+- built-in power supplies vs. single-supply via wall wart (+1176 content)

[midwayfair]

One thing I like about guitar pedals is that their power requirements are fairly well-standardized.

Preamps and compressors, not so much. But there are a lot of common requirements, and a lot of "close" requirements. For instance, almost all mic preamps require +48V phantom power. +15/-15 isn't really that far away from that. So I've done a few preamps now that just run on 48V, and regulate down if I need to.

I also dislike building power supplies, so this has let me use wallwarts for all my stuff. While it's slightly inconvenient to have four different warts attached to my power strip, this also lets me hide the power supplies at least a foot away from any audio devices.

Assuming that I filtered the power section of any build very well and ensure that my devices are safely within the operating range of the wart I want to use, what are the reasons to go with a built-in power supply? It looks like it's a lot of downside: I need additional space inside the case, have to worry about orientation of the transformers to avoid hum, etc. Are there any benefits to the stability of most circuits (e.g. the ability to use ground as a rock-solid -- but not adjustable -- bias point).

Specifically, I'm looking at the possibility of building an 1176-style compressor that just runs on the 48V warts I use for mic preamps, and knowing that I don't have to build a power supply in the box would make the project very economical and the box much smaller.

 

[JohnRoberts]

Back at an earlier day job I dealt with similar concerns. I had over a dozen SKUs running from the same 16VAC wall wart. With an AC wall wart, I could scale voltages up with doublers, down with DC-DC , etc.  Of course some of my trick power supply applications could not be used in parallel with each other.

Looking for ways to power multiple units from a common PS you need to be careful about the common PS ground and keeping signals from being corrupted by currents flowing between chassis grounds (possible but should be anticipated).

JR

 

[midwayfair]

Back at an earlier day job I dealt with similar concerns. I had over a dozen SKUs running from the same 16VAC wall wart. With an AC wall wart, I could scale voltages up with doublers, down with DC-DC , etc.  Of course some of my trick power supply applications could not be used in parallel with each other.

Looking for ways to power multiple units from a common PS you need to be careful about the common PS ground and keeping signals from being corrupted by currents flowing between chassis grounds (possible but should be anticipated).

JR

Thanks as always. In my case, I'm specifically looking and dangers that come from using them in parallel. The +48V DC warts I have are HUGE -- one's 5A and the other's 10A and a single daisy chain has done a good job powering up to 6 peramps without a measureable (at least, with my crude method in the DAW) difference in noise or voltage loss, but I haven't tried it with a compressor of any sort. They're also reasonably quiet, as long as I'm careful about filtering regulators.

In the case of the 1176, -10V and +30V are really, really close to 48V, and under load after the PSU filtering, I typically get 40V. I'm speculating that there's some way for me to eliminate the negative rail entirely. I need to study the schematic more closely to figure out how to go about that, though. I don't think I can elevate ground, but the only - rail connections are the release and part of the ratio buttons, plus the meter ... FET? It's hard to tell because the schematic lines overlap.

I think I need to update the title of the thread.

 

[abbey road d enfer]

I'm speculating that there's some way for me to eliminate the negative rail entirely. I need to study the schematic more closely to figure out how to go about that, though. I don't think I can elevate ground, but the only - rail connections are the release and part of the ratio buttons, plus the meter ... FET?

Indeed, the current draw on the -10 rail is minimal.
Elevating ground presents too many risks (certainly makes it incompatible with other non-elevated ground units running on the same PSU); in regard with the light cost and ease of use of DC-DC converters, I would use one to create a -12V rail and drop it by 2V using a string of three 1N400x.

 

[JohnRoberts]

I'm speculating that there's some way for me to eliminate the negative rail entirely. I need to study the schematic more closely to figure out how to go about that, though. I don't think I can elevate ground, but the only - rail connections are the release and part of the ratio buttons, plus the meter ... FET?

Indeed, the current draw on the -10 rail is minimal.
Elevating ground presents too many risks (certainly makes it incompatible with other non-elevated ground units running on the same PSU); in regard with the light cost and ease of use of DC-DC converters, I would use one to create a -12V rail and drop it by 2V using a string of three 1N400x.

