| amo |
| I have been using the cool little spreadsheet from the National site to help design an amp based on the 3886 chip. I put in 26 volts for the supply voltage and 4 ohms for the load. This yields 62.61 watts of power at 1 % THD, and calculates Vpeak at 22 v and Ipeak at over 5 amps. Is this cutting it a little too close for a 5 amp regulator such as the lm338? Do I have to use the steel package and a massive heatsink? I just hate the idea of the voltage going through the giant heatsink before it gets to the chip, as the steel package has no output pin… Is this totally irrational? I have a feeling that there is a hole in my understanding of how supply voltage relates to the Ipeak/Vpeak. And I am assuming that to get the supply voltage of 26 volts I need (26-3)/1.41 volts from the trafo. |
|
|
| saabie22 |
Amo,
If you use a mica insulator and grease or a silicone rubber insulator with insulating plastic bushings for the screws you will isolate the case from the heatsink.
Here's a Digikey.com link that has transistor mounting hardware:
http://dkc3.digikey.com/PDF/T042/0567.pdf
Mike |
|
|
| carlmart |
| quote: | Originally posted by amo
I have been using the cool little spreadsheet from the National site to help design an amp based on the 3886 chip. I put in 26 volts for the supply voltage and 4 ohms for the load. This yields 62.61 watts of power at 1 % THD, and calculates Vpeak at 22 v and Ipeak at over 5 amps. Is this cutting it a little too close for a 5 amp regulator such as the lm338? Do I have to use the steel package and a massive heatsink? I just hate the idea of the voltage going through the giant heatsink before it gets to the chip, as the steel package has no output pin… Is this totally irrational? I have a feeling that there is a hole in my understanding of how supply voltage relates to the Ipeak/Vpeak. And I am assuming that to get the supply voltage of 26 volts I need (26-3)/1.41 volts from the trafo. |
The hole in your understanding is that you think current demand will be conitnuous. The 3875/3886 chips are class-B amps, so they will demand current in peaks.
Perhaps if the amp owner listens to rock music all the time at high level there might be a problem, but not all the time. The heat might be a problem, but neighbours will probably complain first.
In any case, any conscious DIY owner will test his amp thoroughly and feel how hot the heatsink is. If it's too hot you may need a bigger heatsink or a lower voltage. IMO the chip protection will kick first.
Why do you say the voltage goes through the heatsink? The chip should be isolated from heatsink, to there's only heat being sinked.
The 5amp mentioned are peak, not continuous.
Carlos |
|
|
| johnnyx |
On the steel version (TO3 type), the fixing bolt or nut can be soldered to the PCB, or a wire attached to the bolt with a tag (like a washer, but with a solder or crimp lug attached).
Current doesn't have to pass through the heatsink:) |
|
|
| carlosfm |
| quote: | Originally posted by carlmart
The 5amp mentioned are peak, not continuous.
Carlos |
Yes.
Another detail: the LM338 gives 12A peak current.;)
These chips don't get hot if the difference between input and output (regulated) voltage is around 5V, max. 10V.
And this is independent of the PSU voltage, high or low.
I regulate at a voltage differential of around 5V, when possible, and they run:cool: . |
|
|
| amo |
| quote: | Originally posted by carlosfm
Yes.
Another detail: the LM338 gives 12A peak current.;)
These chips don't get hot if the difference between input and output (regulated) voltage is around 5V, max. 10V.
And this is independent of the PSU voltage, high or low.
I regulate at a voltage differential of around 5V, when possible, and they run:cool: . |
Does this rule out the very cool PLITRON LONO 30 Volt trafos for 4 Ohm loads??? By my calculations I will have to drop close to 16 volts on the regulator heatsink to get the rail voltage of 25 volts (30*1.451-3-25=16). I really want to use these trafos, don't know why, but they are sooo sexy. (If my girlfriend would read this post, she would be laughing her @$$ off, calling me a nerd, and I guess I must be, after all, who else thinks a trafo can be sexy?)
