[f] AMB on Facebook

The σ78 and σ79 ultra low noise voltage regulators

Upgrades for the 78xx and 79xx 3-pin TO-220 fixed voltage regulators

The σ78 and σ79 ultra low noise voltage regulators

Postby amb » April 4th, 2016, 5:27 am

Status

- April 8, 2016: Updated to v0.2 - Minor PCB layout changes.
- April 10, 2016: Updated to v0.3 - more PCB layout adjustments, add more text about physical design and pin configurations.
- April 14, 2016: Prototype σ78 PCBs being ordered. PM me if you would like to get one to play with.
- April 30, 2016: Prototype σ78 PCBs are here.
- May 7, 2016: Prototype σ78 configured for 5V output is up and working. Testing in progress.
- July 4, 2016: Production PCBs ordered.
- July 6, 2016: σ78/σ79 website is now live. For most update information please visit the website.
- July 27, 2016: Production PCBs arrived.
- July 28, 2016: σ78 and σ79 released and available at AMB audio shop.

Introduction

As many of you know, the venerable 78xx and 79xx 3-terminal monolithic voltage regulators are not great performers compared to some of the modern LDOs, or PSUs like the σ11/σ22. But they are easy to use, requires minimum external parts, can pass a decent amount of current, and work adequately well for many applications. For more critical circuits such as low noise amplifiers (e.g., preamps, headphone amps) and ADCs or DACs, something better is desirable.

I am now developing the σ78 and σ79 ("sigma78", "sigma79"), an upgrade for the 78xx and 79xx. Each will be a small board made to be compatible with the TO-220 3-pin package. They can replace the 78xx and/or 79xx regulators found in a myriad of devices, including AMB's CK²III headphone amp (78xx and 79xx), the σ25 positive PSU board (78xx only) and σ26 negative PSU board (79xx only).

Here is a 3D render of a σ78, to give you an idea what it looks like, placed next to a TO-220 package. The σ79 is similar. Note that the top soldermask is not rendered correctly where there are rectangular regions of copper.

Image

Board photos

These are photos of σ78 boards, with and without parts mounted on it, and in comparison to a TO-220 device and a U.S. $0.25 coin.

Image

Image


Technical information

The σ78 is based on the Texas Instruments TPS7A4701RGW ultra low-noise (4µV RMS), wideband LDO voltage regulator rated up to a maximum of 35V input and 1A of current. The σ78 PCB carries the TPS7A4701RGW and its supporting resistors and capacitors, making it a self contained unit that should just "drop in" (after assembly) and replace a standard 78xx TO-220 regulator. The regulated output voltage is set by two resistors on the board (R1 and R2), allowing a range from 1.8V up to 24V (the latter is limited by the voltage rating of the output decoupling capacitor). This is even better than the standard 78xx series, where the lowest output voltage is 5V (the 7805). The pin-out is the 78xx series standard: IN-GND-OUT

The σ79 is similar, except it's based on the negative output voltage LDO regulator variant TPS7A3301RGW. The pin-out is the 79xx series standard: GND-IN-OUT.

These voltage regulators provide dramatically lower noise and better transient response than the 78xx/79xx, as well as excellent PSRR, load regulation and line regulation.

The schematic diagram is basically the standard application circuit from the Texas Instruments datasheets. The positive regulator operates in "adjustable mode" (i.e., output voltage is set by two resistors) rather than the "PCB-programmable" mode (output voltage is set by connecting various configuration pins to GND, see TPS7A47 datasheet). The negative regulator also operates in "adjustable mode", the TPS7A33 regulator does not support "PCB-programmable" mode.

Both the TPS7A4701RGW and TPS7A3301RGW are in a small 5mm x 5mm SMD quad flat no-lead package (QFN20), with a thermal pad on the bottom which need to be soldered to the board. On the σ78 and σ79, 9 vias on the thermal pad bonding area connects to the back side copper plane which acts as a heatsink, and the device can be mounted to an additional heatsink similar to a standard TO-220 device. Usually, no isolation pad is required, just use some thermal compound on the mating surfaces. Note that the screw hole (and the exposed back side copper plane) is GND on both σ78 and σ79. The standard 79xx series has the mounting tab connected to the IN pin instead. Most of the time this should not matter, but if you don't want the GND to be electrically connected to the heatsink, then use a TO-220 heatsink mounting kit (which should include an isolation pad, nylon shoulder washer, mounting screw and nut).