For light current draw you can also use a charge pump (capacitor charged to +V, then the top of that cap is driven to ground, charging a -V reservoir cap with a diode from the bottom of the charge pump cap).  I think I've even seen ICs that do this.

The good news is this doesn't create a HF magnetic field like a typical DC to DC switcher while the current spikes in the ground will still need to be managed.

JR

 

[midwayfair]

Thanks for the extra thoughts.

I think my main question has been answered, that as far as noise performance goes, I shouldn't expect any problems. This leaves a couple questions about the 1176, but rather than start a new thread I guess I can ask them here.

Indeed, the current draw on the -10 rail is minimal.

I think more to the point: Why does there need to be a NEGATIVE rail at all, instead of simply using ground as the lowest point? DC is all relative to another DC point. The amplifiers in the audio and envelope path will be just as happy on +40V as they are on +30V (or it's a simple transistor swap for something that's fine on the higher voltage). The decay capacitance shouldn't care if it's draining to ground instead of a negative voltage. The FET is being driven positive, so going up from 0V should be the same as going up from 10V. The op amp shouldn't care if it's 0/40V with a voltage divider Vb or -10/30V.

This leaves the handful of negative rail connections in the ratio. I *think* this is the only important part, as the negative rail sets a lower point than 0V when the audio path is running on 0/30V ... but I don't know what benefit that has.

As usual, I think the solution is "just try it and see what happens." The worst that happens is that it doesn't work perfectly and I have to build a power supply and redo a couple wire connections.

I think I've even seen ICs that do this.

LT1054 etc. can put out a negative voltage (they're not just doublers). But They're 15V max input voltage.

oh ... also: When I said elevate ground, I only meant making a voltage divider that puts a very well-filtered +10V spot, which is where all the current "0V" connections for the active elements go, and 0V is treated as the -10V rail, as well as just being the ground connection for the chassis, XLRs, transformers, etc. I still think it's a bad idea (all my DC biases would be messed up), but it shouldn't be incompatible with the power supply being daisy chained. Just, again, based on the understanding that DC is relative.

 

[JohnRoberts]

Audio is AC generally referenced to 0V.

Indeed you could set up an artificial ground at V/2 and many older designs did it to save cost.

There is no problem in theory with biasing audio up to V/2 and many modern A/D convertors do that inside the IC.

In practice you pass through multiple DC blocking caps in a given audio path. Again not a problem in theory.

Back in the bad old days they even made some audio power amps using single supply with big honking caps in series with the output (never a great idea).

JR

 

[Chris_V]

I also hate designing power supply, and put it inside the enclosure.
What I have done in recent design is to power the unit from a "standard" smps (12V, 24V or even from those cheap 19V laptop chargers). I then build all the necessary voltages using small DC/DC converters for each channels.

This is not the cheapest option, but it is very practical.
I have done a Neve style mic preamp, a tube preamp (based on the Old UA 610A) and I can build 1, 2 or 4 channels versions very easily, only one supply cable on the board. Testing and maintenance is also simplified.

TI as a usefull online tool to design DC/DC (and their controllers are quite cheap). The only drawback for DIY is that you have to deal with SMD component and at least a double side PCB (I am now using 4 layers PCB)

 

[PRR]

> Why does there need to be a NEGATIVE rail at all

Or why does there need to be a Positive rail?

The JFET is N-type, and connects to to ground. The control voltage must be negative. There are other ways to do it, but this designer picked a small negative supply.

The stuff positive of ground is very ordinary amplifier. It could have been built negative-hot. However it uses a Power transistor. At that time, Si NPN was a buck cheaper than Si PNP. Not that the cost was killer; in a failure, the tech was more likely to find a spare power NPN than PNP. All of this was probably adopted from some existing line-amp. The cost of merging a mostly-positive amplifier with a mostly-negative control circuit was considered cheap compared to other possible paths.

Another point is that the amplifier chain is BIG power, 30V 50mA. Inefficient Class A design. Simple, clean, serviceable. 1.5 Watts DC here. The control side is much smaller, like 0.1 Watts. "Stacking" them may be like stacking the lobster-pot on the tea-cup; stability is unsure.

