LM-3875 turn on/off noise
Hello sek,
could You please elaborate a little more detailed about power-up/down problems of the LM3875`s?
Do they make a big noise (maybe a loud low frequency bump) while turning on/off?
As I plan to LM-3875 power a tweeter in an active system, I`m a bit worried about this. Not only that I hate such noise when powering up a system, I fear it could actually damage a tweeter when directly coupled to the amp output.
Hello sek,
could You please elaborate a little more detailed about power-up/down problems of the LM3875`s?
Do they make a big noise (maybe a loud low frequency bump) while turning on/off?
As I plan to LM-3875 power a tweeter in an active system, I`m a bit worried about this. Not only that I hate such noise when powering up a system, I fear it could actually damage a tweeter when directly coupled to the amp output.
Hi Cocolino,
I've done this in this thread - in theory. I've studied datasheets, read application notes and asked people on the forums.
The difference between LM3875 and LM3876 is explained in some of those documents.
From both the LM3875 and the LM3876 datasheet:
From National's application note AN-898 "Audio Amplifiers Utilizing SPiKe Protection":
So they are different in that regard.
I haven't tested the LM3875, so I just can't help you with that decision. But it is said that the common parts in the internal circuit of LM3875 and LM3876 are identical, so I decided for the '76. Besides that, it's pin compatible to the LM3886, so the more powerful one would also fit in the board.
I want to use the boards for tweeters, too, so I want to be somewhat sure. And guess what, my parts dealer sells the '76 even some cents cheaper:
Sebastian.
PS: Last weekend I set up a prototype with an OPA541. This IC has no such protection circuitry, and the effect is quite awfull: Spike on turn on, distorted fade out on turn off (depending on the power supply regulaton, e.g. capacitor size).
I've done this in this thread - in theory. I've studied datasheets, read application notes and asked people on the forums.
The difference between LM3875 and LM3876 is explained in some of those documents.
From both the LM3875 and the LM3876 datasheet:
From National's application note AN-898 "Audio Amplifiers Utilizing SPiKe Protection":
So they are different in that regard.
I haven't tested the LM3875, so I just can't help you with that decision. But it is said that the common parts in the internal circuit of LM3875 and LM3876 are identical, so I decided for the '76. Besides that, it's pin compatible to the LM3886, so the more powerful one would also fit in the board.
I want to use the boards for tweeters, too, so I want to be somewhat sure. And guess what, my parts dealer sells the '76 even some cents cheaper:
Sebastian.
PS: Last weekend I set up a prototype with an OPA541. This IC has no such protection circuitry, and the effect is quite awfull: Spike on turn on, distorted fade out on turn off (depending on the power supply regulaton, e.g. capacitor size).
Hi tbla,
Yes, only a couple of days ago I've read an application note about controlling the feedback loop's phase margin via the output capacitor's ESR. I just seem to loose PDFs after I read them, I can't find this one any more!
Okay, I'll replace the 1uF caps near the elkos with bleeder resistors. That's an increase in stability and safety.
20k should be fine here, drawing between 1.25mA and 1.5mA at 25V to 30V. That's below 0.05W, so a standard resistor will do fine. I can even imagine dedicated bleeder resistors for the purpose of draining the caps so that the undervoltage-protection circuit switches off quickly, but I don't want to waste power (component size, heat and cost).
Thank you for all the insight, that's clearly knowledge and experience speaking here.
Yes, only a couple of days ago I've read an application note about controlling the feedback loop's phase margin via the output capacitor's ESR. I just seem to loose PDFs after I read them, I can't find this one any more!
Okay, I'll replace the 1uF caps near the elkos with bleeder resistors. That's an increase in stability and safety.
20k should be fine here, drawing between 1.25mA and 1.5mA at 25V to 30V. That's below 0.05W, so a standard resistor will do fine. I can even imagine dedicated bleeder resistors for the purpose of draining the caps so that the undervoltage-protection circuit switches off quickly, but I don't want to waste power (component size, heat and cost).
Thank you for all the insight, that's clearly knowledge and experience speaking here.
Good, this is version 0.3 of the schematic. An attempt to the power supply bridge rectifier and reservoir capacitors.
