'rail-splitter' 0V vircuits

I've seen plenty of discussions about creating a 'virtual 0V' at the midpoint for a single-rail power supply, e.g. turning a 48V rail into a +/- 24V equivalent.
  What I haven't seen is much, if any, discussion on the best way of splitting a floating power supply, e.g. switchmode. I'm looking at using one of the Meanwell 36V supplies split to +/- 18V for a preamp project and have ended up a bit confused about analysing possible 0V currents etc.

  Methods I've commonly seen for splitting rails are :
1- simple resistive divider, commonly only seen in guitar pedals etc
2- opamp buffered divider for better stability / lower impedance
3- variations on this using power opamps or more complex splitters to drive the opamp buffer, including dedicated 'reference' chips
4- discrete circuits like the 'Sijosae' circuit commonly used in battery powered heaphone circuits.

Any opinions on the relative merits or otherwise of these circuits for a mic preamp application with switchmode PSU? I'm leaning towards the discrete option, but would it be overkill for relatively small preamp currents?

Also, best point to connect Chassis Earth / external 0V to the circuit.

This depends entirely on what the virtual ground impedance needs to be. If you are just making a VG reference for an op amp that isn't throwing a lot of current around, then the VG impedance doesn't have to be perfect. It can be just a voltage divider with bypass caps. But if you are dumping a lot of current into the VG (like from a line output stage), it is not really practical at all because a positive current would have to be diverted to the negative rail and negative current would have to be sourced from the positive rail which means you have to have pass devices that can absorb that which is difficult and inefficient. So it depends on the circuit but generally trying to do an entire mic pre with a virtual ground is dubious.

The correct method is to use two SMPS and "stack" them where you connect the negative out of one SMPS to the positive out of the other SMPS and use that as 0V like in any other circuit. That will be a very solid ground. That have to be floating outputs but virtually all are because it is inherent to the design of SMPS.

But your probably not going to find 18V. Probably more like 24V in which case you can regulate down (or chokes with capacitance multipliers if you want to be pedantic about it).

I am no op amp expert but AFAIK there is absolutely no reason why am op amp has to be tun from split rails. It is simply a convenience that simplifies biasing at the mid point of the power supply. Equally, there is no reason the bias circuit need to carry amy more current than that required to bias the op amps. Just treat it as bias rather than a virtual 0V because if you run an op amp from a single rail supply then 0V is the 0V of that supply.

The only downside are you need to remove as much PSU noise as possible from the bias voltage (and a SPMSU may actually be an advantage in that respect) and your coupling caps need to be able to withstand at least half the supply voltage. The idea of running an entire preamp or mixer from a single 48V supply is very attractive and something I have toyed with on occasion.

Cheers

Ian

NOON said:

I've seen plenty of discussions about creating a 'virtual 0V' at the midpoint for a single-rail power supply, e.g. turning a 48V rail into a +/- 24V equivalent.
  What I haven't seen is much, if any, discussion on the best way of splitting a floating power supply, e.g. switchmode. I'm looking at using one of the Meanwell 36V supplies split to +/- 18V for a preamp project and have ended up a bit confused about analysing possible 0V currents etc.

  Methods I've commonly seen for splitting rails are :
1- simple resistive divider, commonly only seen in guitar pedals etc
2- opamp buffered divider for better stability / lower impedance
3- variations on this using power opamps or more complex splitters to drive the opamp buffer, including dedicated 'reference' chips
4- discrete circuits like the 'Sijosae' circuit commonly used in battery powered heaphone circuits.

Any opinions on the relative merits or otherwise of these circuits for a mic preamp application with switchmode PSU? I'm leaning towards the discrete option, but would it be overkill for relatively small preamp currents?

Also, best point to connect Chassis Earth / external 0V to the circuit.

Click to expand...

In fact that "rail splitter" is not very smart (IMO) for battery operation since it is always drawing (wasting) current.  In an application like that if you absolutely, positively, need a low impedance v/2 I would throw a low voltage op amp in there. The op amp internal current is wasted so a low current op amp might be advised.

Many many audio circuits work fine with only a single supply. Most older ones did.  Biasing audio 0V at real world 0V can eliminate a number of DC blocking caps, and their real (and imagined) audio problems.

Most op amps do not care if they are running from +/-15V or +30V/0V, while they do not like to operate very close to either rail so an op amp running from a single rail with input voltage that swings lower than it's -V rail will not be happy. That said specialty op amps have been engineered that will operate all the way down "to" the - rail without losing it's mind.

Finally I recall one type of op amps called Norton Amps that were intended to be operated from single supply with signals referenced to that - rail. The LM3900(?) a quad was popular for a while, but I don't see many around these days.

JR

@ Scott kind of early in the day to be drinking?

The idea of running an entire preamp or mixer from a single 48V supply is very attractive and something I have toyed with on occasion.

Click to expand...

I would have thought that 6.3V rails would be your forte

Martin Griffith said:

I would have thought that 6.3V rails would be your forte

Click to expand...

Since 48V is widely available as phantom voltage on many mic inputs, it is attractive to consider for simple audio circuits.

Back last century I made an active discrete FET input buffer, built into a 1/4" jack so musical instruments could plug direct into a mic input.

Carrying that concept to one extreme I designed a mixer input expander where the phantom power from one mic input supported multiple (non-phantom mics) with gain trim, pad, and polarity for each.    While clever this didn't make a lot of market sense when Peavey dealers would much rather sell the customer a new bigger mixer.  :

JR

PS: I have even seen phantom powered DSP with switching supply to better extract useful power from the soft 48V rail.

