Hi bjorn,
...."I'll just try to increase the impedance of the ribbon instead, i.e. making more strips. This would also increase the efficience, which is nice."
Good idea. You could start with a 2.& ohm 10W resistor in series with your current ribbon and try it with an amp you have access to. See how it sounds, the efficiency - I don't know which magnets you have used, but ribbons are notoriously inefficient. No doubt you built it for quality not gross output.
Then you can decide whether you need a serious increase in efficiency, power or both! If you wish to build an amplifier which is designed for conventional 4/8 ohm loads but which can easily be adapted to your special needs check my website.
I have a simple modification that adds extra MOSFETs, if and when needed, and can assist.
Cheers,
Greg
...."I'll just try to increase the impedance of the ribbon instead, i.e. making more strips. This would also increase the efficience, which is nice."
Good idea. You could start with a 2.& ohm 10W resistor in series with your current ribbon and try it with an amp you have access to. See how it sounds, the efficiency - I don't know which magnets you have used, but ribbons are notoriously inefficient. No doubt you built it for quality not gross output.
Then you can decide whether you need a serious increase in efficiency, power or both! If you wish to build an amplifier which is designed for conventional 4/8 ohm loads but which can easily be adapted to your special needs check my website.
I have a simple modification that adds extra MOSFETs, if and when needed, and can assist.
Cheers,
Greg
And here is the simple output stage mod -
http://www.diyaudio.com/forums/showthread.php?postid=743777#post743777
Cheers,
Greg
http://www.diyaudio.com/forums/showthread.php?postid=743777#post743777
Cheers,
Greg
Allowing for the losses, with 20V rails the LM3886 puts out about 17V peak. With 6 LM3886 using .1 ohm paralleling resistors peak current at .4 ohms with no other losses would be ~41 amps. Still a whole lot of current. Max current on the LM3886 is 8 amps. It will be pushing ~360W RMS at this absolute limit occurring at ~1/3 ohm impedance. Minimum recommended gain is 20dB.
All information was derived from the Overture Design Guide spreadsheet.
All information was derived from the Overture Design Guide spreadsheet.
It occurs to me that it would actually be easyer to modify a regular, well regarded and known MOSFET amp for this purpose, than paralleling LM3886s.
Given 20V peak output, a good low voltage power supply would have to be on the order of 25-27V or so unloaded. This is ordinairly quite low for even moderate power (50W into 8 ohms) designs' input stages. It would be a fairly easy affair to generate higher voltage rails for the drivers using voltage multipliers, and even include regulation. 30V or so power rails for the driver circuits would suffice, even for source follower topologies to be driven right to the high power rails. All that remains is choosing the proper high current MOSFETs. At such low voltages, it may even be possible to find reasonably performing 100A complementary pairs, so that only a single pair could be used without source resistors (which would ordinairly be needed for good current sharing) and thus efficiency would increase.
Another approach could capitalize on the fact that ribbons are mostly resistive with relatively little reactance. Because of this, and the low driving voltage, even parallel BJTs could be used. The advantage with BJTs is, that due to much higher Gm, smaller emitter resistors can be used to achieve good current sharing, so the complete contraption may still be reasonably efficient. Because the output voltage is low and the load is not very reactive, BJTs can be found where SOA is not a major issue for the currents and voltages in question. Even so, you would need AT LEAST 3 pairs, mor probably 4 or 5. MOSFETs seem the better option.
LM3886 with their 8A current limit imply at least 7 and preferably 8 chips in parallel. The 0.1 ohm series resistor for current sharing amount to 0.0125 ohm equivalent series resistance, which will dissipate about 15.6W at full power and present a loss of about 0.65V, not too bad - however, MOSFETs in the output can manage lower Rdson at clipping, and therefore, lower losses. The internal losses of the LM3886 will be far higher, about 3-4V off either rail, which will mean that higher rails for required output. The losses in the LM3886s could amount to about half of the output power, which will not be a trivial matter to deal with. MOSFETs win again on account of heatsink and power supply cost, and I would be prepared to bet that the MOSFET discrete amp would be simpler.