Also, I guess it was wrong for me to say that the voltage will flow "through" the heatsink, because the heatsink it self will be isolated from the case. What I meant to say is that the sink will be "live", but no current will flow through it. Does this somehow degrade performance? I would prefer to stay away from isolating the chip. Thanks for your help as always! |
|
|
| carlosfm |
Regulate for 30~31V, it will work fine for 4 ohm loads.
The more voltage you drop, the more heat you generate and the less current you get out of the LM338.
Oh, and you can use those sexy trafos.:D |
|
|
| johnnyx |
A heatsink has a large area, so if it was connected to a noisy source, then the noise could capacitively couple to sensitive parts of the circuit.
For a regulator, I think it will be be solved by moving sensitive inputs away from the power supply, something you'd do anyway.
If it was at a high voltage there are safety issues, but for an amp the worst that could happen is accidentally shorting something with a test probe. Make a "Caution, live heatsink" label to remind you.
I prefer grounded heatsinks myself, but I have had live ones in some projects. I have seen it in commercial products on many occasions.:) |
|
|
| amo |
| quote: | Originally posted by johnnyx
A heatsink has a large area, so if it was connected to a noisy source, then the noise could capacitively couple to sensitive parts of the circuit. |
I had a feeling that something like this might be going on, but wanted to hear it from someone with more experience. I will have to reconsider my strategy. I was going to chop up a copper bar of 1.5 x 1.5 inches into 1.5 inch sections, creating tiny cubes weighing over a pound each! I was hoping to mount the chips right to these things (after polishing) and isolate them from the case using PVC bar stock. Some vertical holes could also be drilled in the cubes to promote air movement. But I guess I will have to rethink this plan. |
|
|
| amo |
| quote: | Originally posted by carlosfm
Regulate for 30~31V, it will work fine for 4 ohm loads.
The more voltage you drop, the more heat you generate and the less current you get out of the LM338.
Oh, and you can use those sexy trafos.:D |
Ok, I have no idea if what I am about to say makes any sence at all, but here it goes anyway: Would it be possible to put a Zenner diode before the regulator circuit to drop some voltage, to help out the lm338 chip. (Keep in mind I do not know what a Zenner diode really is). I have a feeling that if I put a power resistor before the regulator to drop some voltage it will suck up the current as well. Just shooting in the dark here.... |
|
|
| amo |
| quote: | Originally posted by amo
Ok, I have no idea if what I am about to say makes any sence at all, but here it goes anyway: Would it be possible to put a Zenner diode before the regulator circuit to drop some voltage, to help out the lm338 chip. (Keep in mind I do not know what a Zenner diode really is). I have a feeling that if I put a power resistor before the regulator to drop some voltage it will suck up the current as well. Just shooting in the dark here.... |
Well, I though about this some more, and now I understand why no one bothered to answer.... After all, it is a regulator circuit, and can drop more voltage at expence of more heat..... :smash:
Then I checked out the lm338 data sheet, specifically the output current vs input-output differential voltage graph, and everything fell into place. Thank you all for your help! |
|
|
| metalman |
I don't mean to threadjack, but my problem is similar, so I thought it might fit nicely here. I built, more or less, Pedja's regulator using the LM338's, being fed from 25V secondaries rectified by MUR860 bridges. It is feeding both channels of my non-inverting GC ala BrianGT's boards, and a more advanced version of my own design. Amp build is standard stuff, with 1500uF Panasonic FC's at the LM3875 chips. With the power supply disconnected from tha amp circuits, the regulators work fine giving me exactly the 29V I was shooting for. However, when I connect the power supply to the amp circuits and turn it on, the LM338's cease to work.
By my calc's the 338's should easily handle the demand of the operating amp, but I think the turn on current surge is blowing the regulators. I had hoped that the 338's internal protection mechanisms would allow them to survive this, but I guess not. Anyone else have similar experiences?
My plan at this point is to augment the regulator circuit with the slow turn on option that is detailed in the LM338 data sheet. Has anyone tried this yet?
Cheers, |
|
|
| carlosfm |
The LM338 has absolutely no problems driving two chips.
