BrianGT said:
You're welcome to play off some of the stuff I've done as well (PCB-wise). I haven't posted the latest tweaks and have yet another round in mind, but whatever. I was figuring I'd just do the PCB myself, but if there's any interest in the paralleled version I may opt to order a set of boards to distribute to a few folks to try.
Sometimes it's aggrivating that this project of mine is taking its time. 🙂
C
Variac said:
Please send it to Bill Ceruzzi, the owner of #0 and generous
contributor to DIYAudio.com
😎
planet10 said:
"If you're going to build one chip amp this year, make it this one!"
Actually I was wondering what a good rail voltage will be?
Terry was using ~+/-30V. I was looking at the LM3875 data
sheet and it seems the 4ohm power output drops off at
around +-27V. Given that this is a bridged amp, I wonder if
it'll be beneficial to use lower rails.
This is such a fun and educational thread. I hope to get
a chance to actually build this.
Cheers,
Dennis
I know this may make things a little more complicated but..............
What about using two LM4780's in a paralled/SS configuration. This would eliminate any problems with driving a 4-ohm load, and would put out a little bit more power than using two LM3875's. Any thoughts? I would be extremely interested in PCB's using this configuration.
Cheers,
Zach
What about using two LM4780's in a paralled/SS configuration. This would eliminate any problems with driving a 4-ohm load, and would put out a little bit more power than using two LM3875's. Any thoughts? I would be extremely interested in PCB's using this configuration.
Cheers,
Zach
Well, here's where I am at in the decision making process for the pcb's.
Part of me says "You have a proven design using the LM3875, don't fix what ain't broken." Which, by the way has no problems with 4 ohms, but speakers with 3 or 2 ohm impedance dips will present a problem to this variation at high playback levels.
I was leaning towards the LM3886 for it's superior current capacity, but as BrianGT pointed out to me, it's pinout arrangement is really ugly! Why they need power delivered to so many different pins is absurd.
BrianGT suggested going with paralleled LM1875's, but it really doesn't provide a significant advantage over the LM3875, and adds complexity to the circuit.
I've been looking at using a LM4780, which would allow one chip per channel, but again, there isn't a significant advantage over the LM3875 design which I've already sorted out.
So this leaves me thinking that I should stick with my original LM3875 design for now, and use this as an excuse to indulge my jealousy and punish owners of Wilson Audio or Thiel speakers.
One last tidbit, I also have on the backburner a Turbo version that uses 4 LM3875's per channel to provide current for lower impedance speakers, but it isn't a paralleled design (It's better!). But as I haven't got very far in its development, I don't want to release it as a GP yet.
Please, discuss amongst yourselves.
Part of me says "You have a proven design using the LM3875, don't fix what ain't broken." Which, by the way has no problems with 4 ohms, but speakers with 3 or 2 ohm impedance dips will present a problem to this variation at high playback levels.
I was leaning towards the LM3886 for it's superior current capacity, but as BrianGT pointed out to me, it's pinout arrangement is really ugly! Why they need power delivered to so many different pins is absurd.
BrianGT suggested going with paralleled LM1875's, but it really doesn't provide a significant advantage over the LM3875, and adds complexity to the circuit.
I've been looking at using a LM4780, which would allow one chip per channel, but again, there isn't a significant advantage over the LM3875 design which I've already sorted out.
So this leaves me thinking that I should stick with my original LM3875 design for now, and use this as an excuse to indulge my jealousy and punish owners of Wilson Audio or Thiel speakers.
One last tidbit, I also have on the backburner a Turbo version that uses 4 LM3875's per channel to provide current for lower impedance speakers, but it isn't a paralleled design (It's better!). But as I haven't got very far in its development, I don't want to release it as a GP yet.
Please, discuss amongst yourselves.
If you are creating a PCB then the pinout for the LM4780 does not seem too bad at all really. You could use one LM4780 per channel and get all the benefits of using two LM3886s and get a more compact layout at that. It really should not that terribly more complex than the paralleled 4780 Brian already produces (and they are nice). It seems so much cleaner to use two chips for a stereo amp than to use four.
Russ White said:
This all makes sense to me....circuit-wise it shouldn't be much -- if any -- different.
dave
metalman said:
Not a severe restriction since many amps are in the situation.
