Hi ebijma,
sorry for the late response, quite busy these days...
Well, you can easily use one positive regulator per supply voltage. One LM338 or LT108x for the positive and one for the negative side. But I'm afraid not for both channels, sorry. Please read on ebjima, I'll explain it.
The schematic is straightforward, though. Only positive regulators, as you know
Hmmm.
Your questions from the first post should have been answered. You can go with the voltages you suggested, or something lower if you don't want to waste too much power and build up unneccessary heat inside the amp enclosure.
How much voltage you have to drop is written in the datasheets of the LM117 series, LM138 series and the LM1083 series regulators. How much current capability your transformer has to have for a desired worst case minimum drop (transformer load regulation at the regulator's input!) is something you have to make sure with your transformer maker, as suggested.
That sounds completely reasonable to me, too.
And remember, the fine thing about having a regulator is the fact that you can set it to any voltage you want. Finding out that load regulation under worst case current is not good enough is just a matter of changing a single resistor per regulator, setting them for a volt less or two. This won't affect sound too much.
Sorry about that. It's the other way round for me, sourcing the LM338 in TO220 would require ordering it separately (with additional shipping). But I wouldn't pay more than 10 EUR for the LT1084...
Transformer output voltage is specified under rated load (remember: talk to your supplier about that point!). Without reasonable current flowing, it's (unloaded) output voltage will always be a little higher, then drop under load. It is this particular property which determines the smalles possible regulation drop.
Unloaded, a 25V per rail power toroid (for example) will supply like 27Vrms per rail or more. Loaded with it's rated current, it shouldn't drop below 25Vrms if it wants to be called a good transformer.
Now there is the unavoidable voltage drop across the rectifier bridge. Let's account between 0.6V and 1.2V per diode (depending on diode type). So the smoothing caps see between 23.8Vrms and 26.4Vrms per rail in our example, depending on the current flowing.
The caps smooth that to the actual peak-to-peak value (1/sqrt(2), factor 1.4142) of something between 33.6Vpp and 37.3Vpp per rail in our example, with a ripple (rms voltage) depending on your smoothing caps (that's why they shouldn't be too small here).
Then comes the regulator. The LM338 has a minimum required voltage differential of 2.5V to 3V (under load) between input and output, depending on it's temperature (diagram 5, page 3 in the datasheet). The LT1084 is rated for the same current (5A), but requires only a drop of less than 1.5V (under load) over temperature (worst case, diagram 4, page 4 in the datasheet). Unloaded and cooled, these figures get even better for both ICs, but especially so for the LM1084 (<1V required drop!).
That's why I suggested either the definition of the required transformer winding specs, or the use of the LT108x series regulators. Actually, I suggested both
But the bottom line is: even under full load, the LM338 will work with the 25V transformer (in case it's a good enough one). Unfortunately, this doesn't account for power line fluctuations, so a couple of volts more don't hurt too much. But 30V secondaries on the transformer is as high as I would go, higher doesn't sound reasonable to me. As Pedja explained, he suggested a higher voltage than actually needed for worst case regulation, but then was comfortable using a lower voltage anyway.
If you want to know what you need to get the amplifiers to play, even the LM317 will work here.
But to make it usable and safe, use one regulator per rail and chip. That's four regulators for two channels, sorry. The peak current requirement of speakers (and thus the amplifiers) at higher volumes requires a higher amount of current than a single regulator could supply. Perhaps the LT1083 with it's 7.5Arms and 10App capability, but I have never heard of someone trying a single regulator for multiple channels. Consult the datasheet of your amplifier IC and compare with the current capability of the regulator ICs.
That's probably the right moment for me to mention my current GC project
Have you had a look here? It's a board with the LM3876/86 and two regulators in an extremely compact form factor. You probably don't like my parts choice, but the schematic has been 'approved' by the same nice forum members who helped you in this thread
The schematics and board for both regulated supply and amplifier are in posts #35 thru #37. I didn't apply important changes to the topology since then. Note that they are still untested in this particular implementation, but the supply schematic is essentially built around Pedja's regulator circuit.