The regulator chips have built-in over-temperature and over-current protection. However, for safe operation, be sure to mount the σ78 and σ79 on sufficiently sized heatsinks for your specific application. If you are designing a new circuit based on the σ78/σ79, keeping the dropout (input-output difference) voltage low will help minimize the heat dissipation. Since these are based on LDO regulators, you don't need the 2+ volts dropout for the standard 78xx/79xx device. When replacing the 78xx/79xx in existing devices, you can use the original heatsink.

The PCBs have two copper layers, with the back side layer serving as a built-in heatsink and mating surface to an additional heatsink. The board material will be FR-4, 0.4125" x 0.8000" (10.48mm x 20.32mm) with thickness of 0.062" (1.6mm). The width of the PCB is about the same as a standard TO-220 package, but it is taller than the TO-220 body + mounting tab. The spacing between the top edge and the mounting hole is TO-220 compatible, but the bottom edge extends lower than the standard TO-220 body for the pin-mount area.

Physical design

A goal of the σ78/σ79 is to be physically-compatible with as many existing 78xx/79xx installations as possible, so that a direct swap can be done with minimum fuss. This requirement dictates that the σ78/σ79 be no wider than a standard TO-220 package, and minimally taller. If these requirements are not met, then there is a high chance that it may either not fit on an existing heatsink, or may physically interfere with other parts or the chassis (almost all of the other 78xx/79xx-upgrade implementations out there do not meet these requirements). The back side of the σ78/σ79 board must be flat (i.e., containng no parts or solder joints), so that it can be mated to a heatsink surface. There must also be no exposed pads or vias on the back side other than those for GND.

With these constraints, there is very limited amount of front-side PCB real estate for parts and traces. The result is a dense layout, yet it follows the layout recommendation from the datasheet very closely for the best performance,

TO-220 packages are often used in two configurations: horizontal-mount and vertical-mount. An example of horizontal is the 78xx on the σ25 PSU board. Examples of vertical are the 78xx/79xx on the CK²III, or the MOSFETs on the M³, β22, σ11 and σ22. Note that the subject of discussion here is just physical configuration, not to imply that a voltage regulator can be used in place of a MOSFET or vice-versa!

The LDO chip and its five passive components used up all the equivalent height of the TO-220 body + mounting tab. Thus it is necessary to make the σ78/σ79 board slightly taller in order to provide the solder pads for the three pins. The two mounting configurations, various heatsink designs and clearances became a challenge, because there isn't a single solution that will work for all configurations. With help from MisterX, we searched many different pin types and essentially came up empty. The current solution I am going with depends on the configuration.

I designed the σ78 such that when it is secured with its mounting screw on the σ25 PSU board in the default horizontal configuration (with or without the extra heatsink), the three "pin" solder pads are aligned perfectly with those on the σ25. You can then simply insert a resistor lead into each hole, through the two boards, and solder it on the front side of the σ78 and bottom side of the σ25. Thick resistor leads are recommended for lowest impedance.

Not every horizontal-mount configuration use the same distance (as the σ25) between the screw hole and where the TO-220 pins are bent downwards. For these other configurations, the resistor leads may be soldered to the top of the σ78 pin pads (laying flat, not inserted into the holes), and then bent down 90° to go into the matching holes on the PCB on which the 78xx was mounted.

The following picture shows the two horizontal configurations.

Image

For vertical mounting, a 3P 0.1" pitch straight pin header strip may be used as the three pins. It should be soldered to the σ78 pin pads (laying flat, not inserted into the holes). There are two ways to do this, as shown in the picture below:

Image

The first method will work if there is enough distance between the screw hole and the board surface to accomodate the pin header strip with its plastic portion in place. This is heatsink-specific. If it doesn't fit, then you could cut off the plastic, or use the second method instead. With the latter method, you install the pin header strip the opposite way, and cut the pins just above the plastic piece, which is easier to do. After you soldered the pins on the σ78 and cut off the plastic, you should use a clip to temporarily secure the pins on the σ78 while you solder the other end of the pin to the board below, so that it does not become desoldered from the heat.