Given 20V peak output, a good low voltage power supply would have to be on the order of 25-27V or so unloaded. This is ordinairly quite low for even moderate power (50W into 8 ohms) designs' input stages. It would be a fairly easy affair to generate higher voltage rails for the drivers using voltage multipliers, and even include regulation. 30V or so power rails for the driver circuits would suffice, even for source follower topologies to be driven right to the high power rails. All that remains is choosing the proper high current MOSFETs. At such low voltages, it may even be possible to find reasonably performing 100A complementary pairs, so that only a single pair could be used without source resistors (which would ordinairly be needed for good current sharing) and thus efficiency would increase.
Another approach could capitalize on the fact that ribbons are mostly resistive with relatively little reactance. Because of this, and the low driving voltage, even parallel BJTs could be used. The advantage with BJTs is, that due to much higher Gm, smaller emitter resistors can be used to achieve good current sharing, so the complete contraption may still be reasonably efficient. Because the output voltage is low and the load is not very reactive, BJTs can be found where SOA is not a major issue for the currents and voltages in question. Even so, you would need AT LEAST 3 pairs, mor probably 4 or 5. MOSFETs seem the better option.
LM3886 with their 8A current limit imply at least 7 and preferably 8 chips in parallel. The 0.1 ohm series resistor for current sharing amount to 0.0125 ohm equivalent series resistance, which will dissipate about 15.6W at full power and present a loss of about 0.65V, not too bad - however, MOSFETs in the output can manage lower Rdson at clipping, and therefore, lower losses. The internal losses of the LM3886 will be far higher, about 3-4V off either rail, which will mean that higher rails for required output. The losses in the LM3886s could amount to about half of the output power, which will not be a trivial matter to deal with. MOSFETs win again on account of heatsink and power supply cost, and I would be prepared to bet that the MOSFET discrete amp would be simpler.
Thanks for this input guys. Right now I have a build that use simple ferite magnets, which obviously could be upgraded to neodyn to get more efficient speakers. Also, I use four strips and I guess I could cram in six, i.e. higher efficienty. Dahlberg build one with neodyne and six strips which according to him have an efficience of 92 Db (!). I guess that's enough. The current implementation is roughly one meter long and runs with an audiolab 8000A and sounds really great. I'm not kidding when I say it sound better (in my opinion, I guess I'm kind of biased here though) than a pair of Martin Logan Odyddey hooked up with a really expensive amp. The only thing I miss here is the last bit of resolution, where the martin logans might win but my ribbon kick hard when it comes to dynamics and sound stage.
I am interested in seeing what a better amp might do for them. I have been thinking about the Leqach amp or the Krell KSA clone. Let me ask you amp gurus which one you think is the easiest to build vs. cost and overall quality. I have the impression that both are really good actually.
Also, if someone of you would ever want to build a couple of ribbons it my utmost recommendation. Materials cost is quiet low, less than $300 in my case.
I am interested in seeing what a better amp might do for them. I have been thinking about the Leqach amp or the Krell KSA clone. Let me ask you amp gurus which one you think is the easiest to build vs. cost and overall quality. I have the impression that both are really good actually.
Also, if someone of you would ever want to build a couple of ribbons it my utmost recommendation. Materials cost is quiet low, less than $300 in my case.
I have a pair of Martin Logan so if your efforts are better than ML, it mustn't be "kattpink". I know what you mean though. It's a certain feeling to do things yourself.bjorn.lindberg said:
Exactly what I was thinking. You should look at some power transconductance amplifiers.
Nelson Pass did some experiments in that direction:
(2nd paragraph;2.5Mb! ) http://www.passdiy.com/pdf/cs-amps-speakers.pdf
Building a high power chip amplifier and letting go all that voltage and gain go to waste is a waste of time and money.