Metalman, you have a problem there... |
|
|
| metalman |
| Thanks Carlos. It seemed strange to me that the 338's weren't surviving. Are you using a thermistor on your AC primaries? I've seen the schematic of your regulators, and it doesn't look like you are using one. This is one of those times I wished I had a scope to watch what is happening. |
|
|
| carlosfm |
| quote: | Originally posted by metalman
Are you using a thermistor on your AC primaries? |
No.
Remove that and test without it.
Are you sure the LM338s blow up?:eek: |
|
|
| metalman |
Carlos, I haven't been using a thermistor, I was just wondering if you were. I do not know with 100% certainty that the 338's are destroyed, but this is my basis for assuming they are, chronologiclly speaking:
1) After adding the regulator circuit to the power supply, but without the power supply being connected to any load, I switched on the power. The pre-regulator voltages were +/- 34.4 volts, and the post-regulator voltages were +/- 29.3 volts.
2) I switched off the power, and drained the capacitor charge by connecting a 200ohm 25W resistor to between +/- rails of the regulated supply and ground. Then repeated Step 1), checked the voltages again getting the same result, and drained the charge again.
3) Next I connected the power supply to the GC amp, switched on the power and checked the voltages again. Pre-regulator was still +/- 34.4 volts, but post-regulator read +/- 0 volts. Then switched off the PS and drained the charge off the off the un-regulated and regulated +/- rails.
4) Repeated Step 1) again, and the pre-regulator voltages were +/- 34.4 volts, but the post-regulator voltages were still +/- 0 volts.
To me this indicates the 338's are now an open circuit connection. However, if I am missing something, enlightenment is a good thing. |
|
|
| carlosfm |
Have you checked the polarity of the caps?
Maby some mistake there?
Please post here the exact parts you used on your PSU.
What do you have after the regs? |
|
|
| metalman |
Carlos, I checked the polarity of the caps and protection diodes, and they are correctly oriented. Here's a description of what I am using:
500VA Toroidal Transformer with dual 25V secondaries.
Dual full bridge rectifiers consisting of 8 MUR860 diodes, one for positive rail, one for negative rail.
The following is the regulator stage, one per supply rail (i.e. one regulator stage for the positive rail to Gnd, and one stage for Gnd to the negative rail):
2 x 4700uF Electrolytic Capacitors bypassed with a 0.1uF film capacitor, connected between V+in and GND
LM338 regulator with Vin pin connected to V+in, Vout pin connected to V+out, and Adj pin connected to various as described below.
47uF Panasonic FC capacitor connected from Adj pin to Gnd
97ohm resistor connected from Adj pin to Vout pin
22.1Kohm resistor connected from Adj pin to Gnd
1N4004 diode connected from Adj pin to Vout pin (orientation permitting current discharge from 47uF capacitor to Vout pin)
1N4004 diode connected from Vout pin to Vin pin (orientation permitting current discharge from Vout pin to Vin pin)
0.1uF film capacitor connected from Vout pin to Gnd.
The output of the regulators are connected as follows:
The Vout from the positive side regulator is connected to the V+ of the amplifier circuit.
The Gnd from the positive side regulator and the Vout from the negative side regulator are connected to the Gnd of the amplifier circuit.
The Gnd from the negative side regulator is connected to the V- of the amplifier circuit.
For each LM3875 chip there are 1500uF Panasonic FC capacitors connected:
Between the V+ pin and Gnd.
Between the V- pin and Gnd.
(In other words, after the regulation, there is 6000uF total of capacitance).
I modelled the power supply using Duncans Amp Tools, and it indicates that I could see transient current peaks up to 25A within the first 1/2 second after turn-on, but then again I don't completely trust circuit modelling.
Thanks for having a look. I appreciate it! |
|
|
| carlosfm |
22.1k resistor from Adj. to Gnd?:eek:
Are you serious?:bawling:
You mean... 2.2k? |
|
|
| carlosfm |
| quote: | Originally posted by metalman
For each LM3875 chip there are 1500uF Panasonic FC capacitors connected: |
That's too much and when you have this fixed you'll find out that it sounds bad.