It features two different types of diodes in the rectifier bridge: single and dual types, a.k.a. the MUR860 or the MBR20200 style in TO220 packages. Both "soft recovery" and Schottky types will work. I'll try to arrange them so that axial diodes will fit too, but I can't promise...
The bridges are bypassed with 100nF ceramics. Film is not neccessary here, right?
I've drawn in three reservoir capacitors per rail. This is no final decision, just an approach on how to handle board space and parts arrangement (it would be two larger or four smaller caps, then). Each big cap is bypassed with a small ceramic, too. You could argue about capacitance here. Then there are separate bleeder resistors for each cap, just to distribute power.
I've thought about putting in an (optional) smd LED anywhere between the bridge and the regulators. Should I?
Here You are again:
What did I do wrong? Or what could be done better?
Thanks,
Sebastian.
It features two different types of diodes in the rectifier bridge: single and dual types, a.k.a. the MUR860 or the MBR20200 style in TO220 packages. Both "soft recovery" and Schottky types will work. I'll try to arrange them so that axial diodes will fit too, but I can't promise...
The bridges are bypassed with 100nF ceramics. Film is not neccessary here, right?
I've drawn in three reservoir capacitors per rail. This is no final decision, just an approach on how to handle board space and parts arrangement (it would be two larger or four smaller caps, then). Each big cap is bypassed with a small ceramic, too. You could argue about capacitance here. Then there are separate bleeder resistors for each cap, just to distribute power.
I've thought about putting in an (optional) smd LED anywhere between the bridge and the regulators. Should I?
Here You are again:
An externally hosted image should be here but it was not working when we last tested it.
What did I do wrong? Or what could be done better?
Thanks,
Sebastian.
Talking about regulator stability? Yes, You're probably right. But besides that the circuit is intended for both LM3876/86, those resitors have a different function.
R9 and R10 are in place to drain C3 and C4 after power-off. The internal undervoltage-protection circuit "mutes" the output at below +-9V supply voltage, probably reducing current consumption.
Now, when power is returned very quickly, the chip turns on the output again while surrounding circuitry might not be stable already...
You might be right that R9 and R10 aren't needed, as page 1 of National's AN-898 doesn't indicate that in the internal circuit (including protection blocks). The circuit seems to simply mute the input stage, but doesn't turn off the power supply of the bias and output stages.
According to the datasheet, the LM3876 draws at least 30mA muted and with no signal applied. But how much does it draw below it's +-9V threshold? Does anyone have experiences with periodically switching the power of a regulated GC?
I'm just a little concerned, because when I asked about turn-on noise in another thread, everybody reported some pops sometimes...
Sebastian.
I experience that. In one instance, I put off the amp, and unknowingly switched it on quite immediately, and I heard a pop and a little smoke in my regulator. Good, my speaker is not connected during this time.
Upon visual inspection and retesting, the regulator was not damaged.
Thus, I thought that time, that the amp needs to have quite a time to discharge cap charge before turning it on again.
Then I noted STK recommended power supply, putting 500R each in the power rail caps for gracious discharging away from the amp package.
But in your case, r9 and r10 may not be needed at all. You see, you already provided the 2 protection diodes (D1 & D2, and D3 & D4), which directs discharging of C3 & C7 to those resistors in the power caps away from the regulator and the chip amps.
Further, when it starts discharging, the mute resistance will not have enough voltage to switch-on the amp out, so your output will just be muted (I think which is why you have to at least configure the Rm as close to the rail voltage, like in national datasheet, if you use +-16V, use 16K ohm, so that it can immediately mute the OUT when voltage drops during discharge).
I am still looking for some other explanation just to see if I have overlooked something else.
Upon visual inspection and retesting, the regulator was not damaged.
Thus, I thought that time, that the amp needs to have quite a time to discharge cap charge before turning it on again.
Then I noted STK recommended power supply, putting 500R each in the power rail caps for gracious discharging away from the amp package.
But in your case, r9 and r10 may not be needed at all. You see, you already provided the 2 protection diodes (D1 & D2, and D3 & D4), which directs discharging of C3 & C7 to those resistors in the power caps away from the regulator and the chip amps.
