted looks like far too many parts. If I was going in that direction, I'd use an op-amp and get better performance with less messing about.

[amptramp]

If you want a buffered Vcc/2, an LM386 will do the job as one of its advertised features is self-centering quiescent output voltage and you can make use of the 50 db power supply rejection ratio.  The output impedance is very low as it can supply typically 700 mW and a minimum of 500 mW into the load, which is the Vcc/2 supply.  If you have more than one stage that is biased from Vcc/2, this prevents signal mixing.

[R.G.]

As A.T. says, an LM386 works fine, and self biases to half the power supply with no extra parts. You can also do what the TLE2426 does - use an opamp with a resistor string on its input to buffer the resistor string's voltage to its output. That's what the TLE is inside.

But I'd try it first with the resistor divider. I looked at the TLE or one of its progenitors years ago, and came to the conclusion that if I ever ran into a real need for it, I'd use it. Haven't run into that need yet. A real need for it would be a significant DC current into or out of the Vbias source, or a significant AC signal "current" into or out of the Vbias node.

For my designs, I concluded that either a significant DC or AC current in or out of the Vbias node is a design mistake. Not saying that about all designs, just the ones I've done. Both types of currents at the Vbias voltage were handled other ways. I have used the equivalent opamp buffer for Vbias for special quietness needs, discovering along the way that the buffered version also buffers supply noise and ground noise into the Vbias, so you have to do it carefully.

Separate resistor strings for multiple bias voltages introduce issues with the bias voltages not being exactly the same, which just means that you have to block DC between stages using the different bias voltages to avoid introducing unexpected oddities with DC levels.

[tubegeek]

...you can make use of the 50 db power supply rejection ratio.

As A.T. says, an LM386 works fine, and self biases to half the power supply with no extra parts.... I have used the equivalent opamp buffer for Vbias for special quietness needs, discovering along the way that the buffered version also buffers supply noise and ground noise into the Vbias, so you have to do it carefully.

Am I misunderstanding or does this mean that the LM386 is a superior choice as compared to an opamp fed by a resistive divider because of the PS noise rejection?

[R.G.]

Maybe. The devil will always be in the details, and especially when you start comparing things in db.

Many opamps have much more than 50 db rejection of power supply noise. We can usually discount the opamp as a source of power supply noise, even though you do have to be careful about what you call the power supply to make this simplification.

You also have to be careful about what you call "ground". Power supply noise has to be compared to something, and it's easy to compare it to a noisy ground point. In the LM386, you're stuck with the ground that the chip sees, so ground noise, which adds directly to power supply noise, is tacked on. I haven't looked at the LM386 internal circuitry for how they derive the "middle point" to bias the output to, but it can't be any better than "in the middle of the difference between the most positive and least positive".

With a resistor divider and opamp, you get to stick a BFC from the reference point to a ground of your choosing, which can be (if you're careful) a quiet reference ground to tie that point as tightly to a quiet ground point for AC. This will, of course, be layout dependent.

My best guess is that an LM386 will be far better than a resistor reference designed without much thought in cases where there is a lot of AC or DC Vbias current; about the same for a quick and dirty resistor reference and an opamp follower in modest current cases; and worse than a carefully designed, good-reference opamp followed divider, and about the same as a resistor divider and BFC for low current cases.

[tubegeek]

The "F" in BFC stands for "Filter," right, Tex?

Sorry I missed you on my trip to Austin. Hope to see you one day.

[Ben N]

I can't honestly say I have a clue what the vbias current draw of this thing is; it's not a digital current monster, but it it's not a Fuzzface, either. It's a booster with eight discrete devices (6 x jfet, 1 germ, 1 mosfet) doing some interesting things with eq and harmonics, in addition to gain, and not a very familiar topology to me at all. Maybe current draw is the reason the original design went with the TLE.

I am working with what is now a fully populated pcb, and a 3-hole socket (not the 8 pin version) for the TLE. Based on the diagrams at the web page I referenced above, I have an idea how to connect wire up an opamp + resistors [caps?] on a daughterboard and plug that in, but how would a 386 connect there?

PMing Dino, as I imagine he would have some circuit-specific insight.

[PRR]

> LM386 internal circuitry

I think the '386 has only 6dB PSRR until you put the cap on the Bypass pin.
https://protosupplies.com/wp-content/uploads/2018/04/LM386-Schematic.jpg

It is "half" because there is 15k one side and 15+15k on the other side. For DC *or* AC/buzz.

Gosh, they even drew you a picture! No-cap is 6dB PSRR. 10uFd cap on the "Bypass" pin may be good enough (50dB) for a 4-inch speaker. 47uFd will be better for a large speaker with good 100/120Hz response.



The advantage over a 15k+15k+47uFd is that the '386's output impedance is far lower than 7k||47u over most of the audio band.

[digi2t]

I don't see anything that looks particularly critical, so I'd be inclined to at least give it a try with the usual resistive divider.

The "discrete rail splitter" you posted looks like far too many parts. If I was going in that direction, I'd use an op-amp and get better performance with less messing about.

I have to echo Tom on this one. Keep two points in mind;

- When I trace something out (especially Spaceman stuff), I simply try to maintain all the original topology and elements. My goal isn't to reinvent it, simply to trace it out, in it's completely original form.
- Spaceman has built a reputation on three fronts; stellar design execution, interesting circuit concepts, and.... quirky component selection.

Would a good old fashioned pair of resistors divider work here? Sure, just as it has in hundreds of other designs. Insofar as this project is concerned, a traditional resistor pair will work fine. It's what I used on the breadboard while I was waiting for the 2426 to show. It worked just fine with a pair of 10K resistors. Would it look as cool? Nope. Would it throw off the amateur circuit sleuth? Possibly. The list of Spaceman circuits that I've traced out to date is getting fairly heavy, and just when I think I've seen everything, they throw me an interesting curve.