Use 100uf on the chip. |
|
|
| metalman |
Carlos, you're right it was a typo, the Adj to Gnd is 2.21k. As to the caps, I hadn't originally intended to run this off a regulated supply, thus the 1500uF caps, but I will be trying different cap values with the regulated supply to see how things change. I have some 220uF amd 100uF on order for other purposes, so I'll give them a try. In the meantime I also have some PNP bipolars coming to try the soft start option and see if that changes the regulator behaviour with what I have right now. Not to mention that ultimately I intend to provide each channel a separate regulator stage. In my experiences you get 99% of the benefit of true dual mono supplies by using only one trafo and rectifier section and then splitting to dual regulators.
After building half a dozen or so discrete regulator circuits, I figured getting a chip regulator up and running would be a cinch. Sometime simpler is more complex I guess. Oh well, no one ever said DIY was for the faint hearted, otherwise everybody would be doing it. |
|
|
| carlosfm |
| quote: | Originally posted by metalman
In my experiences you get 99% of the benefit of true dual mono supplies by using only one trafo and rectifier section and then splitting to dual regulators. |
To get that benefit you need independent rectification too.
Don't forget that you are using positive regs for both rails.
It works either way, but to be truly independent you need separate rectification for each channel. |
|
|
| metalman |
Interesting comment Carlos. I suspect our different views comes from the fact that my previous power supplies have a minimum 25000uF of filter capacitance (that was my headphone amp, my poweramps usually have a lot more) in either a CLC or CRC configuration before I split to the regulators. With that kind of pre-regulator filtration, the rectifiers seem to have much less influence overall. With that setup, my friends and I had great difficulty discerning between true dual mono and the pre-regulator split.
Cheers:D |
|
|
| carlosfm |
No, I mean: in this case (with positive regs for both rails), if you have common rectification for 4 regs (independent per channel), on one side you will end up with two regs in parallel output, for instance, the neg. rail for each channel.
It works fine, but it's not precisely so independent as the intention you have.
What you are saying works with positive/negative regs.
I know it's not easy to understand by words.
Make a quick sketch of how you would connect it and you will get there. :) |
|
|
| amo |
| I am glad this thread has proven useful. Carlos- How do you feel about your circuit being augmented by a pre-regulator or pre-trafo capacitor. Do you think it would be of any benefit? |
|
|
| carlosfm |
Don't use pre-regs here.
I'm almost sure that it will sound worse.
78xx/79xx regs benefit from pre-regulation, but good variable regs don't sound good with regs behing them.
At least that's my experience.
Pre-trafo cap?
You mean a mains filter?
You can try that on the amp. Not sure if you will detect any difference though... |
|
|
| metalman |
OK, I've confirmed a couple of things about the LM338 regulators by using a high speed datalogger to monitor various voltage points during turn-on. They are most definitely not capable of handling the inrush current when supplying 2 GC channels with a total of 6000uF capacitance after the regulators. Even with the slow turn-on circuit implemetation, the transient currents well exceed what the LM338 can handle.
Basically, if you want to try regulating your GC, you'll have to significantly reduce the value of the capacitance at the LM3875 chips if you want to use the LM338 regulators. From my results I'd estimate a max total capacitance after the regulators of approximately 2000uf. More than that and the 338's will self-destruct at turn-on.
With that said, I'm going back to building discrete regulators for my GC. |
|
|
| carlosfm |
Metalman, I told you to use 2x100uf on each LM3875 chip if you use LM338 regs.
Even 1000uf per rail/chip sounds very bad (but it works).
Why would you use 1500uf per rail/chip?:eek: :hot: |
|
|
| metalman |
First, I recognize that everybody has different priorities in what sounds good to them. This even applies to esteemed gentlemen such as John Curl and Nelson Pass, whose abilities have been proven time and time again. So I like to try out the different approaches myself and identify what best matches my personal tastes as to what sounds good. In short, I listen to everyones opinion, pick and choose what seem like good ideas, try them out and see if I like it.