Further, when it starts discharging, the mute resistance will not have enough voltage to switch-on the amp out, so your output will just be muted (I think which is why you have to at least configure the Rm as close to the rail voltage, like in national datasheet, if you use +-16V, use 16K ohm, so that it can immediately mute the OUT when voltage drops during discharge).
I am still looking for some other explanation just to see if I have overlooked something else.
aHobbit,
thanks for sharing this experience. It's very much appreciated, as I now know that my assumption and suspecting wasn't without reason.
As for the parts values in the schematic: the bleeder resistors, the regulator voltage dividers, the feedback network calculation and the input impedance setting are not final values. Those are first attempts from recommendations (mostly datasheets and the mentioned sources) and will be verified and fine tuned in a first prototype "series", where I want to mutually compare two circuits with different combinations and settings.
In particular, the bleeder values seem too high, the time constant t = R*C of 20kOhm and 1mF is too long, I guess (it would take some time to discharge deeply, not enough safety when fiddling with the circuit).
The regulator output voltage is currently set to ca. 25V. This has to be adjusted for the desired value and/or low loss.
The gain and impedance settings are just employments of parts I had available. They are not optimized for low offset etc.
Sebastian.
thanks for sharing this experience. It's very much appreciated, as I now know that my assumption and suspecting wasn't without reason.
As for the parts values in the schematic: the bleeder resistors, the regulator voltage dividers, the feedback network calculation and the input impedance setting are not final values. Those are first attempts from recommendations (mostly datasheets and the mentioned sources) and will be verified and fine tuned in a first prototype "series", where I want to mutually compare two circuits with different combinations and settings.
In particular, the bleeder values seem too high, the time constant t = R*C of 20kOhm and 1mF is too long, I guess (it would take some time to discharge deeply, not enough safety when fiddling with the circuit).
The regulator output voltage is currently set to ca. 25V. This has to be adjusted for the desired value and/or low loss.
The gain and impedance settings are just employments of parts I had available. They are not optimized for low offset etc.
Sebastian.
Okay, version 0.4 so far, please collect complaints for the next couple of days.
I'll be on vacation for a week or two.
Good news is: due to the reduction of regulator input capacitor size, I gained back the needed board space for both an RL Zobel and an RC snubber network (while both are optional)!
When I'm back, I'll post the board layout, the first version is nearly finnished.
To make this project interresting enough for you to stay in mind while I'm awayD),
here are some features that the final version will very likely have:
- One board per rectifier bridge with capacitors, one board per amplifier with regulators,
- board sizes are 75mm x 30mm (ca. 3" by 1.18"),
- there will be an additional board for more reservoir capacitance, which may also be used in conjunction with a separate bridge rectifier in order to save the rectifier board,
- boards can be stacked to form a module of N amplifier channels with 75mm x 30mm x N*40mm (ca. 3" by 1.18" by N*1.57") volume, e.g. two channels and a rectifier in 75x30x120mm (or a little less in height, depending on max. component height, mainly capacitors),
- the module is connected via bolts/screws and only needs a transformer (approx. 2x25V, >3A per wanted amplifer "channel"), an audio source connection (one per channel), a speaker connection (one per channel) and a heatsink (common to the whole module, big enough to mount whatever module size (channel count and capacitance options) you want, low thermal resistance according to LM3876/86 and regulator needs)
- very variable in component count (see schematic for optional components)
- not so variable in exotic component choice (tight layout due to size constraints),
- not a beginners project (due to two sided mixed thru-hole and SMD component mounting), but everybody who knows the own soldering station well enough should be able to build it.
I would say that this actually fits the specification that I've had in the beginning, a compact design that fits in almost any back side compartment of an active speaker box system.
Thanks to everybody who helped so far, that thing is going on,
Sebastian.
I'll be on vacation for a week or two.
Good news is: due to the reduction of regulator input capacitor size, I gained back the needed board space for both an RL Zobel and an RC snubber network (while both are optional)!
When I'm back, I'll post the board layout, the first version is nearly finnished.
