'Discrete' Voltage reference...

[liquids]

While this is probably an amateur question, it's been on my mind, so I figured I'd ask even if it discloses my (lack of)  knowledge about basic electronics.   

I like to use an op amp stage for a reliable voltage divider / Vref supply rather than just two resistors and a capacitor, when possible.  But there are those times where I just don't want to add another op amp, if I don't have to...

Got to thinking about the mini-booster and discrete buffers in general.  An op amp, you feed it a voltage and hook it up like a buffer...and it's a voltage source.   

So, does a JFET or BJT (etc) with it's drain/collector hooked up to V+, and it's gate/base with a voltage divider on it, and no other signal....does the emitter/source function as a good Vref?

And if not, is there another way to use Jfets or BJTs to act as a adequate voltage source/voltage divider, similar to an op amp stage?

[frequencycentral]

I'm actually surprised not to see 78L05's used as voltage references. Close enough I would think.

[trjones1]

I wonder about the 78l05 too.  My onespot power supply puts out 9.5V, so the 5 V output would be almost exactly half.

[ianmgull]

I wonder about the 78l05 too.  My onespot power supply puts out 9.5V, so the 5 V output would be almost exactly half.

Although I always assumed that the extra .5-.6v was to compensate for any voltage drop that occurred from a series polarity protection diode.

[R.G.]

1. Having a Vref track 1/2 of the power supply as it falls due to battery life is a good idea in general; only apps which need a fixed voltage would want Vref to stay at 4.5 or 5V when the battery declines to 7V.
2. Most people use ac adapters so 1. may not matter.
3. If an opamp isn't available, there is a device which integrates a 1/2 V+ divider and an opamp into a single TO-92 package. TLE2425 as I remember, but look up the exact number.
4. A bipolar emitter follower is a medium-good Vref. It suffers from drift and loading issues compared to an opamp, but isn't that bad overall. You have to remember that the base is 0.4 to 0.7V higher than the emitter, depending on current levels, temperature, hfe change with current and hfe and the difference between its impedance pulling up as opposed to down. But they've been used a lot.
5. A resistor/resistor reference with a BFC (Big Freaking Capacitor) gets relatively better as the cap gets bigger and the resistors get smaller.
6. The LM386 8-pin DIP all by itself, no resistors, caps, etc.  is a good, low impedance Vref provider.
7. It is easy to use up more parts and PCB space doing something like a transistor buffer than an opamp would use. To do a comparison, it helps to count an 8-pin opamp as two TO-92's and then  count the Rs and Cs. It creeps up quickly.

[earthtonesaudio]

6. The LM386 8-pin DIP all by itself, no resistors, caps, etc.  is a good, low impedance Vref provider.

It would be difficult to find a simpler, lower-parts-count, and physically smaller solution than this one here.

[head_spaz]

TLE2426 is the dang deal doodad you're looking for.

[PRR]

There are tradeoffs with "solid" and power consumption.

Two 1-Ohm resistors would be pretty solid, but eat 4.5 AMPERES or 41 Watts. Cook your foot.

Two 100K resistors is low-low power, but not solid enough to reference several things, or even one beefy thing.

Sometimes the DC is not the issue, but you must smooth-off the audio. Largish resistors with R.G.'s BFC often work fine. Don't discount that approach.

Emitter/Source followers are very stable for loads that pull "down". If the load pulls up, the transistor goes out of action, so you need a pull-down resistor and it may have to be hungry.

The TLE2426 is made for such use. Output current is not huge but ample for most pedals.

LM386 is a cute trick, but note that it idles at many mA power drain. Not an issue for a wall-wart, but will shorten battery life. And in a few cases, the noise or crossover distortion can be issues, suggesting another R-C at the output. And of course a LM386 with bypass and output filter is as much work/space as the opamp that you wanted to avoid.

A push-pull (two transistor) buffer is a good reference, but tricky to design for good balance of power and quality, and really more work (pins/parts) than a chip opamp.

Anything you can do with an opamp, you "can" do another way. The advantage of modern opamps is that you don't really have to think about it. The 39 cent opamp will do the job in most any stompbox project. Anything else, first you must define what you really need (1K ohm impedance? 1 ohm impedance? Uni- or bi-directional? Short-proof?) and then analyze your multiple alternatives.

Unless your brain-time is worth a lot less than 39 cents, the opamp is the universal answer when you need something more than the simple R-C network.

> probably an amateur question

It's not "amateur". It's professional. Experienced designers stay up at night thinking how best to avoid more parts, including reference voltages. For you it is 39 cents and a half-inch of board. So frikkin what? When you make a million, the costs make the difference between profit and bankruptcy.

[liquids]

I guess I should detail that, working with vero as I usually do, I personally have a harder time working around op amps than TO-92s. There's always the factor of total component count, and the space they take up, and cost.  But it seems like I'm back to using half of a dual op amp as the best answer in general, and if that's an issue, a simple voltage divider works fine (if it's limitation considered and/or compensated for), or, well, tough luck!    