That leads me to my next reason. This is only one small section of a much larger project. One part of this will be doing direct comparisons of non-regulated vs. regulated GC's, and then both of those with different amount of capacitance at the chip, and then finally I have a couple of prototype amplifier designs using LM3875's that I will be comparing against the common GC designs. I want to be able perform my evaluations changing only a single variable at a time.
So, as the powersupply is by far the most costly part of this project, I'm only building only one powersupply in modular form so that it can be used to power all the amplifier variations. One of the power supply requirements will be the ability to drive amps that may have up to 12,000uF of capacitance after the regulators. My original thought was to keep with the spirit of GCing and use IC regulators to shortcut through the effort of designing and building a custom regulator stage. But those nifty LM338's just aren't robust enough to keep up with everything I am going to be throwing at my regulation stage, so I'm going o build something that will.
Please don't get me wrong. I'm not saying that your approach isn't correct, and in all ernest, the LM338's have actually quite impressed me. I just have other plans that will require a different approach. In the end I may find your approach is the best, I just haven't got that far yet.
Cheers, |
|
|
| johnnyx |
| quote: | Originally posted by metalman
One of the power supply requirements will be the ability to drive amps that may have up to 12,000uF of capacitance after the regulators. |
Why? |
|
|
| carlosfm |
| quote: | Originally posted by johnnyx
Why? |
Yes.
Why?:confused:
If the amp produced good sounds with high capacitance I would not bother to regulate.
These regs don't like high capacitance after them.:att'n:
Is there any reg that does?
That capacitance?:eek: |
|
|
| SheldonD |
Carlos:
It would be more correct to say the Amps do not like large caps unless for bass only. (I sure found that out).
Metalman: why the 12000 ufd at the chips after the regs?
Why not before regs? |
|
|
| carlosfm |
| quote: | Originally posted by SheldonD
It would be more correct to say the Amps do not like large caps unless for bass only. (I sure found that out). |
Right.
These amps. |
|
|
| metalman |
You guys aren't reading my posts!
I have built a few different variations of amps to perform comparison testing of the differences. All use LM3875's. Only two variations would likely be considered gainclones. Others are amps that use the LM3875's as only one of two or more gain stages. Some use only one LM3875 per channel, and I have one built so far with 4 LM3875's in a single gain stage in one channel. Depending on how testing and evaluations go, I may go even higher than that. None of these amplifier topologies are described in any National Semiconductor Application Note or Product Data Sheet. One variation uses a topology developed by Nelson Pass, others are completely my own design. Not all of the capacitance I am describing is for the LM3875's. Some of the circuit sections I am trying to develop require extremely quiet power, far quieter that the LM338's can achieve. These need extra filter capacitace after the regulators and then low pass filtering after that. For other circuit sections, the LM338's do not have a sufficiently low output impedance, which can be overcome with capacitive reservoirs after the regulators.
So, to answer your questions:
Sheldon: Why so much after the regs and not before?
1) My supply already has ~40,000uF before the regs.
2) There are always more than one way to get good performance from any given general topology, and I want to play around with some other ideas.
3) Some versions may have several LM3875's, each requiring seperate capacitive reservoirs that when added tegether, even if they are individually small amounts, can present as a single large amount to the regulators.
Carlos: Why?
1) Because I want that much! I don't need any other reason.
2) See some of the reasons above, i.e. I'm not just building standard gainclones!
3) I'm not a sheep, and although I believe you sufficiently to try it your way and see how I like it, I'm not so convinced of your perfection that I'll only do it your way simply because you say so.
4) This whole project for me is an experiment to identify what variables are important to achieving an amp that sounds the way I like it too sound. I already know from reading many of your posts that your preferences are different from mine. I really don't care if the final product of all this work sounds good to anyone else, as long as I like it.
To close out my participation in this thread, I want to point out that I have never once asked for opinions on how much capacitance should be present after the regs, nor have I asked any questions about the amplifier sections of the amplifiers I am building. I only asked one question: Can the LM338's handle the transient turn-on current in-rush with up to 12,000uF after the regs? It is sad that no-one ever did provide any insight into that question, and instead focused on convincing me to change everything else to match with the LM338's operating envelope.
