Power Supply Questions

[Electron Tornado]

I'm working on a solid state amp project that I would like to run with a supply voltage of 24V. The schematic can be found here:  http://www.ssguitar.com/index.php?topic=1789.0  . The data sheet for the amp chip can be found here:  http://pdf1.alldatasheet.com/datasheet-pdf/view/25041/STMICROELECTRONICS/TDA2009A.html  . My amp is based on the schematic in Figure 13.

I would like to use an unregulated power supply, but I have a question - how do I determine the current requirement for the amplifier? I know I = SQRT(P/R), but with P=10 Watts and R = 4 ohms, then I = 1.58 amps. From what I've read, that would call for a transformer rated at 3 amps. Seems a bit much. 

Another question. The chip data sheet gives specs for a supply voltage of 24V and it says the chip's absolute max voltage is 28V. Would I be safer to use a supply voltage of 20V or 22V?

Thanks for any help.

[iccaros]

ET

So when you say unregulated, are you going to do the rectification and filtering yourself? if so 24v is hard to get you would need  a 18v transformer

24v will be fine, and 3A is not that much, but 2amps would work if the output amp and load was all there was to finding the power draw, the preamp will draw current and along with diodes and tone stack.. if you are willing to use some voltage regulation parts express has some 24v 2 and 3 amp supply for cheep 

by the way if you use a  4 ohm load and you are getting the 11.5 volts out you would pull 11.5/4 = 2.88 amps from that amp.The Chip is rated to pull 3 amps, and put out up to 20watts max. 
@18v that Chip will put you closer to 10 watts.

[R.G.]

You would think this would be an easy question to answer, right? It's not.

First off, +5 points for thinking to look at the data sheet first. The absolute maximums are the place where the manufacturer says that if it lives past this, you're lucky. It's a good idea never to get there. And that's where the design process starts.

You get to the highest power supply voltage in general at lowest load and highest incoming power line. The AC power line in the USA is usually 120-125V, and may go higher than that for brief periods. I've seen 132Vac. If your power transformer primary is rated at 120Vac, and it's being fed 132Vac, the output voltage wants to be the rated output times 132/120 = 1.1 times the normal value. If it's rated at 115Vac, and fed 132Vac, then the output tries to be 132/115 = 1.15 times the rated voltage.

Worse, there is transformer "regulation" to take into account. A transformer rated for 120Vac in, 10Vac out at 2A (just to pick an example out of thin air) is only going to put out 10Vac with a resistive load of 5 ohms. There is some resistance in the windings, and to get 10V out at full load, the transformer maker actually made the transformer secondary put out, say, 11Vac at no load. It sags down to 10Vac because of the winding losses at full load. They call this "regulation"; actually, it load induced sagging. What that means is that when your amplifier is not pulling much current, the power supply voltage increases because the amplifier is not pulling the transformer down.

On top of that, you will be using a full wave bridge rectifier and capacitive filter. The nature of these things is that you lose two diode drops in the bridge, and the capacitor charges up to max when the rectifiers conduct, then runs down linearly with load until it gets the next charge pulse. So there is ripple on the DC you make, and at light load, the capacitor charges up to the highest possible voltage.

So you have to start at the biggest-ever voltage and work down. 28Vdc is absolute max. We'll say that happens at high AC line voltage and lowest load. The datasheet says the quiescent current for the chip is 60ma or less for normal power supply voltages. This is so much smaller than it would use under large signals, we'll ignore it and say the idle current is zero. So the filter caps are charged to 28Vdc by the full wave bridge. The diodes will have small drops under these conditions, so we'll guess they are losing 0.6V each, and the peak voltages feeding the diodes to get 28Vdc peak are 28Vdc + 1.2V = 29.2Vpeak.

Transformers are always rated in RMS voltage. The RMS value of a sine wave is related to the peak by a factor of 1.414..., so the transformer secondary could only put out 29.2/1.414 = 20.65Vrms and not be killing the chip.