This isn't "fake bi-polar" like an opamp biased half-supply. This is two separate loads doing different things.

Also the control path "may" benefit from not sharing a power supply too-closely with the big audio amp. I don't think so, but it is something to fret about.

Step back. All inputs and outputs are FLOATING. Transformers. If you insulate the chassis and don't connect the G pin, it seems you "could" tie V- to your shared common audio/power return. However any crap picked-up on the chassis, or the internal common, which should have a short direct path to system common, now has to flow through the V- filter cap. It's big. It may not be a problem. Note that there is also a Zener, so any RF which the filter cap doesn't completely clamp will be rectified, potentially police/taxi talk injected into the limiter guts. This is the kind of exposure which works fine one day one place and is un-usable another time and place. This drives traveling audiomen crazy.

There is also MUCH to be said for separate power transformer for every bit of kit. Shared audio common is curse enough. Also sharing power common begs for power crap to become audio crap. The shared common is often tolerable on a pedalboard: everything in a small space (short wires) and total demand quite small. When the system spreads over a 2-room studio, crap happens. When I managed a 660 seat concert hall, common anything was hopeless, each rack had to float from all others. (3VAC from stage to booth, varying AC from fly to recorders depending on recent rains.)

That said: figure what connects to "G" and don't expose it outside the box. Stack 10V and 30V BIG (10W) Zeners, feed about 100mA from your 48V supply. Your -48 system common is now "G" for this box.

I have done such a last-minute conversion on a limiter. At the very end I realized my meter-LEDs really had to go to common, and so the chip had to, but my control voltage was up at +14V center. I hacked-up a diff-amp on perfboard, and it worked (drifted like heck). Also all control voltage measurements had to be mentally corrected for the half-supply level. (Still using a hard-grounded VTVM, not a floating battery-meter.)

 

[midwayfair]

First: PRR, I hear you on avoiding daisychaining power supplies and the dangers thereof. Daisy chaining is at least as likely to insert crud into the audio as an internal power supply. On the one hand, the parts and transformer for putting the PSU in the case isn't THAT high -- I can do it for about $25. On the other hand, even if I have to use a separate power supply, it's the same number of spaces on the power strip, and even if I give up on the idea of converting it to single-supply and run it off a +15/-15 (or whatever) external AC supply, it's STILL the same number of spots on the strip. So ... I consider it at least worth pondering the possibilities here.

> Why does there need to be a NEGATIVE rail at all

Or why does there need to be a Positive rail?

The JFET is N-type, and connects to to ground. The control voltage must be negative. There are other ways to do it, but this designer picked a small negative supply.

Hrm.

I tried duplicating the function of the input stage on the breadboard with a single supply, and I need the source higher than ground if the gate is at ground, and then any voltage positive of the source drives the resistance down.

Gee, that sounds familiar. So, well, I know that works on some level; and if I just crib the 1176's method, the SOURCE would be the one to get that trimpot adjustment which will end up ... somewhere north of 0V and somewhere south of 9V, and the ratio buttons ... uh not sure about that yet.

Just to be clear, negative control voltage on the source of an N-channel FET drives the resistance UP.

The audio path's amplifiers don't have to change. It's only the envelope circuit and where the source goes

Some drawbacks: this means that the source can't actually go to ground. It has to a spot that just "sounds like ground," which means a very, very heft HPF (meaning, the elevated voltage is filtered). That's problematic -- it means that the gain reduction is now frequency dependent! The Orange Squeezer gets around this problem simply by being incredibly NOISY.

--

I don't have a PFET around to try that ...

--

The possibility of dropping in a charge pump as suggested above might work; an LT1054 can supply up to 100mA, which is way more than what's needed, and -10V is easy enough to get (I just tried it). This is the best option so far if I decide to skip the PSU in the box. Mainly because I don't have to change anything on a PCB designed for the circuit -- I just drop in the charge pump for the negative rail spot. I don't have to buy anything extra and it's a $2 chip and 15c in caps on perfboard.

---

For completeness, here's another option, though not a great one after messing around with it:

Use a MOSFET instead of a FET.

The drain-source resistance of a FET with the source at ground is several MOhms. A positive voltage of 1.5V is enough to drive it into oblivion. At half a volt, it's several dozen KOhms, and at a quarter volt isn't back up over 1M.