Besides, some of us need to use up our stash of LM3875s from
aborted GC projects... 🙂
Hey Terry,
Please don't tease us like that. You're starting to sound like
the Master. 🙂
(Congrats and thanks for the great work you're done on the SusyGC, BTW.)
Dennis
stappvargen said:
I have no idea. I bought it at a shop (Busan, Korea) where music lovers often visit.
Regards
From a DIY perspective, the 4780 looks very tempting. Using the spreadsheet that National provides one sees that the 4780 requires good heatsinking and simple screws may not suffice and clamps are recommended.
I think that using two 3886s may provide a good compromise. Why? the pinout is not as dense as the 4780, thus more tolerant of sloppiness 2. You can get insulated and noninsulated versions 3. You get more surface area to dissipate the heat. This will tolerate non ideal heatsinking and provides that extra margin.
Don't forget that with the GCSS, the bridging arrangement provides for nearly doubling the apparent rail voltage. So operating the GCSS at 20-25 volts provides for apparent high rail voltages.
http://www.national.com/appinfo/audio/files/Overture_Design_Guide14.xls
I think that using two 3886s may provide a good compromise. Why? the pinout is not as dense as the 4780, thus more tolerant of sloppiness 2. You can get insulated and noninsulated versions 3. You get more surface area to dissipate the heat. This will tolerate non ideal heatsinking and provides that extra margin.
Don't forget that with the GCSS, the bridging arrangement provides for nearly doubling the apparent rail voltage. So operating the GCSS at 20-25 volts provides for apparent high rail voltages.
http://www.national.com/appinfo/audio/files/Overture_Design_Guide14.xls
I definitely think either the dual LM3886 or single LM4780 per channel is the most ideal. Having build both the 3875 and 3886 I find the 3886 is able to drive hard loads much better. The 3875 really limps on the 4 ohm loads I have tested. Don't get me wrong, they both sound great. But a lot of us really need that extra current the 3886 can deliver.
Ultimately a think two paralleled 4780s per channel would be an incredible solution able to drive most anything, but some may find it complex.
FWIW I have found that heat disapation for the 4780 is not a problem for most uses, it seems really that only when you start bridging them that heat can be a significant issue. I have 6 channels of parraleled 4780s running on a single 12"x4"x1" 16 fin passive aluminum heatsink with no overheating issues whatsoever. So heat dissapation should not be a major consideration here.
I do a agree that the 3886 pins are much easier to work with, those pins on the 4780 are really close together, but on a PCB I really find it not to be a big deal. Now for P2P on the other hand....
Ultimately a think two paralleled 4780s per channel would be an incredible solution able to drive most anything, but some may find it complex.
FWIW I have found that heat disapation for the 4780 is not a problem for most uses, it seems really that only when you start bridging them that heat can be a significant issue. I have 6 channels of parraleled 4780s running on a single 12"x4"x1" 16 fin passive aluminum heatsink with no overheating issues whatsoever. So heat dissapation should not be a major consideration here.
I do a agree that the 3886 pins are much easier to work with, those pins on the 4780 are really close together, but on a PCB I really find it not to be a big deal. Now for P2P on the other hand....
Richard Bona
Reverence
GCSS/ MSS(paralleled): GCSS has bass guitar sound, which tails fade quickly. I do not know why. It is as if to be due to thin sound waveform near the crossover line. It sounds as if the diminishing amplitude of the sound wave suddenly disappears near the crossover line. By the way, voice quality of GCSS has more charm than MSS(paralleled).
Regards
Attachments
For some reason I missed that thread (and the recent developments). I would definitely like to try that circuit. My latest LM4780 board may come quite useful here, as it allows bridge configuration now (as well stereo and parallel), providing completely separate two channel inputs and outputs. One only needs to add the input circuit, which I would be very tempted to build around 2SK389.
Attachments
I have six of Brian's LM4780 boards each with one LM4780 with its two channels paralleled. I am running two snubberized PSs each with an Avel 500VA 25V dual secondary toroid. So three on each PS.
I have no idea what the thermal resistance is, because I do not know how to measure it. I bought the heatsink at a scrap metal shop.