That's got to be under high line conditions and no load. To figure out what to buy, we have to correct down for the presumed high AC power line and down again because the transformer is not being loaded, so the regulation "sag" is not in there. You get to pick what tolerance the AC power line has based on what you think will happen, and you're betting your chip's life that you're right forever more.    So let's say we will buy a transformer rated for 120Vac on the primary and the highest tolerance up will be +10%, or 110% of nominal. So to figure out what the maximum rated secondary voltage, we have to subtract 10% from the 20.65Vrms, to get 18.77Vrms.

That's the peak AC rms voltage at nominal AC power line voltage, and low load. It's not the voltage rating printed on the transformer box. If you pick a transformer and read its data sheet carefully, you may find an "open circuit voltage" rating. This is the voltage the secondary puts out with nominal AC line input and no load. It will be higher than the rated secondary voltage under full load. The smaller the transformer the worse regulation sag gets. For about 20W transformers, it's in the range of 8-12% usually. This varies by manufacturer, construction, phase of the moon, and other things. Read the datasheet for the transformer you're thinking of buying. Let's pretend we get one with a sag of 11%. So the unloaded 18.77Vrms becomes (finally!) a rated 18.77/1.1 = 17.06664... volts AC rms.

What did we just figure out? To keep from exceeding 28Vdc under low/no signal conditions and high AC power line voltage, you can't use a transformer of more than 17Vac with the additional conditions that (1) the primary is rated for 120Vac (2) the AC power line will never go above 132Vac (3) the transformer sag is 11% or more.

It also means that for normal AC power line voltages and modest output levels, the output voltage will be lower, and the chip will not be able to reach full power because its power supply won't be as much as 28Vdc. That difference is the price we pay for the chip not dying under bad conditions. If you get a transformer with a tighter "regulation" (i.e. smaller sag) you can up the voltage because it will not sag as much under load and that means it won't go up as much under low load. A reasonable DC voltage to expect for this kind of setup is maybe 21.5-22Vdc under normal usage. 

Squinting at the charts on the data sheets, they suggest we can get about 6W into 8 ohms and 14W into 4 ohms for each channel. So it might put out 12W total into 8 ohms, and 28W into 4 ohms, if it's well enough cooled by a heatsink. Let's say you use 4 ohms. You know you will. 

The audio power out MIGHT BE as much as 28W. Each channel will do 14W into 4 ohms. From ohm's law and the definition of power, we get that the RMS current into each load is P = I²*R, or I = SQRT(P/R) = 1.87A. Both channels doing this makes a total current of 3.74A rms going out if I did the math right. What's the average DC current, which is what the DC power supply sees? The peak currents are 1.87A times the same 1.414 factor times two, or 5.29A peak. The average is 0.639 times that or 3.38A.

If that much DC comes out of the filter caps, what is the AC RMS current (which is how transformers are rated) out of the transformer. There's a lot of math I'll skip and just tell you that the RMS in the transformer is about 1.6 to 1.8 times the DC average, depending on a lot of stuff. The transformer secondary would need to be rated at 3.38*1.6 = 5.4Arms.

Uh, that's big. Yes, it is. And it's unrealistic. That stuff was derived assuming the amps were putting out max-power sine waves continuously. Music, even headbanger metal music doesn't do that. Normal pop music has about a 20db peak to average loudness (power) ratio. Heavily compressed music can be as little as 10db peak to average. This is called the "crest factor" of the music.  In power terms, 10db is about 10:1 power ratio. So the average current pulled from the power transformer secondary could be as little as 0.54Arms. You place your bet on how loud and compressed your music will be.

So you got the facts now. You need a transformer rated at 120Vac in, no more than 17Vac out, at somewhere between 0.5A and 5A secondary current.

If it were me, I'd get a 16Vac transformer rated for 1A to 2A, depending on what the deals where where I bought transformers.

[Electron Tornado]

WOW!   Great response!  R.G., thanks for taking the time to go through all of that. It helped immensely in clarifying some other things I've been reading about power supplies and filled in some blanks.

Yes, I am using 4 ohm speakers, because I have a pair and decided to try making an amp with them. The choice of 10 Watts was just a good round number and to keep things simple. The TDA2009A was chosed because I found it quickly and it already had an application circuit for a 10W amp using two 4ohm speakers, so a lot of work was done for me.

I do have a few more questions:

So you got the facts now. You need a transformer rated at 120Vac in, no more than 17Vac out, at somewhere between 0.5A and 5A secondary current.

