'rail-splitter' 0V vircuits

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
> I have a feeling split rail supplies only really came about because of the need for dc vacuum tube amplifiers.

Operational amplifiers generally. Mechanical or thermonic.

The natural field of Operational Amplifiers is Real Numbers. Both sides of Zero. (And in practice something less than infinity; we do not have infinite grasp.)

Some problems appear to have only positive solutions. Gun-aiming, you want the shell to go toward the enemy not behind you on your own mess-hall. The altitude of an airplane should always be positive. However the partial products needed to compute error of angle and powder or rate of climb will often run negative. Windage may be east or west, climb may be up or down, which are +/- analogs naturally.

We could run a single +400V rail, define "+200V = Zero", subtract 200V from all our inputs and outputs. This is cumbersome, messy, and error-prone. It seemed to be (usually is) simpler to run bipolar rails (mechanical hangers or electronic power supplies).

 
PRR said:
> I have a feeling split rail supplies only really came about because of the need for dc vacuum tube amplifiers.

Operational amplifiers generally. Mechanical or thermonic.

The natural field of Operational Amplifiers is Real Numbers. Both sides of Zero. (And in practice something less than infinity; we do not have infinite grasp.)
digital computers have trouble with infinity too... try to divide by zero.... :eek:

JR
Some problems appear to have only positive solutions. Gun-aiming, you want the shell to go toward the enemy not behind you on your own mess-hall. The altitude of an airplane should always be positive. However the partial products needed to compute error of angle and powder or rate of climb will often run negative. Windage may be east or west, climb may be up or down, which are +/- analogs naturally.

We could run a single +400V rail, define "+200V = Zero", subtract 200V from all our inputs and outputs. This is cumbersome, messy, and error-prone. It seemed to be (usually is) simpler to run bipolar rails (mechanical hangers or electronic power supplies).
 
Fascinating where these discussions go sometimes, via WWII radar and opamp theory.  :)

No firm conclusions yet though.
To get more specific about it, if I have an existing preamp board (based around a THAT 1512 preamp chip and THAT 1646 output driver) designed to work with a conventional linear +\-18V supply and I have a Meanwell LRS-35 36V power supply  https://www.meanwell.com/productPdf.aspx?i=395 what is the general consensus on the best way of using them together?

There's a reasonable explanation of the options here. http://tangentsoft.net/elec/vgrounds.html

Cheers!
 
NOON said:
To get more specific about it, if I have an existing preamp board (based around a THAT 1512 preamp chip and THAT 1646 output driver) designed to work with a conventional linear +\-18V supply and I have a Meanwell LRS-35 36V power supply  https://www.meanwell.com/productPdf.aspx?i=395 what is the general consensus on the best way of using them together?

The simplest solution IMO would be using LM317 reg for creating +18V from +36V and using it as reference (common, ground). 
 
moamps said:
The simplest solution IMO would be using LM317 reg for creating +18V from +36V and using it as reference (common, ground).
I don't think that will work. LM317 is not designed to sink current.

As long as only relatively high impedance bias voltages are being used, then a simple virtual ground can be used. An op amp will make it stiffer if you have an extra one anyway.

The THAT1646 can sink and source a lot of current but it is a differential driver so it should be ok with a simple virtual ground.

Just remember that you need DC blocking caps. And because the load on the output can be low, you will need pretty large electrolytics on the THAT1646 output lines (like 220u).

But in general, I personally think using a VG is dubious. It will not universally work with all circuits. If you have a filter for example that is shunting frequencies to ground, that is a possible path for high current which may strain the VG. So it just requires extra thought. It would be much better to just use a proper bi polar supply but I recognize that it is useful if you only have single supply (because you want the whole thing to run on a laptop brick).
 
Running audio from single voltage power supplies is very old, so there are many examples around.

The details of what is optimal will depend on the specific application.

JR

 
squarewave said:
I don't think that will work. LM317 is not designed to sink current.
It will work, IMO.  If the output current from LM317 is too small, it can be loaded with 390 ohms/1W resistor for example  to ensure stability.  In the attachment is a similar solution.
As long as only relatively high impedance bias voltages are being used, then a simple virtual ground can be used. An op amp will make it stiffer if you have an extra one anyway.