I think a lot of my trouble is, when I first started working with basic resistor/cap voltage dividers, I was thinking that if Vref from the junction of two small resistors was good, one from two large resistors was better!  When my results were not satisfactory, I asked a very knowledgeable co-worker...and for all his useful tips, I missed the part about usingsmaller resistors, which might provide a more stable voltage reference.  

Anyhow, thanks for all the good ideas here.  Each has it's pros and cons, but it's always good to have those kind of options and know how to utilize their advantages and exploit their limitations!  

[head_spaz]

The beauty of the TLE2426 is that it's a high impedence voltage divider, with an opamp follower, in a three legged TO-92 case.
The high input impedence of the voltage divider draws extremely little current. Optimized for battery use.
The built in opamp buffer has a low impedience output whether you're trying to pull it up or push it down.
All that in a tiny little TO-92.
What's not to like?

OH wait... they're 72 cents EACH!!! What the hell was I thinking?
nevermind...
... nothing to see here...
... ... these are not the droids you're looking for...

[trjones1]

Is 72 cents too much to spend for a little elegance?  Believe me, as an amateur a little is all I'm looking for.

[Eb7+9]

So, does a JFET or BJT (etc) with it's drain/collector hooked up to V+, and it's gate/base with a voltage divider on it, and no other signal....does the emitter/source function as a good Vref?

practically speaking, if you can worse case the load currents and keep them under +/-10% of the bias/idling current then a BJT follower gives the best drive between it and the jFET one - and it does so without ringing like an op-amp follower does ... the BJT gives an output impedance that results from dividing what's seen at the Base by the device's current-gain - set the bias via a 10k trimmer and Zout is guaranteed less than 5k/100 = 50 ohms assuming beta of 100, with real devices it will be lower even ... apply Darlington and you roughly get 0.5 ohm as an estimate ... unless you need super low Zout this is plenty fine for biasing many types of circuits - but just in case the drive specs of one can be adjusted against the load specs of other, or vice versa - easily ... temperature drift is negligible in these circuits by way of NFB in the emitter circuit

[Jarno]

Not sure about the performance, but I've tried this:

It didn't track exactly in the middle though, so I replaced one of the resistors with a trimpot (one of the 4.7k ones).

[earthtonesaudio]

PRR, I totally agree with your post except for this part:

LM386 is a cute trick, but note that it idles at many mA power drain.

The LM386 datasheet states the quiescent current is 4mA.  Not as good as some other methods, but still pretty good for the price point.


But, if you're using Veroboard and size/ease of layout is paramount, that TLE2426 sounds really cool.

[liquids]

But, if you're using Veroboard and size/ease of layout is paramount, that TLE2426 sounds really cool.

Agreed - The TLE2426 is neat, and yeah its $1 or so, but I can see it being worth it in many cases. I  Had never heard of this device.   Only problem is, it's greatest strength is also a weakness - it's going to give me exactly 1/2 power supply.  But if it stays there under load, that's good enough for most circumstances.    But At 9v, 5v regular is the same size, lower cost, and offers that advantage of likewise being just enough above 1/2V to be 'good enough' for many circumstances.    At higher voltage designs, if a spare op amp isn't available and maximum swing isn't an issue, the TLE2426 it seems like a great fix.  Wish I had one right about now!  

The 386 makes sense, and seems easiest. Enter higher voltages... it can't handle voltage from a charge pump at 18v, so it's limited in that regard, outside of 9v.  Yeah, t's fewer parts, but resistors are not a big deal on the veroboard.   It's one chip on 8 pins when an op amp can be dual, and it's costlier to boot...back to the 5v regular if I just need a 9v Vref, or an op amp.

I've got a design I'm finishing up, where I've got ground, a charge pump voltage (24V+),  and then 3 other voltages levels going on-- one is just above ~1/2 the charge pump supply, another is a variable voltage supply controllable via a pot (op amp required) and then a ~1/2v off that variable voltage supply.  That's 5 voltages!    Can you say DIY?    

The variable voltage supply is definitely best with an op amp Vref.  So it's a no-brainer to use a dual op amp and use the other side for one of the Vrefs (ahh, but which one?   )...  

The design is probably going PCB due to it's size anyhow, but even so, I keep going back and forth about that 3rd 1/2 variable voltage supply - to op-amp or not to op amp.  And then, if only one of the remaining 2 use op amps, which one...  My compulsive side says, 'do it right, add an op amp!" while my conservationist side says, "don't waste 8 pins and space on one Vref."  I'll order a TLE2426 and demo it for future use if not this one if impatience doesn't get the best of me....

5. A resistor/resistor reference with a BFC (Big Freaking Capacitor) gets relatively better as the cap gets bigger and the resistors get smaller.

Likewise, I'm going to go back to this, since I'd mistakenly considered it 'poor design' previously.   I'll see how it works here with smallish resistors and that 'big cap,' and measuring the voltage under load to see how much voltage loss I can accept or compensate for.  Of my two remaining options for this particular project, I'd have to use 2 Rs and a C with an op amp anyway, so it may be the right compromise if the results are satisfactory.  

The TLE2426 will be a great addition to my parts stash in the future.   The 5v regulator is something I've rarely turned to, but now will for simple designs.  That will come in handy in short time.  For me, sometimes it's the obvious that's...well, the least obvious.  

