Rail to rail, single rail op amp

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

ruffrecords

Well-known member
Joined
Nov 10, 2006
Messages
16,107
Location
Norfolk - UK
As part of the design of a LED VU meter I need to design a precision rectifier. There are plenty of circuits for this but I need a single supply version (to work on a 12V supply). There are plenty of circuits for too but they rely on  a rail to rail voltage swing capable single supply op amp. There are plenty of these too but they all seem to be micro power types and they are quite expensive - over £3 for a single. What I am looking for is jelly bean op amp of this type - kind of a single rail, rail to rail swinging TL072.

Any ideas?

Cheers

Ian
 
Microchip has a decent cmos opamp process while not as quiet or cheap as jelly bean TL07x

http://ww1.microchip.com/downloads/en/DeviceDoc/mic7122.pdf

TI has a nice 12v CMOS opamp http://www.ti.com/lit/ds/symlink/opa743.pdf  but I have no idea of price point.

CMOS and higher rail voltages are not compatible.

I have no hands on experience, and speed/gain bandwidth may be an issue for accurate low level HF rectification, but that is true for standard op amps too.

I have seen very old school approaches to use CMOS gates in series with standard op amps to provide rail to rail capability while you need to use tricks to stabilize them and current consumption when CMOS gates are used linear can rise.

JR
 
Thanks for the input JR.I should also have mentioned I prefer it in a DIP package - a single SMT device in amongst a load of tubes and through hole components seems a bit odd.

Microchip's  MCP601 seems to be in common use, it is dirt cheap and available in PDIP but its max supply volts is only 6V. I guess I could just use a 78L05 to power it from 12V and pad the input signal appropriately.

Cheers

Ian
 
ruffrecords said:
Thanks for the input JR.I should also have mentioned I prefer it in a DIP package - a single SMT device in amongst a load of tubes and through hole components seems a bit odd.

Microchip's  MCP601 seems to be in common use, it is dirt cheap and available in PDIP but its max supply volts is only 6V. I guess I could just use a 78L05 to power it from 12V and pad the input signal appropriately.

Cheers

Ian
If you can run it from 5V do you really need rail to rail op amps?

I have done too many meters/rectifiers to offer one best way.

JR

PS: FWIW I did have an issue last century when a meter front end running from only 12V saturated from hot audio levels and caused perceptible tracking errors driving old school true VU mechanical meters.. Of course with LED meters you can scale levels as appropriate.

 
> that is necessary to do precision rectification with a single rail supply.

For some definition of "precision". (It always falls short... how much low-level and high-frequency can you spare?)

IMHO, it is quite hard to beat the Neve. They got PPM-deep audio bandwidth, and the rectifier is one transistor and two diodes, all jellybean. (Analysis is left to the student.) The log converter is a fistful of diodes and gives couple-dB accuracy without trims; if you have a better/smaller logger then this all reduces to a resistor.
 

Attachments

  • Neve-PPM-hack-42.gif
    Neve-PPM-hack-42.gif
    12.3 KB · Views: 53
ruffrecords said:
If you can run it from 5V do you really need rail to rail op amps?
Yes, because that is necessary to do precision rectification with a single rail supply.
I don't get it.
I understand the need for single-rail operation, actually most opamps are capable of single-rail operation. I guess you mean the capability to operate with one or both inputs at ground or V+ potential. But does it need to be rail-to-rail also?

https://www.analogictips.com/amplifiers-rail-to-rail-single-supply-mean/
 
PRR said:
> that is necessary to do precision rectification with a single rail supply.

For some definition of "precision". (It always falls short... how much low-level and high-frequency can you spare?)

IMHO, it is quite hard to beat the Neve. They got PPM-deep audio bandwidth, and the rectifier is one transistor and two diodes, all jellybean. (Analysis is left to the student.) The log converter is a fistful of diodes and gives couple-dB accuracy without trims; if you have a better/smaller logger then this all reduces to a resistor.
Very interesting. I shall investigate this further.

Cheers

Ian
 
abbey road d enfer said:
I don't get it.
I understand the need for single-rail operation, actually most opamps are capable of single-rail operation. I guess you mean the capability to operate with one or both inputs at ground or V+ potential. But does it need to be rail-to-rail also?

https://www.analogictips.com/amplifiers-rail-to-rail-single-supply-mean/
A Typical single supply precision rectifier uses the op amp in an inverting configuration with the + input grounded. When input goes negative the output goes positive. When the input goes positive the output clamps at 0V provided the output will swing down that far. So rail to rail per se is not required but it is necessary for the output to swing down to 0V.