Biasing THAT1512/1646 combo isn't  simple:
http://www.proaudiodesignforum.com/forum/php/viewtopic.php?t=191

 

Attachments

  • rail.png
    rail.png
    75.3 KB
It's a floating 36v supply and I'd prefer to treat the preamp circuit like a 'black box' with no modifications like biasing it to a halfway voltage, adding blocking caps etc.
Simple answer is add a couple of resistors, one from each rail to circuit 0v then nail that point to chassis ground. I assume there is a more optimal answer, but most people seem to be as confused as me about it. :)
I assume worst case scenario would be running unbalanced into a low impedance, so one side of the 1646 is being driven hard, but I keep thinking myself into unrelated backwaters and getting distracted.
 
NOON said:
It's a floating 36v supply and I'd prefer to treat the preamp circuit like a 'black box' with no modifications like biasing it to a halfway voltage, adding blocking caps etc. 
"floating" is important if you want to make things simple.

Simple answer is add a couple of resistors, one from each rail to circuit 0v then nail that point to chassis ground. I assume there is a more optimal answer,
Certainly. It may work somewhat, but any current unbalance between the positive and negative rail is going to be a problem. Indeed you need at least a big capacitor to decouple the rail.
I would feel better with an active rail-splitter; after all, the difference between a passive (R-R-C) rail-splitter and an active one is about the same as the difference between an unregulated supply and a regulated one. Both may work, but the pros and cons of each solution must be known and understood.
I suggest you study the schematic of the original dbx units such as the 160/161/162; it starts with a 723-based single rail regulator and adds an active rail splitter based on an LM301 and a couple of complementary transistors.
 
I'm not absolutely sure you can do that because all of your power supply bypass caps will be connected to the VG.

But otherwise yes, if you connect the VG to the chassis (pin 1s of XLRs), the SMPS is floating so you can skip the blocking caps and your circuit is totally standard.

But you will need a stiffer VG than some resistors. Connecting it to 0V will not "nail" it down because the current running into / out of ground has to go to the corresponding rail. I think that is probably the part that is confusing you. The impedance of the 0V needs to be MUCH lower impedance than anything connected to it. So if you're doing a basic THAT 1512 / 1646 circuit, then the lowest impedance nets are going to be the 5K on the ref and din- pins (assuming you use the 22K resistor on the inputs which you should). For 5K, a simple voltage divider with a unity gain op amp buffer should be good enough.

But again the power supply bypass caps might cause instability because, in the case of the VG op amp buffer at least, you would be connecting the output of the OA to the power supply rails through caps which is a dubious thing to do.

Also, any offset from the op amp is going to be reflected on the outputs.

So you want a low offset, low noise, slow op amp for the VG buffer. An NE5534 with offset adjusted might actually not be horrible. No doubt there are specialty op amps that would be better though.
 
squarewave said:
But again the power supply bypass caps might cause instability because, in the case of the VG op amp buffer at least, you would be connecting the output of the OA to the power supply rails through caps which is a dubious thing to do.
That's why some precautions must be taken in order to ensure stability with a heavy capacitive load, which is not that hard to do actually.

Also, any offset from the op amp is going to be reflected on the outputs.
I don't get it. Since the output of the rail-splitter is connected to ground, what offset are you considering?
 
abbey road d enfer said:
I don't get it. Since the output of the rail-splitter is connected to ground, what offset are you considering?
True. There is an offset but since everything is hanging on the output it's not an issue.
 
> I assume there is a more optimal answer, but most people seem to be as confused as me about it. :)

Me not confused. Sell the single supply and buy a dual supply.

Sorry if you do not like this answer.

 
"Sell the single supply and buy a dual supply. "

If this was a professional design group, I would heartily support this answer, however I think the OP deserves an other answer:
A) because this is DIY, and particularly the OP expressed his wish to use something he has in hand
B) because he admits having trouble understanding the issue, and the Drawing Board has an educational vocation
 
Back
Top