Unfortunately, I'm too busy to finish this development at the moment. I would be glad to present results, but it's too early, sorry.
Ciao,
Sebastian.
sorry for the late response, quite busy these days...
Well, you can easily use one positive regulator per supply voltage. One LM338 or LT108x for the positive and one for the negative side. But I'm afraid not for both channels, sorry. Please read on ebjima, I'll explain it.
The schematic is straightforward, though. Only positive regulators, as you know
Hmmm.
Your questions from the first post should have been answered. You can go with the voltages you suggested, or something lower if you don't want to waste too much power and build up unneccessary heat inside the amp enclosure.
How much voltage you have to drop is written in the datasheets of the LM117 series, LM138 series and the LM1083 series regulators. How much current capability your transformer has to have for a desired worst case minimum drop (transformer load regulation at the regulator's input!) is something you have to make sure with your transformer maker, as suggested.
That sounds completely reasonable to me, too.
And remember, the fine thing about having a regulator is the fact that you can set it to any voltage you want. Finding out that load regulation under worst case current is not good enough is just a matter of changing a single resistor per regulator, setting them for a volt less or two. This won't affect sound too much.
Sorry about that. It's the other way round for me, sourcing the LM338 in TO220 would require ordering it separately (with additional shipping). But I wouldn't pay more than 10 EUR for the LT1084...
Transformer output voltage is specified under rated load (remember: talk to your supplier about that point!). Without reasonable current flowing, it's (unloaded) output voltage will always be a little higher, then drop under load. It is this particular property which determines the smalles possible regulation drop.
Unloaded, a 25V per rail power toroid (for example) will supply like 27Vrms per rail or more. Loaded with it's rated current, it shouldn't drop below 25Vrms if it wants to be called a good transformer
. Now there is the unavoidable voltage drop across the rectifier bridge. Let's account between 0.6V and 1.2V per diode (depending on diode type). So the smoothing caps see between 23.8Vrms and 26.4Vrms per rail in our example, depending on the current flowing.
The caps smooth that to the actual peak-to-peak value (1/sqrt(2), factor 1.4142) of something between 33.6Vpp and 37.3Vpp per rail in our example, with a ripple (rms voltage) depending on your smoothing caps (that's why they shouldn't be too small here).
Then comes the regulator. The LM338 has a minimum required voltage differential of 2.5V to 3V (under load) between input and output, depending on it's temperature (diagram 5, page 3 in the datasheet). The LT1084 is rated for the same current (5A), but requires only a drop of less than 1.5V (under load) over temperature (worst case, diagram 4, page 4 in the datasheet). Unloaded and cooled, these figures get even better for both ICs, but especially so for the LM1084 (<1V required drop!).
That's why I suggested either the definition of the required transformer winding specs, or the use of the LT108x series regulators. Actually, I suggested both
But the bottom line is: even under full load, the LM338 will work with the 25V transformer (in case it's a good enough one). Unfortunately, this doesn't account for power line fluctuations, so a couple of volts more don't hurt too much. But 30V secondaries on the transformer is as high as I would go, higher doesn't sound reasonable to me. As Pedja explained, he suggested a higher voltage than actually needed for worst case regulation, but then was comfortable using a lower voltage anyway.
If you want to know what you need to get the amplifiers to play, even the LM317 will work here.
But to make it usable and safe, use one regulator per rail and chip. That's four regulators for two channels, sorry. The peak current requirement of speakers (and thus the amplifiers) at higher volumes requires a higher amount of current than a single regulator could supply. Perhaps the LT1083 with it's 7.5Arms and 10App capability, but I have never heard of someone trying a single regulator for multiple channels. Consult the datasheet of your amplifier IC and compare with the current capability of the regulator ICs.