[R.G.]

The 386 makes sense, and seems easiest. Enter higher voltages... it can't handle voltage from a charge pump at 18v, so it's limited in that regard, outside of 9v.  Yeah, t's fewer parts, but resistors are not a big deal on the veroboard.   It's one chip on 8 pins when an op amp can be dual, and it's costlier to boot...back to the 5v regular if I just need a 9v Vref, or an op amp.

Welcome to the design swamp. There are more alliga... er, design options than you can attend to at any one moment. 
LM386N-4 is rated 18V normal operation, 22V absolute max.
LM380 in an 8 pin dip is also rated for 22V continuous.
TLE2426 is in a TO-92, but costs the same or more, and is much less available; but goes to 40V input.
Charge pumps from 9V don't get to 18V - they get to 16 or 17V because of switching and diode losses.

The design is probably going PCB due to it's size anyhow, but even so, I keep going back and forth about that 3rd 1/2 variable voltage supply - to op-amp or not to op amp.  And then, if only one of the remaining 2 use op amps, which one...  My compulsive side says, 'do it right, add an op amp!" while my conservationist side says, "don't waste 8 pins and space on one Vref."  I'll order a TLE2426 and demo it for future use if not this one if impatience doesn't get the best of me....

It gets tricky relying on unconscious factors (conservative vs compulsive vs space vs cost vs quick vs performance) to choose a design approach. The standard approach to this is to make a conscious decision what is most important in a design, then second, and so on, to help make these decisions. For instance, for a cell phone design, the most important factor could well be space and size, with cost secondary. For an audiophile amplifier, it's distortion and mojo-claiming rights, and cost/size/heat/etc. can go hang. In a DIY setup, it can well be whether you can get part A versus part B, or whether you can etch PCBs or not. Impatience in the form of needing to meet a release schedule is also a big factor to the pros. In professional design work, you either get the design done on time, or work nights and weekends and then lose your job if you can't. All of these things can matter, so it helps to set out what the objective are, in priority order, ahead of time. This can help a lot in picking whether you choose designs on the basis of good, cheap, quick, small, etc.

Likewise, I'm going to go back to this, since I'd mistakenly considered it 'poor design' previously.   I'll see how it works here with smallish resistors and that 'big cap,' and measuring the voltage under load to see how much voltage loss I can accept or compensate for.  Of my two remaining options for this particular project, I'd have to use 2 Rs and a C with an op amp anyway, so it may be the right compromise if the results are satisfactory.  

You might want to read http://geofex.com/circuits/Biasnet.htm from 2002, which is a pop-technical version of how one designs these setups.  The bottom line is that (1) you have to know the size and variability of the currents pulled from your Vref; if you can't at least bound this, you are relegated to trial and error to get it right;  (2) you have to design the standing current through the resistors to make the output current negligible compared to the resistor currents; this is usually taken to be 1/10 or less of the standing current; (3) you have to make the AC impedance of the capacitor to ground be small enough to act as a good "ground" for AC signals at all frequencies of interest. That's really all there is to it, but you realize that this is a highly compressed rendition of "all there is to it". 

R-R-C networks are neither good nor bad design. It depends entirely on how you measure success. And sometimes it's not trivial to define what success is. Note that there is no smiley attached to that last statement.

[Tube2stomp]

Related issue, instead of Vref and staying single rail, I need/want to generate a dual rail and virtual earth from a single supply.
For a small preamp and not much current I know I can get away with a simple "med/lowish impedance" RCR divider and battery... but not this time
I'm building an on-board guitar preamp with about 3-4 dual opamps, so instead of battery I'm thinking to "phantom power" it with a single rail DC source and generate the dual rail/virtual earth in the guitar.

R.G. (because his article is the blame for this idea but anyone can chime in ), can you describe or sketch a practical circuit of how to use the LM386 for this?
I can also "donate" any of the usual suspects (TL07x/8x, NE5532/4, LF353, LM833) for the cause if they will do a better job.
A while ago, a google search revealed some discussions around the net on this topic, but most were for more powerful needs (like headphones amps) or not audio related... and agreement was never always reached

Thanks
BTW, I remember seeing P.R.R on some of those discussions, and if it's the same PRR here on this thread, then I must say... I bump into your posts all over the net all the time!
Impressive *thumbs up*

[liquids]

Not sure I totally understand what you are doing, but have you looked at the LT1054 datasheet?

Page 15 (read all you can manage though) Figure 19  http://focus.ti.com/lit/ds/symlink/lt1054.pdf  Less than $2, too ~100ma current.

Simple way to bipolar rails with that chip and two extra caps, basically (and maybe two regulators).  You'll have true ground as your reference, too, not a voltage approximation.

Probably not ideal for a battery though.  If you phantom power it, why not?   I'm using this myself...it's a good thing.  More headroom for your op amps...bias voltages are the assumption around here, but bipolar have their benefits and are easier to work with I think, too, when your rails are equal and ground is voltage stable ole' ground.  

That said, this thread proves there's more than one good way, let alone more than one way to do it...depending on your needs, likes, and priorities.