The classic precision rectifier needs a split rail supply. I could produce an artificial 0V at half the supply voltage and use regular op amps but the dc output would then be referenced to the artificial 0V but the LED driving comparator resistor chain is referenced to 'real' 0V - unless I reference it to the artificial 0V. Hmmm, something to think about.

Cheers

Ian
 
ruffrecords said:
A Typical single supply precision rectifier uses the op amp in an inverting configuration with the + input grounded. When input goes negative the output goes positive. When the input goes positive the output clamps at 0V provided the output will swing down that far. So rail to rail per se is not required but it is necessary for the output to swing down to 0V.

The classic precision rectifier needs a split rail supply. I could produce an artificial 0V at half the supply voltage and use regular op amps but the dc output would then be referenced to the artificial 0V but the LED driving comparator resistor chain is referenced to 'real' 0V - unless I reference it to the artificial 0V. Hmmm, something to think about.

Cheers

ian

Cheers

Ian
OK, got it, so your suggestion of referencing to the virtual V/2 seems to be the way to go.
 
ruffrecords said:
Yes, because that is necessary to do precision rectification with a single rail supply.

Cheers

Ian
Perhaps you can expand on what you mean by "precision" rectification. 

I generally took that to mean an active rectifier topology that did not exhibit diode drops from the forward voltage of diodes. Further precision rectifiers can detect full wave or both polarities of signal (more important for peak meters than average).

Lastly as I have shared before, typical active rectifiers run out of gas (actually open loop gain) to track very low levels at high frequency. Active rectifiers generally have to slew from -1 diode drop to + 1 diode drop every time the signal passes through zero.

JR

PS: The fanciest precision rectifier I did was inside my TS-1 design... but that was far more than a LED level meter, it actually read out in dB. I used a cheap transistor array for the log conversion.  I have posted that schematic before again TMI for this thread. 
 
abbey road d enfer said:
OK, got it, so your suggestion of referencing to the virtual V/2 seems to be the way to go.
I sketched it out and it looks as though it should work. Just need to remember the LED comparator resistor chain is in parallel with the top resistor of the divider creating the virtual 0V so just need to make it perhaps 10x the value or add a compensating resistor to the bottom leg of the virtual 0V divider to compensate. If I was very smart I could make the comparator resistor chain BE the top resistor of the virtual 0V divider.

Cheers

Ian
 
ruffrecords said:
If I was very smart I could make the comparator resistor chain BE the top resistor of the virtual 0V divider.
I would think you're smart enough to figure this out. Anyway you don't really need the virtual reference to be exactly V+/2; actually it may be beneficial to shift it lower in order to increase the positive output headroom.
 
abbey road d enfer said:
I would think you're smart enough to figure this out. Anyway you don't really need the virtual reference to be exactly V+/2; actually it may be beneficial to shift it lower in order to increase the positive output headroom.
I am not sure shifting it would work because the precision full wave rectifier needs to swing as far in each direction to produce both half cycles. Here is a link to the precision rectifier schematic:

https://www.eeeguide.com/precision-full-wave-rectifiers/

Cheers

Ian
 
ruffrecords said:
I am not sure shifting it would work because the precision full wave rectifier needs to swing as far in each direction to produce both half cycles. Here is a link to the precision rectifier schematic:

https://www.eeeguide.com/precision-full-wave-rectifiers/

Cheers

Ian
Actually you could scale the R values to deliver similar headroom with asymmetrical artificial ground (aka v/2 but not /2).

Since they don't number the Rs I can't say which exact R but imagine making the two Rs in the top middle of the fig 2.63 , 1/2 or less the resistance of the others Rs. As long as those two Rs are the same as each other the signal at the output will track in both directions. This would allow it to handle full signal in both directions with less swing from the op amp in the one direction with lower resistance values.

JR
 
Another solution is to use the alternative circuit (see attachment).
On negative signals the upper opamp can go into clipping but it does not  matter since it is disconnected from the output. Slower opamps may introduce parasitics due to recovery, but a TL0xx should not cause any serious problem, unless high accuracy at 20kHz is a requisite. Adding a small cap between output  and negative input should tame this.
The attack time is governed by the current capability of the opamps and can be lengthend with the series resistor. The release time is govened by the capacitor and the 100k resistors AND the load resistance, if any.
 