That's probably the right moment for me to mention my current GC project
Have you had a look here? It's a board with the LM3876/86 and two regulators in an extremely compact form factor. You probably don't like my parts choice, but the schematic has been 'approved' by the same nice forum members who helped you in this thread
The schematics and board for both regulated supply and amplifier are in posts #35 thru #37. I didn't apply important changes to the topology since then. Note that they are still untested in this particular implementation, but the supply schematic is essentially built around Pedja's regulator circuit.
Unfortunately, I'm too busy to finish this development at the moment. I would be glad to present results, but it's too early, sorry.
Ciao,
Sebastian.
Thanks Sebastian for the input,
You realy know what your talking about.
I thought that everyone was building there regulated gainclone with only one pair of LM338 for both channels.
I can see that it may be a little to much current for the LM338.
So i'll order another two of these reguators.
Thanks for your advice on this.
But i am in doubt about a 5 ampere transformer, is this to small too?
Do i need more ampere to feed both channels?
Erwin
You realy know what your talking about.
I thought that everyone was building there regulated gainclone with only one pair of LM338 for both channels.
I can see that it may be a little to much current for the LM338.
So i'll order another two of these reguators.
Thanks for your advice on this.
But i am in doubt about a 5 ampere transformer, is this to small too?
Do i need more ampere to feed both channels?
Erwin
ebijma said:
Single regs for both channels work fine (better than unregulated IMHO), but of course, it's better to use independent regs for each channel.
And why not independent rectification too?
Independent trafos?
There's a lot of combinations you can make, it depends on your money, time, patience, etc.
It may even depend on the speakers you're gonna use.
It depends on what you really want to do, a small amp or a serious one?
BTW, the power amp I use on my main system has independent regs per channel, but I have other amps with common regulation.
ebijma said:
I thought so, too, until carlos just taught me different.
I really can't recommend using a single pair of regulators. Not only does it limit the amount of power you have available (e.g. the regulator has a current limiting built in and shuts off once it's overloaded for too long), but it's not as safe as it should be, too. Imagine some occasional overload gets your amp IC into it's protection mode, or an accidental short circuit happens.
This could drive a single regulator into shutdown. No problem, says the datasheet, but unfortunately your circuit will be heated up to more than 100 degree already! The Safe Operating Area of e.g. the LT1084 (diagram 6, page 5 of it's datasheet) could be exceeded and that's it; bang!
As carlos stated, it depends on your speakers, your power requirements and how you listen to what
It's, of course, your choice whether you want to try separate regulators, but I would suggest it for safety. Not that I really fear harm to anyone by an overheated LT1084, but a failure could render the whole project useless...Hmm, 27V * 5A = 135VA, that's sufficient for a single channel GC, I'd say. But then, wait until I have thought about that.
Here it appears that you never actually named a particular ampflifier IC. If I knew which you plan on using, I could have a look into the datasheet and find the current requirements for a particular output power. You also didn't specify a power demand, so it's difficult to tell what those 5A should be enough for.
Wait, I read your mind and go for the LM3875/76 class
You thought about regulating to +-30Vdc. According to the maximum output voltage graph (diagram 9, page 8 of the LM3876 datasheet) one can expect an output voltage swing from -26V to +26V. As the THD is too high at 26V, I'll assume an output voltage swing of +-18V for low THD (as derived from a voltage swing ratio of 60% of the supply voltage, 21.15V@8Ohm for 56W@1%THD with +-35V on page 2 of the datasheet).
Hmm, you didn't specify a load impedance. As you are a tupe preamp user, I miraculously guess that you prefer 8Ohm speakers.
That accounts for P = U*U/Z, that's 18V*18V/8Ohm > 40W. So, fourty watts, still plenty. That's I = P / U, about 40W / 18V > 2.2A. This is the estimated average current for an average power of 40W into 8Ohm. This current, of course, has to flow through the whole circuit; through the regulator, the amplifier and the speaker.