To make this project interresting enough for you to stay in mind while I'm away
D),here are some features that the final version will very likely have:
- One board per rectifier bridge with capacitors, one board per amplifier with regulators,
- board sizes are 75mm x 30mm (ca. 3" by 1.18"),
- there will be an additional board for more reservoir capacitance, which may also be used in conjunction with a separate bridge rectifier in order to save the rectifier board,
- boards can be stacked to form a module of N amplifier channels with 75mm x 30mm x N*40mm (ca. 3" by 1.18" by N*1.57") volume, e.g. two channels and a rectifier in 75x30x120mm (or a little less in height, depending on max. component height, mainly capacitors),
- the module is connected via bolts/screws and only needs a transformer (approx. 2x25V, >3A per wanted amplifer "channel"), an audio source connection (one per channel), a speaker connection (one per channel) and a heatsink (common to the whole module, big enough to mount whatever module size (channel count and capacitance options) you want, low thermal resistance according to LM3876/86 and regulator needs)
- very variable in component count (see schematic for optional components)
- not so variable in exotic component choice (tight layout due to size constraints),
- not a beginners project (due to two sided mixed thru-hole and SMD component mounting), but everybody who knows the own soldering station well enough should be able to build it.
An externally hosted image should be here but it was not working when we last tested it.
I would say that this actually fits the specification that I've had in the beginning, a compact design that fits in almost any back side compartment of an active speaker box system.
Thanks to everybody who helped so far, that thing is going on,
Sebastian.
That's for power handling. With one resistor per capacitor, I can use standard 1/4W resistors. As I already wrote, I probably want to decrease the resistor value for a shorter time constant. A single bleeder resistor would then have to handle more power and thus I had to use smd power resistors, which are more expensive and more difficult to find.
The resistor really mounts under the cap, and eaven if you leave it out, you don't waste unpopulated board space.
Thanks for the hint. Other things to mention?
Cheers,
Sebastian.
But now I'm away...
The resistor really mounts under the cap, and eaven if you leave it out, you don't waste unpopulated board space.
Thanks for the hint. Other things to mention?
Cheers,
Sebastian.
But now I'm away...
And this is version 0.5 of the amp's schematic.
Some minor changes here, too.
C12* is now connected to power ground instead of signal ground, as pin 4 sources current into the supply, not the signal source (and it is therefore easier to route on the board).
The output filter network now consists of multiple standard resistors. This is smaller, easier to get and looks better (and makes the layout smaller and more convenient). Thanks to Per-Anders and Pedja for the inspisration.
Sebastian.
Some minor changes here, too.
C12* is now connected to power ground instead of signal ground, as pin 4 sources current into the supply, not the signal source (and it is therefore easier to route on the board).
The output filter network now consists of multiple standard resistors. This is smaller, easier to get and looks better (and makes the layout smaller and more convenient). Thanks to Per-Anders and Pedja for the inspisration.
Sebastian.
Attachments
First board version
And now something new
The first version of the board, which I will call 0.5 too.
As it's typical for Eagle, red shows the upper and blue the lower side, while 'lila' just indicates the presence of copper on both sides.
To make it usefully readable, I have included the component placement. It makes 'reading' the picture a little difficult at first, but helps a lot compared to four views of two layouted sides and two matching component placement graphs.
Some components overlap, those are the alternatives in the schematic.
Some component names are mirrored, this indicates that the component is placed on the other board side.
The board is cut in half horizontally. The upper half is the amp circuit, the lower half is it's rectifier and reservoir capacitor board.
On the amp board one can see the regulators (with all their adjacent circuitry) in the upper left and right corners.
In between them sits the amp chip with input filter, feedback and output filter circuitry.
On the rectifier board we have sixteen diodes (!) as two alternative sets of eight pieces. The large capacitors can be combined by mounting the bigger under the board, while putting the smaller one as usual.
Sizes: length 75mm (3"), width 30mm (1.18"), mounting heigth determined by capacitors and ICs (max. 5mm SMD height on the bottom side).
It's not optimized, but fulfills the usual design rules of board manufacturers, has no (!) vias and only needs two-sided soldering where it's possible (e.g. on the large caps and the ICs).
What does everybody think of it?
Sebastian.
And now something new
The first version of the board, which I will call 0.5 too.