You can respond to this however you like, but I won't be returning to this thread to read it. I found the answer I needed and have better things to move onto than arguing with someone who isn't listening. |
|
|
| carlosfm |
Hey Metalman,
Cool down, we are just trying to help, and that means understanding what you are doing.
In fact, you had a problem and I tried to help.
I'm puzzled with your reaction.:xeye: |
|
|
| moamps |
| quote: | Originally posted by metalman
.....I only asked one question: Can the LM338's handle the transient turn-on current in-rush with up to 12,000uF after the regs?.... [/B] |
Hi,
I built power supply with 338 ( with soft start circuit (1sec) and protection diodes) and ca 20000uF at output (distributed) for one another application. It works without any problem.
Regards
Milan |
|
|
| johnnyx |
You can use the Charge - Balance equation to good effect in power supplies.
CV = q = It
CV = 15000uFx30v, load capacitance to charge up x the
supply voltage
how long for t? say 1ms, fast but not too fast, the time for one cycle of 1kHz sine wave.
so I = 15000ux30v/1ms = 450A
No wonder the poor LM338 got upset.
The regulator has to respond quickly for good transient response, so I would expect t to be much less.
The feedback network within the regulator should track the amp's current demand, otherwise it isn't a regulator, the cap of 100u or so as suggested by Carlos keeps the regulator output impedence low at high frequencies, when the regulator feedback loop runs out of gain, due to the inevitable compensation required for stability. |
|
|
| johnnyx |
| quote: | Originally posted by moamps
[B]
( with soft start circuit (1sec) and protection diodes) |
In that case, t = 1 , so I becomes 450mA, much better.
I still want to know why though.
Does the soft start slow it down all the time, or just at start up?
With a regulator circuit there are two feedback loops in the amp, and both have to be unconditionally stable. Actually three loops, one reg for each rail. Slugging the regulator with loads of capacitance may or may not help the situation. I trust Nat Semi to have a good design without such tinkering. |
|
|
| amo |
| quote: | Originally posted by carlosfm
In fact, you had a problem and I tried to help.
I'm puzzled with your reaction.:xeye: |
Perhaps some members will benefit from Prozac...:smash: |
|
|
| carlosfm |
| quote: | Originally posted by amo
Perhaps some members will benefit from Prozac...:smash: |
Forget it, the man went away.:bawling:
He kept his promise not to come back.:angel:
Go figure.:eek:
It's difficult do give Prozac to some patients.:dodgy: |
|
|
| moamps |
| quote: | Originally posted by johnnyx
Does the soft start slow it down all the time, or just at start up?
|
Just at start up. |
|
|
| d3imlay |
I think it might be a good idea to review the requirements of a voltage regualtor. Most of the regulators I've seen by diy's look only at the output voltage with normal input voltage. However, it's important to consider the worst case current, and input voltage.
A standard practice for worst case input voltage is +/_ 10 percent of nominal. Assuming nominal to be 120 volts then the power supply must regulate the output voltage between 108 and 132 VAC. Rarely is the low end taken into consideration. It appears that the LM338 input voltage needs to be 5 volts greater than the output voltage. Consequently the the raw DC voltage at Low Line (108) has to be 5 volts above the output. At the high end the Vin-Vout will be much higher and the regualtor must have enough heatsinking ability to keep the regulator from failing or shuting down. This is the biggest disadvantge of linear regulators. You end up with a lot heat to get rid of.
The regulator should be monitored with an oscilloscope while the input voltage is changed with a variac while providing full load current. I have seen many regulators become unstable and break into oscillations at low input voltages. |
|
|
| erictoucan |
I think because the capacitor at the output discharge to the input through the regulator and damage the regulator. So just connect a diode across output and input can serve the task without worrying about blow up the regulator.
I think it is hard to say what value of capacitor at the 3875 is safe. Using a diode is the way. You may refer to the application notes of national pdf file for details. |
|
|
|