If it were me, I'd get a 16Vac transformer rated for 1A to 2A, depending on what the deals where where I bought transformers.

With a current demand calculated to be between 0.5A and 5A, how did you decide that a 1A to 2A transformer would be adequate for the job?


Another question - I've read about using capacitors in parallel with the rectifier diodes, although some amp power supplies use them and others don't. How would I determine whether I should use them or not?

[R.G.]

With a current demand calculated to be between 0.5A and 5A, how did you decide that a 1A to 2A transformer would be adequate for the job?

Oh, sure, ask the pertinent, demanding, accurate question of the moment... 
It's a guess.

It's all tied up in that "crest factor" thing and the fact that transformers have a long thermal inertia.

Transformers are hunks of metal with a little insulation inside. They average the heat generated inside them very well indeed. So it may take minutes or hours for a transformer to come up to full temperature based on its average load. And you can burn out an insulated wire inside them in a few seconds if the power/current pulse is very quick and very massive because there are insulation sheets between wires. Both are true. But if the power demands are just the rise and fall of audio being generated, the overloads (if any) are short, and the average of the audio going out is what the transformer heats up with.

So - what's the average load? It's the average output power level, and that's usually 10-20db below the peak power out, that being the definition of crest factor: the peak to average power level in the audio. A 10db crest factor translates to only 10% of peak power being the average power. Relating that back to the 0.5A to 5A thing, if 5A is the necessary power for a sine wave running right at the edge of max power, then music with a 20db crest factor which is hitting the max power out on peaks will have an average power of 1/10 of that, and so the average power implies a current of 0.5A.

So the only good way to pick a transformer rating is either to measure your music output in detail and make you swear you'll never do anything worse; or to say, well, he might go crazy and play full-power sine waves through it for hours for some reason, and choose the 5A version. Or to guess. This is a classic problem in the audio industry, by the way. People who design shaker table amplifiers (giant subwoofer amplifiers which are used in mechanical simulations) or power line testers (which have an output of 120, 240... Vac to test equipment at multi-KW power levels) have to go for full continuous power. So the full-bore, military, NASA, Star Trek rated answer is to use a 5A transformer. And you can do that, and it will always run cool as a cucumber with your music playing.  Or you can pick something smaller that's probably good enough and make the internal bet that you'll never overheat it. Or that you won't overheat it enough to cause damage.

I shortened all that up and guessed at about 1A to 2A as a reasonably safe rating for a not-too-high-priced thing you could actually find.

Another question - I've read about using capacitors in parallel with the rectifier diodes, although some amp power supplies use them and others don't. How would I determine whether I should use them or not?

You should always use them unless you have a spawn-of-the-devil MBA sitting on your shoulders demanding you do the cheapest possible thing, or unless you have bought rectifiers which are both fast and soft recovery special diodes. All common "normal" rectifiers have a slow turn off that ends abruptly and makes the wires around the diodes ring with RF spikes at the rate they turn on and off. This is picked up by the amplifier wiring as a buzzy sounding hum. Snubber caps are not a panacea for this, but they help a lot. The correct panacea is to tune an RC snubber for each diode so it damps the RF ringing. That takes a lot more measurement and math.

[Electron Tornado]

So - what's the average load? It's the average output power level, and that's usually 10-20db below the peak power out, that being the definition of crest factor: the peak to average power level in the audio. A 10db crest factor translates to only 10% of peak power being the average power. Relating that back to the 0.5A to 5A thing, if 5A is the necessary power for a sine wave running right at the edge of max power, then music with a 20db crest factor which is hitting the max power out on peaks will have an average power of 1/10 of that, and so the average power implies a current of 0.5A.

Thanks again. So the transformer current rating used is based on the average output power, which can be thought of as the power level it would see on a continual basis.

I also see some power supplies using caps to ground on the incoming AC lines, but one source also warns against using them unless you know just what you're doing. How useful are these caps, and other than making sure their voltage ratings are high enough, what is critical about using them? Do they essentially do the same thing , filter out high freq noise, that a small value cap in parallel with the main filter cap(s) do?