You may need to bend the pins slightly forward in order for them to line up with the holes on the PCB below. Do this before soldering the pin header strip to the σ78 so that you don't rip the solder pads off.

The above discussion about the σ78 also applies to σ79. The negative version of the σ25 PSU is the σ26.

How will these be offered

The LDO chip is difficult to hand-solder not only due to the tiny leadless package, but also the need to solder the thermal pad to the PCB. Hence it should be done in a SMD reflow oven. My current thinking is to offer the σ78 and σ79 with the LDO chip pre-mounted on them. The builder would solder the passive parts. There are only 3 capacitors and 2 resistors. C1 and C2 are SMD 1210 package and should be easy to solder, C3, R1 and R2 are SMD 0603 and much smaller, but still manageable. In addition, a 3-pin header can be installed as the "TO-220 pins" as described above.

Technical highlights, schematic diagram, parts list, and other information

Please see the σ78/σ79 website.

5 people like this post.
User avatar
amb
Site Admin
 
Posts: 10355
Likes: 175 posts
Liked in: 478 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby linux-works » April 4th, 2016, 9:40 am

cool ;)

wonder if it makes sense to think about having an nvram-backed i2c digipot so that the voltage can be set via a controller, or 'programmed' as needed via simple i2c and no soldering needed.

circuit could also be soft-controlled over i2c from any controller (arduino, pic, rasp pi, etc) and be true adjustable regulator, in case you needed variable voltage for some application.

if you don't like that idea, perhaps at least bring out the programming resistor lead to a 'public' pin (yes, 4 wires instead of the classic 3) so that us controller guys can at least get easy access to it.
Sercona Audio (Control Systems for Analog and Digital)
Photos: http://www.flickr.com/photos/linux-works/
User avatar
linux-works
 
Posts: 2065
Likes: 1 post
Liked in: 40 posts
Joined: March 31st, 2010, 9:04 am
Location: Santa Clara, CA
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby amb » April 4th, 2016, 12:40 pm

@linux-work
Remember that this is a linear regulator. If you pull a non-trivial amount of current into a load, the heat dissipation will become a problem if the dropout voltage becomes large, which is what will happen if you set the output voltage much below the input voltage. I don't think this is a good candidate to have its output voltage be programmable "on the fly". It should be designed-in to a fix value. Not to mention that the little board is already very packed, with no room for another circuit trace and an extra pin, even if the pin is made to be optional.
User avatar
amb
Site Admin
 
Posts: 10355
Likes: 175 posts
Liked in: 478 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby linux-works » April 4th, 2016, 12:46 pm

sot23-6 is the size of a modern i2c digipot. there's room for that and 1 more pin.

since the voltage is settable by a programming resistor, I'm not following why setting it over a digipot is any different, in terms of heat (?)
Sercona Audio (Control Systems for Analog and Digital)
Photos: http://www.flickr.com/photos/linux-works/
User avatar
linux-works
 
Posts: 2065
Likes: 1 post
Liked in: 40 posts
Joined: March 31st, 2010, 9:04 am
Location: Santa Clara, CA
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby amb » April 4th, 2016, 1:16 pm

The programming resistors set output voltage once at build-time, and the input voltage is also pre-determined by the design of the circuit. So the dropout voltage is a known and set quantity. As I said in the first post, keeping the dropout voltage low will help keep the regulator running cool. If you make the output voltage adjustable "on the fly" without also changing the input voltage to match, then as you dial down the output voltage, the dropout voltage becomes larger, and the regulator will run hotter.

To give you a concrete example, let's say the input voltage is 30V and you set the output to 24V, and pull 100mA into a load. Then the power dissipation will be (30V - 24V) * 0.1A = 0.6W. If you dial the output voltage down to 5V while the input is still 30V, and pull 100mA into a load, then the dissipation becomes (30V - 5V) * 0.1A = 2.5W. Big difference! The situation gets even worse as the load current increases.