Attachments

  • full wave rectifier.jpg
    full wave rectifier.jpg
    24 KB · Views: 23
> Here is a link to the precision rectifier schematic:

There are dozens/hundreds of "precision rectifier" schemes.

With limited supply options, one cute one is an inverter with input biased at the negative supply. For negative inputs it makes a positive output. For positive inputs it just slams the negative supply; so we add a resistor so the positive input drives the output positive. (This will usually need buffering so may not be a minimum-parts solution.) That's where a true To-Ground In/Out (and no overdrive quirks) opamp is wanted.

The problem with biasing a conventional +/-15V scheme with +12V zero and a 6V Vref is the output rids on the Vref. I painted myself in a corner this way once (changed my paln mid-build and didn't think-out all the consequences), and perf-boarded a diff-amp to get back to zero ref (and had to fine-trim that to keep the bottom LED from glowing).

The "crude" Neve plan is not wide-range precision like a hearing aid WRDC (which does useful math from 20dB SPL to >100dB SPL) but is ample for most program monitoring.

Oh, I forgot to add-back the peak-catching capacitor in my simplified plan. A good implementation may want Q4 and an offset diode to free time-constants from the rectifier impedance.

If you were another kind of geek, the $1 PIC processors are THE 21st century way to go. JR has done some with this but he's busy brewing beer. The algorithms I see outlined for hearing aid level sensing don't make sense to my VU-oriented experience, but frankly it's just dancing lights and different fiddlers don't much change that party.
 
PRR said:
> Here is a link to the precision rectifier schematic:

There are dozens/hundreds of "precision rectifier" schemes.
yup
With limited supply options, one cute one is an inverter with input biased at the negative supply. For negative inputs it makes a positive output. For positive inputs it just slams the negative supply; so we add a resistor so the positive input drives the output positive. (This will usually need buffering so may not be a minimum-parts solution.) That's where a true To-Ground In/Out (and no overdrive quirks) opamp is wanted.
I recall that approach being pimped by the op amp company who made an op amp that worked that way... clever in a gee whiz way...
The problem with biasing a conventional +/-15V scheme with +12V zero and a 6V Vref is the output rids on the Vref. I painted myself in a corner this way once (changed my paln mid-build and didn't think-out all the consequences), and perf-boarded a diff-amp to get back to zero ref (and had to fine-trim that to keep the bottom LED from glowing).

The "crude" Neve plan is not wide-range precision like a hearing aid WRDC (which does useful math from 20dB SPL to >100dB SPL) but is ample for most program monitoring.

Oh, I forgot to add-back the peak-catching capacitor in my simplified plan. A good implementation may want Q4 and an offset diode to free time-constants from the rectifier impedance.

If you were another kind of geek, the $1 PIC processors are THE 21st century way to go. JR has done some with this but he's busy brewing beer. The algorithms I see outlined for hearing aid level sensing don't make sense to my VU-oriented experience, but frankly it's just dancing lights and different fiddlers don't much change that party.
Ian may be that kind of geek but has expressed disinterest in clock frequencies et al.

I killed a few too many brain cells and even started at least one thread here chewing on the idea of an universal digital LM391x replacement.

I generally ended up including too many features and optimal tolopogy (IMO) was cheap 3.3V micro pushing serial data to a LED latch/driver. Using unused pins of the micro to select between multiple modes allowed the general approach to cover many bases.

I did not perceive enough interest to pursue, and didn't want to end up with even more kit components in my back room.

JR

PS: I prefer to perform the rectification after the A/D despite costing one bit of resolution. You gain the benefit of complete control of time constants and the like without any more glue components.
 
> pimped by the op amp company who made an op amp that worked that way... clever in a gee whiz way...

I very seriously considered it to monitor street-power into my house. Drive from a Current Transformer was very ample and 2 phases all fit on a quad opamp.

Then I realized the Chinese would sell me the whole package far cheaper than I could buy CTs and meters. One of two $11 meter-modules had a bum back-light so I got two more as spares; still cheaper.

You can buy "LED VU Meter" modules far cheaper than you can build them. Many are aimed at across-speaker use and have terrible input sections (but one TL072 might overwhelm their flaws). Of course long-term supply rules-out auction sites unless you can capitalize the initial try-them-all then place a Lifetime Buy.
 

Latest posts

Back
Top