With your given 27V secondaries, I calculate a rail voltage of +-37.5V. The regulator drops this to your estimated 30V. That's a power requirement of 7.5V * 2.2A = 16.5W for the regulator to withstand (not counting losses inside the regulator). No problem, but it needs good cooling!
Now we have the resulting 30V split between the amplifier (12V) and the speaker (18V). I leave it for you to calculate the average power dissipation of your chip amp, regarding cooling and such.
What's important for your transformer is the fact that 16.5W per regulator are on top of the 40W per amplifier channel. That's 80W for two amplifier channels and 66W for the regulators, thus, 146W as of this calculation!
But don't forget that my calculation is a conservative approach. You won't dissipate 146W for two channels and regulators by listening to music. Your average current requirement will be much lower, while your peak maximum current requirement may be much higher. Your transformer will have no problem to supply this current for short durations. It's output voltage may drop a little below the specified 27V, but that's what the regulator is for, isn't it?
One thing shouldn't be forgotten: I derived an output swing of +-24V from the datasheet. That would mean that the absolute peak power output into an 8 Ohm load would be >80W. Consider what this means to your power dissipation (heat) and transformer current when someone accidentally overdrives your amp and can't get to the power switch in a couple of seconds... the internal limiting would limit dissipation, but your 5A trafo is not suited for abuse, that's for shure.
OTOH, an audiophile would answer: You'd have to tenfold your current supply to be shure, that's 22A in your case dude!
carlosfm said:
Now that's interesting. Did you ever recognize the single regulators to get too hot or influence the sound? I guess not, because you would have changed it, then.
What kind of amplification would you recommend two GCs with a single regulated PSU for (voltage, average power, duration of continuous use). Perhaps what you're doing with it could be sufficient for Erwin.
Ciao,
Sebastian.
PS: Following the above calculation, combining two channels with a single regulator would let the regulator dissipate >30W. Considering the maximum power dissipation of the LM1084 (diagram 9, page 5), this would limit it's safe use to <80 centidegree. Too weak to be shure.
sek said:Now that's interesting. Did you ever recognize the single regulators to get too hot or influence the sound? I guess not, because you would have changed it, then.
As you may have noticed, where I need the power (on my main system), I use independent regs, and they are on a big heatsink.
It's a power amp with LM3886s.
I have other smaller amps with lower PSU voltage and they require much less current from the PSU.
Like an LM4780 amp with a 2x18V 180VA toroid (this is 5A), regulated with LM338s at 20V.
The LM4780 gives around 20 watts/8 ohms at this PSU voltage, and this is more than enough for what I need it for.
Oh, and the LM338s on this amp don't even get warm, and they are on a Pentium heatsink.
Like I said, it depends on what amp you want to do.
Of course I recommend the best, I just said it works fine either way.
The LM3886 power amp above had initially common regs for both channels. I just changed it because... because I wanted to.
Nah, because I got a big heatsink, just ideal for the 4 regs and made a new board with rectification included and 10,000uf caps.
Nice.
sek said:
I have lots of combinations.
Another one will be born soon, with tube buffer.
Allright
BTW Erwin, what was that 5A figure referring to? Did you mean 5A for each of the two secondaries, as in a 270VA trafo? I understood 5A total...
Of course such a donut would easily be sufficient for two channels and four regulators, even for permanent party use
Or you connect four channels and be unsure again...
No, honestly, a 300VA toroid is what I use for a two-channel setup.
Actually oversized! So you could increase your supply voltage and compare the sound (changing only the voltage divider shunt resistor at the regulators).
If we're talking about a 270VA trafo, that is!
Good luck, Erwin.
Sebastian.
BTW Erwin, what was that 5A figure referring to? Did you mean 5A for each of the two secondaries, as in a 270VA trafo? I understood 5A total...