As for room for another SOT23-6 device, have a look at the layout. I put a SOT23-6 next to it and you tell me where to put it and its traces and pin?

Image
User avatar
amb
Site Admin
 
Posts: 10355
Likes: 175 posts
Liked in: 478 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby MisterX » April 4th, 2016, 1:38 pm

Bot regulators are available in a native TO-220 7 pin package.
Seems like that package would a better option for a "true adjustable regulator" hack to me.
1352.5 and counting.
User avatar
MisterX
 
Posts: 1143
Likes: 52 posts
Liked in: 36 posts
Joined: March 31st, 2010, 2:00 am
Location: Kalamazoo, MI
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby amb » April 4th, 2016, 1:49 pm

Only the negative regulator (TPS7A3301) is available also in the TO-220 7-pin package. The positive regulator (TPS7A4701) is only available in the QFN20 package.
User avatar
amb
Site Admin
 
Posts: 10355
Likes: 175 posts
Liked in: 478 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby MisterX » April 4th, 2016, 1:53 pm

Meh, I could have swore they both were. :(
1352.5 and counting.
User avatar
MisterX
 
Posts: 1143
Likes: 52 posts
Liked in: 36 posts
Joined: March 31st, 2010, 2:00 am
Location: Kalamazoo, MI
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby linux-works » April 4th, 2016, 3:35 pm

the trend is to add controllers to things; and it would be a shame to miss an opportunity to make a 'smart regulator' instead of yet another 3-term fixed voltage replacement.

if more leads are needed, so be it. maybe it could be cleverly done so that if you didn't need remote voltage control, you could omit 1 or 2 wires from the board to the user's circuit and maybe it would follow the usual 3 pin reg layout that people are used to. and if you needed extra pins for 'features', you'd be ok with it not being an old school style 3 lead board/module.

an enable line would also be nice. if a problem was detected and the controller needed to remove power fast, an enable line on the reg chip is a nice way to do it and its more common on modern regulator systems. the TI eval boards that I use all have enable lines on them since the chips themselves support that.

I have to admit that I would like to see more smart psu systems. we have enough of the old style; but ones that can take commands or give status are the future.

if you need to go a bit wider, I don't see an issue with that. rarely are to220's mounted neck to neck and many heatsinks will have a bit of space left, even on to220 style geometries.

I like controllers and I like things that offer unique value. I'd really love to see some smart psu elements that people can use to make safer and more efficient power systems.
Sercona Audio (Control Systems for Analog and Digital)
Photos: http://www.flickr.com/photos/linux-works/
User avatar
linux-works
 
Posts: 2065
Likes: 1 post
Liked in: 40 posts
Joined: March 31st, 2010, 9:04 am
Location: Santa Clara, CA
Country: United States (us)

Re: The σ78 and σ79 high performance voltage regulators

Postby amb » April 4th, 2016, 4:21 pm

I hear you about smart control, but this project is not where I would implement such a thing.

The whole point here is to make a better substitute for the 78xx and 79xx. Not to add other features. I don't want to make the board any wider or longer because that would prevent it from fitting many existing locations. For example, the default heatsink used on the σ25 board will not allow anything wider than the standard TO-220 package. I also don't want to make the board any taller, because that would also prevent it from fitting in some devices, due to physical interference with other parts or height of enclosure.

The dropout voltage issue I mentioned remains. If you want smart control, maybe it should best be applied to a switching regulator where efficiency is much higher and dropout voltage/heat dissipation are not usually a big issue. On-the-fly adjustable output voltage and linear regulators simply don't play well together. This is not just a σ78/σ79 issue, but any linear regulator.
User avatar
amb
Site Admin
 
Posts: 10355
Likes: 175 posts
Liked in: 478 posts
Joined: March 29th, 2010, 9:49 pm
Location: Sunnyvale, CA
Country: United States (us)

Next

Return to σ78 and σ79 ultra low noise voltage regulators

Who is online

Users browsing this forum: No registered users and 0 guests

Quick Links

AMB audio DIY resources
AMB audio shop