As it's typical for Eagle, red shows the upper and blue the lower side, while 'lila' just indicates the presence of copper on both sides.
To make it usefully readable, I have included the component placement. It makes 'reading' the picture a little difficult at first, but helps a lot compared to four views of two layouted sides and two matching component placement graphs.
Some components overlap, those are the alternatives in the schematic.
Some component names are mirrored, this indicates that the component is placed on the other board side.
The board is cut in half horizontally. The upper half is the amp circuit, the lower half is it's rectifier and reservoir capacitor board.
On the amp board one can see the regulators (with all their adjacent circuitry) in the upper left and right corners.
In between them sits the amp chip with input filter, feedback and output filter circuitry.
On the rectifier board we have sixteen diodes (!) as two alternative sets of eight pieces. The large capacitors can be combined by mounting the bigger under the board, while putting the smaller one as usual.
Sizes: length 75mm (3"), width 30mm (1.18"), mounting heigth determined by capacitors and ICs (max. 5mm SMD height on the bottom side).
It's not optimized, but fulfills the usual design rules of board manufacturers, has no (!) vias and only needs two-sided soldering where it's possible (e.g. on the large caps and the ICs).
What does everybody think of it?
Sebastian.
Attachments
Hi,
I don't really understand the question, sorry!
The ADJ bypass capacitors are solid tantalums right in front of the regulator ICs.
The output capacitors are the smaller thru-hole electrolytics between regulators and amp.
The bigger electrolytics at the lower edge of the amp boards are the regulator's input capacitors C111 and C115.
Do you mean I should try different types accordingly? I will certainly do that, but as time and budget permits, the first step will be to get the amp to work at all
I decided for tantalums as ADJ bypass capacitors because they are small and inexpensive. The main goal for regulating the LM3876/86's power supply is not ripple rejection, but line regulation. As the PSRR of both the regulator (min. 60dB) and the amp (mind. 85dB) are way good enough, I paid more attention to proper bypassing and left more room for different regulators as well as different input and output caps.
Shure, it's the typical wire-wound 0.7uH coil, just mounted vertically.
Does the grounding look allright so far?
Or does anyone need a better view on the board? Which?
Sebastian.
I don't really understand the question, sorry!
The ADJ bypass capacitors are solid tantalums right in front of the regulator ICs.
The output capacitors are the smaller thru-hole electrolytics between regulators and amp.
The bigger electrolytics at the lower edge of the amp boards are the regulator's input capacitors C111 and C115.
Do you mean I should try different types accordingly? I will certainly do that, but as time and budget permits, the first step will be to get the amp to work at all
I decided for tantalums as ADJ bypass capacitors because they are small and inexpensive. The main goal for regulating the LM3876/86's power supply is not ripple rejection, but line regulation. As the PSRR of both the regulator (min. 60dB) and the amp (mind. 85dB) are way good enough, I paid more attention to proper bypassing and left more room for different regulators as well as different input and output caps.
Shure, it's the typical wire-wound 0.7uH coil, just mounted vertically.
Does the grounding look allright so far?
Or does anyone need a better view on the board? Which?
Sebastian.
sek said:I don't really understand the question, sorry!
The ADJ bypass capacitors are solid tantalums right in front of the regulator ICs.
The output capacitors are the smaller thru-hole electrolytics between regulators and amp.
The bigger electrolytics at the lower edge of the amp boards are the regulator's input capacitors C111 and C115.
Do you mean I should try different types accordingly? I will certainly do that, but as time and budget permits, the first step will be to get the amp to work at all
Sorry my comment wasn't clear enough. I think you should try using the 470uF to bypass the adj pin and the smaller cap at the output.
My experience and other's on this forum may show that using small capacitors at the regulator's output pays off in better sound. Perhaps not as small (10uF), but something around 47/100uF.
What I am suggesting will not affect your budget, as it only involves swapping the capacitors you already have and listening to the results.
Only use those tantalums if you have already tried them in another project. I had many tantalums working bad or even catching fire on some of my projects, even on a digital Turtle Beach I had. Don't trust them too much, particularly if those tantalums are inexpensive. A good Panasonic FC should provide similar results and costs cents, with zero risk.
Carlos
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.