Of course such a donut would easily be sufficient for two channels and four regulators, even for permanent party use
Or you connect four channels and be unsure again...
No, honestly, a 300VA toroid is what I use for a two-channel setup.
Actually oversized! So you could increase your supply voltage and compare the sound (changing only the voltage divider shunt resistor at the regulators).
If we're talking about a 270VA trafo, that is!
Good luck, Erwin.
Sebastian.
carlosfm said:
Hello CarlosFM,
Thanks for your input on this matter.
I am building a serious amp here (riken resistors, blackgate, custom transformer)
And im goning to use 8 ohms speakers.
sek said:
Hello Sek,
Thanks for your input also (again),
Yes i ment a 270 va trafo.
With windings for the valve buffer to.
So the whole thing feeds on one transformer.
I using a E88CC for the valve buffer and two LM3875TF for the clone in a non inverting setting and ofcourse i want it all to be regulated.
Thats go's for the valve buffer to.
I am going to regulated the valve buffer with for positive rail a LM317 and for the negative rail a LM337.
For the filament a single LM317.
Connecting 8 ohms speakers.
carlosfm said:Good.
Then you don't need to ask, independent regulation and rectification.
True, now i don't have to because Sek told me.
carlosfm said:Why not the LM3886TF?
Even more serious.
Why is 12 watts more power more serious?
LM3875 has enough power for me.
And i doubt that the LM3886 will sound better at normal listing levels, with not to difficult speakers.
So................
ebijma said:Why is 12 watts more power more serious?
LM3875 has enough power for me.
And i doubt that the LM3886 will sound better at normal listing levels, with not to difficult speakers.
So................
It's not a question of 12 watts more power.
It's a question of current output.
Around double the current of the LM3875.
What speakers do you have?
carlosfm said:
Yep, because the current capability determines the dynamics of your amplifier-speaker system 'before' clipping.
Erwin, even if you don't consume higher power levels, having more dynamic headroom available and keeping the SPiKe-Protection out of the way can greatly help the sound.
Regarding the difference in sound between the different LM38xx-ICs, you have to judge by yourself (just do a pin-by-pin setup to compare), because it's a matter of taste (it's not even for shure that they'll sound different at lower or medium power levels). After all, it shall become a 'serious amp', right?
Ciao,
Sebastian.
Both chips should sound the same if well implemented. Except if you have current hungry speakers or listen to music at high levels.
If your speaker woofer is larger than 8" it may be wiser using a 3886 or a parallel set of them.
One thing you can do, if possible, is going to the store with a DMM and measuring the DC resistance at the speaker's terminal posts. Easy to do if the speaker is bi-wire ready. This measurement will not tell you the impedance but it's usually a tell-tale of what it may be.
Or check if there are any tests around for the speakers.
Having some power available, even if you don't use it, means that the amp will be working at its most linear response. Particularly not allowing any clippling that sounds nasty on these chips.
Carlos
If your speaker woofer is larger than 8" it may be wiser using a 3886 or a parallel set of them.
One thing you can do, if possible, is going to the store with a DMM and measuring the DC resistance at the speaker's terminal posts. Easy to do if the speaker is bi-wire ready. This measurement will not tell you the impedance but it's usually a tell-tale of what it may be.
Or check if there are any tests around for the speakers.
Having some power available, even if you don't use it, means that the amp will be working at its most linear response. Particularly not allowing any clippling that sounds nasty on these chips.
Carlos
carlmart said:
I say that any modern bass-reflex loaded speaker benefits the extra current of the LM3886.
Independently of the "average" impedance and sensitivity that the manufacturers claim.
I'm only excluding very sensitive speakers (92~93db and up), IF they don't have big impedance dips, specially in the lower frequencies.
Btw, when an amp is not driving a pair of speakers properly it shows even at low level.
It may be subtle, and you only figure it out when you upgrade... the amp.
All, test this with music with strong bass content and dynamics, not choral music.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.