[amptramp]

There is some information here on capacitors used across the line and in line-to-ground applications:

http://www.justradios.com/safetytips.html

Note that if you have a 3-wire grounded plug, the ground should be carried through to the signal ground unless it is also carried through on the amplifier itself - both would give you a ground loop that injects hum into everything if the chassis of both the effects supply and the amp are grounded separately.  There should be one ground to prevent ground loops but the safety ground should be present on all power supplies in use, including the one in the amplifier.  The amp should be grounded and the power supply for effects should have a safety ground to its chassis but this should not be connected to the outputs or the signal ground - that should be done at the amplifier.  Even the cheapest of the old 5-tube AC/DC radios had a capacitor across the line, so it is necessary.  Note that where a cap is used from each line to ground, the chassis will float at 1/2 of the input voltage unless there is a chassis ground that is carried to the external ground via the line cord.  As well as building your own, these or similar filters are commonly used:

http://www.cor.com/Series/PEM/

Silicon diodes have a natural tendency to carry reverse current until all charges are swept out of the junction and this usually results in a pulse with a fundamental in the 27 MHz to 35 MHz range and all harmonics are present.  As well as providing control of susceptibility to extenal noise sources, line capacitors tend to block emission of noise.  The third harmonic is in the FM band and channel 6 of the television band, so you may be able to test for interference using both of these receivers.  Fast recovery diodes and Schottky diodes are also used to push the frequency up and limit the amplitude, but they are not necessarily sufficient to eliminate noise emission and they do nothing for susceptibility.

It may also be a good idea to have a MOV (metal-oxide varistor) or semiconductor surge AC back-to-back zener across the inputs to protect your circuitry.  Typically, you will use a power bar with a circuit breaker inside and you may want to add a surge protector of either type just to eliminate potential spike voltage damage.

BTW, Dave Cantelon of justradios.com is a fellow member of the London Vintage Radio Club and he seems to be a very good guy (and a good parts source).  It is typical to do a wholesale replacement of capacitors in 70-year old radios and I have done several radios myself.

[Electron Tornado]

Note that if you have a 3-wire grounded plug, the ground should be carried through to the signal ground unless it is also carried through on the amplifier itself - both would give you a ground loop that injects hum into everything if the chassis of both the effects supply and the amp are grounded separately.  

Thanks for the links.

I'm not sure I completely follow your statement above.  This is just a simple solid state amp project.

What do you mean by, "...the ground should be carried through to the signal ground unless it is also carried through on the amplifier itself..."  I plan to use a three prong power cord, grounded to the chassis (that did wonders for my old Alamo Jet).

Apart from using an effects pedal in front of the amp, where does the effects supply reference come from?

[R.G.]

I think he means "make only one connection to the chassis with AC power safety ground, and also only one connection to the chassis with the amplifier signal ground. This will [help] prevent hum from grounding currents."

Best safety practice is to take the AC safety ground where it enters the chassis and connect it to a bolt which uses star washers to cut through paint and oxides to make SURE that it forms a tight, low resistance/high current connection to the chassis. The bolt for tying the safety ground to chassis should not be used for anything else. The signal ground usually can't be tied to multiple places on the chassis without hum or oscillation problems, unless you do a lot of experimenting to get it right. simplest way to connect signal ground to chassis is to connect it one and only one place. The input jack is often used, as is the star ground point of the power supply.

[HOTTUBES]

WoW R.G !
 
Its safe to say you have a pretty good handle on this whole " electronics " thing !!!!   LOL !
Alot of tech's are very " protective " of there knowledge , but you share your info to everyone
without any hesitation !!!!  that's very cool in my books !

[amptramp]

R.G. is correct in that there should only be one ground to chassis carrying the signal and one safety ground.  However, i was referring to another possibility: if the amplifier chassis has single-point signal and chassis grounds, the power supply for the effects should have a safety ground connected to the chassis as R.G. described it, but there should be no connection in the power supply to a signal ground - the power supply outputs should be isolated from the chassis.  If the effects have the signal ground connected to the chassis ground, then there will be circulating currents from the signal ground to the chassis ground in the amplifier, to the chassis ground in the power supply for the effects and to the signal ground in the effects.  That is a loop and any current in it will add 60 Hz hum plus any other line noise to your audio.