Ray Ring used a MOSFET on what is, hands-down, my favorite compressor on bass. (One of the notes specifies that "attenuation starts at 0.81V." Seems about right.)

There are a bunch of complications with that, enough that I'm not totally sure I even want to attempt it because of all the moving parts. I'd need to somehow get the control voltage to stay under a couple volts at absolute max and above a half volt, because there's only a volt, maybe 1.5V (if I include a silicon diode drop for rectifying), to play with when the amplifiers are running on 40V ... even if 1.5V is close enough to line level (i.e. as loud as I should be getting anyway), it still seems like it would be a lot of fiddling to get that to work. A FET scales WAY better to the actual audio signal and that means that the controls aren't as finicky.

 

[PRR]

> It has to a spot that just "sounds like ground,"

Well put.

> the gain reduction is now frequency dependent!

You have JFET resistance and some capacitor. If the capacitor impedance is "very small" compared to JFET resistance, it is flat-enough to the frequency where you computed Zc.

Cave-man: assume JFET won't go below 1K Ohms. 10uFd is "small" compared to 1K down to 20Hz. This will work.

Audio-geek: the JFET may go to a few hundred ohms. Then a few dozen uFd is "small". But "small" is still a 1dB error at 20Hz relative to higher freqs. OMG, horrible!! So go 10X bigger at least. 330uFd? 1,000uFd? Since the DC is only a few Volts (Vp + sidechain headroom), it may be a 16V part and won't bulge the budget much.

> Use a MOSFET

You are copying a known-good classic design. If I clone a Jaguar XKE, but stuff a Chevy V-8 in the key spot, it isn't going to work just-like a Jag XKE. It might come out better in some way, but in Artistic Terms that's not always what you want. If you want a "perfect" toy, you'd get a Honda 2000 or a THAT-based limiter: doesn't smear the bumps or leave spots in the driveway. Yet XKEs and JFET limiters have fan-boys.

 

[Andy Peters]

If you want a "perfect" toy, you'd get a Honda 2000 or a THAT-based limiter: doesn't smear the bumps or leave spots in the driveway.

I've had my Honda S2000 for nearly 11 years, and you're right, it doesn't leave spots in the driveway.

 

[midwayfair]

Found a 15V AC, 700mA wallwart in a box of my dad's computer stuff. No center tap, so I have to wrap my brain around that*, and not quite the voltages required by the Hairball. Might be good enough? Guess I'll find out when I get home.


*Do I just stick a third diode at the intersection of D2/D3 for my negative voltage rail?

 

[JohnRoberts]

Found a 15V AC, 700mA wallwart in a box of my dad's computer stuff. No center tap, so I have to wrap my brain around that*, and not quite the voltages required by the Hairball. Might be good enough? Guess I'll find out when I get home.


*Do I just stick a third diode at the intersection of D2/D3 for my negative voltage rail?

No,, your schematic shows a full wave rectifier, so both the positive and negative swings of the transformer are rectified into a single +V supply (there is no negative voltage there to rectify).

How much voltage do you get that way? and how much do you need.?

You can half wave rectify that transformer by grounding one end of the transformer winding and sending the other end through two steering diodes to charge + and - reservoir caps. But with half the current in each supply, vs before with FWR (full wave rectification).

I asked about voltage, because you can split that single  supply into two by letting it float and actively forcing it to split the supplies above and below ground. Note: for a heavy + load and light - load, you might have to dump a bunch of - current to make it share nice.

or use one of the previous suggestions,,,(charge pump, v/2, dc-dc etc).

JR

 

[midwayfair]

urh. Yeah, I'm dumb. It's only 15V total, not +-15V. Oh well.

Moving on!

 

[JohnRoberts]

urh. Yeah, I'm dumb. It's only 15V total, not +-15V. Oh well.

Moving on!

You can half wave rectify that transformer by grounding one end of the transformer winding and sending the other end through two steering diodes to charge + and - reservoir caps. But with half the current in each supply, vs before with FWR (full wave rectification).

If you get +15V FWR you can get +/- 15V using half wave rectification ... like I just said.

JR