I know, good old P48.

For fun I recently tried to layout a poor man's Jensen type 990 DOA  with 0805 surface mount components, It was much more difficult than I thought, with PTH every component leg is also a via, which makes life much easier.  Ideal for +- 24 rails. I might even try it with smaller SMD components, if my plans to get a pick and place machine ever come to fruition. (I'm building a reflow oven at the moment)

ruffrecords said:

I am no op amp expert but AFAIK there is absolutely no reason why am op amp has to be tun from split rails. It is simply a convenience that simplifies biasing at the mid point of the power supply. Equally, there is no reason the bias circuit need to carry amy more current than that required to bias the op amps. Just treat it as bias rather than a virtual 0V because if you run an op amp from a single rail supply then 0V is the 0V of that supply.

Click to expand...

The problem is that if the load on the output of the op amp is not relatively high, whatever current is going into that has to return to the VG (assuming balanced load). So whatever the impedance of your VG is, it's like running the circuit with a resistor in series with your ground and that could affect CMR, noise, distortion, etc. Again it all depends on the circuit. If you drive the line differential with a dual op amp, that might work well because return currents are delivered to the right rail gracefully.

squarewave said:

The problem is that if the load on the output of the op amp is not relatively high, whatever current is going into that has to return to the VG (assuming balanced load). So whatever the impedance of your VG is, it's like running the circuit with a resistor in series with your ground and that could affect CMR, noise, distortion, etc. Again it all depends on the circuit. If you drive the line differential with a dual op amp, that might work well because return currents are delivered to the right rail gracefully.

Click to expand...

Sorry, I think you may have misunderstood me. If the VG only provides bias then there are no other currents flowing into or out of it. The signal 0V is the 0V of the single ended supply not the VG. We should stop calling it a VG and call it a bias voltage.

Cheers

Ian

JohnRoberts said:

@ Scott kind of early in the day to be drinking?

Click to expand...

Ha ha?

Of course?

But Ian's concept was new to me. Although I'm not learning in ss as much as I have been trying to learn tubes, it's an interesting connection and way to look at it for me at least... But I'm super green so, maybe it's old news....

For context: the "sijosae splitter" comes from the Cmoy headphone amp community. The supply is 9V battery. The amplifier is "no cap in signal path" (ha) so heavy headphone current returns to splitter. Cmoy implemented this as resistors and capacitors(!). In a few cases the low headphone resistance leveraged the DC bias off-center. Not understanding this lever, some builders devised stiffer splitters.

A "split" supply derived from a single supply will ALWAYS flow audio in and out through capacitors, one way or another. (Or audio transformers, but they are different.) Ian may be saying: be honest, put a dang coupling cap in/out, then your DC bias only has to be filtered enough to reduce rail-crap and cross-talk.

To clarify the exact situation, I have an existing  preamp board that uses THAT1512 chips for preamp and THAT1646 for balanced out. I have an application for smaller/cheaper/lighter than the current linear power I've been using and I thought this could be a good experiment with switchmode. If it works well I'll consider it for future builds.

I'd rather not modify it for single rail use if possible, a couple of spots would be tricky without a board redesign and I already have a batch of boards.

On a related topic, why does nobody make +/-18V +48V transformers or  switchmode units?! I would have thought there was a small but worthwhile market.

Why not use +/- 15V. Do you need 18?

NOON said:

On a related topic, why does nobody make +/-18V +48V transformers or  switchmode units?! I would have thought there was a small but worthwhile market.

Click to expand...

There are +/-15V +48V ones. They are laptop style bricks with a 5pin XLR looking connector on the end. I saw one on Mouser once for ~$30 USD. I think it was a MeanWell unit. It was discussed here but I don't recall enough about the thread to find it. The API 500 series chassis uses one. Donno if you can buy the PS separately.

Dan offers the 5 pin XLR supply

http://collectivecases.com/?product=psu-cc-psu1

I have a feeling split rail supplies only really came about because of the need for dc vacuum tube amplifiers.  Here is an interesting article on the early origins of op amps:

http://www.analog.com/media/en/training-seminars/design-handbooks/Op-Amp-Applications/SectionH.pdf

I rememeber when I first joined British Aerospace as an apprentice  in 1969 they had only just stopped developing Blue Streak which was the UK's first attempt at a commercial space rocket. Its control system was entirely tube based. During WWII there were lots of servo applications that needed  response down to dc. e.g radar and computers:

http://www.nj7p.org/Manuals/PDFs/Books/MIT-Radiation-Lab-Series-V18-Radar-Engineering.pdf

Cheers

Ian

Analog computers (the earliest application for op amps) were likely DC coupled.

JR

PS: I recall as a freshman in college seeing and old analog computer in a basement room (not working). And an IBM 360 digital computer filling another room that we used for doing our engineering homework on (in batch mode).

JohnRoberts said:

Analog computers (the earliest application for op amps) were likely DC coupled.

Click to expand...

At school, my first brushes with DC amps came with oscilloscope deviation circuits; they also had multiple supplies but were not opamps. Opamps came later in the year.

Completely off-topic, the Blue Streak Ian referred to uses one of the https://en.wikipedia.org/wiki/List_of_Rainbow_Codes, a WWII system for naming secret defence projects

My favourite amongst them is the infamous Blue Circle. This referred to a dummy radar pack used when developing fighter aircraft. The Tornado F2 had concrete ballast instead of the Foxhunter radar which was delivered late

This was an unofficial Rainbow Code, Blue Circle being a UK brand of concrete...

Nick Froome