And why would anyone want to bridge them when all you need is current?😕
You can of course build a current amplifier with chip amps as an easy start which I believe Nelson P. introduced here as well.
greets
PS.: That could get you started:http://www.diyaudio.com/forums/showthread.php?s=&threadid=36585&highlight=
Yeah I guess you're right PA, I think I can attribute very much to that "I-have-built-them-myself" thing. But still they sound great 🙂. Also the Odyssey didn't sound $10000 good, it was a disappointment from that perspective.
Hello haiqu. I have of course concidered that, but may people claim that these transformers are distorting the sound. Maybey you can elaborate on how to create a good transformer soultion?
/Björn
/Björn
hi bjorn,
It's good to see it's up and sounding good. So what is the impedance - you're driving with the 8000A?
Any Class AB amplifier driving 1ohm loads is likely to have serious PS commutation issues and I'd be 99% certain that's where you're resolution is hiding - behind a miasma of intruding PS generated artefacts that have intruded through the limited PSRR.
Every time you halve the impedance you double the current demand and the size of these 'spray of harmonics' superimposed on the supply impedance. Once you get down to around 1ohm these are serious and you need to look to a design that addresses the issue head on. Simply adding ourput devices is only ensuring output current capability.
The Neydyn and 6 strips 'sounds' good. Impedance?
Hope that helps,
Cheers,
Greg
It's good to see it's up and sounding good. So what is the impedance - you're driving with the 8000A?
Any Class AB amplifier driving 1ohm loads is likely to have serious PS commutation issues and I'd be 99% certain that's where you're resolution is hiding - behind a miasma of intruding PS generated artefacts that have intruded through the limited PSRR.
Every time you halve the impedance you double the current demand and the size of these 'spray of harmonics' superimposed on the supply impedance. Once you get down to around 1ohm these are serious and you need to look to a design that addresses the issue head on. Simply adding ourput devices is only ensuring output current capability.
The Neydyn and 6 strips 'sounds' good. Impedance?
Hope that helps,
Cheers,
Greg
Transformers
Everything is a compromise.
If you haven't tried it yet, give it a go. It's a fast, cheap and simple solution compared to getting an amp to perform when driving low impedances. The result is quite good and you'll have a reference to compare whilst playing with amplifiers if you decide to continue on that path. But tying yourself to one purpose-built amplifier is, to my mind, foot-bullet material.
No special transformer is required, just the right ratio and current rating.
Rob
Everything is a compromise.
If you haven't tried it yet, give it a go. It's a fast, cheap and simple solution compared to getting an amp to perform when driving low impedances. The result is quite good and you'll have a reference to compare whilst playing with amplifiers if you decide to continue on that path. But tying yourself to one purpose-built amplifier is, to my mind, foot-bullet material.
No special transformer is required, just the right ratio and current rating.
Rob
Hi,
since a tranformer would only be driving hi frequency into the ribbon, would an air cored coil be suitable?
Whether ferrite, iron or air the ratio would be sq root(ribbon impedance/other speaker impedance) . Try about 2.5 or 3 or 4 to 1, bifilar wound or would quadfilar wound be better?
At what frequency does air cored start to become ineffective?
If using iron would a gapped core be better than a continuous core?
since a tranformer would only be driving hi frequency into the ribbon, would an air cored coil be suitable?
Whether ferrite, iron or air the ratio would be sq root(ribbon impedance/other speaker impedance) . Try about 2.5 or 3 or 4 to 1, bifilar wound or would quadfilar wound be better?
At what frequency does air cored start to become ineffective?
If using iron would a gapped core be better than a continuous core?
amplifierguru: OK, thanks, trying to increase the impedance sounds like a plan. Right now I use the output stage of the 8000A, which is an OK amp but not ampzilla. I have tried gainclones but the 8000A sounds better. I think this is due to the load and the fact hat my gainclones are not that well designed.
The impedance of the speaker is maybe three to four ohms, since i use four strips. It's hard to meassure with a standard multimeter since it is not really that accurate in that range.
I don't really know about the impedance of the six strips solution but I think I recall a value of six ohms or so, which sound reasonable.
haiqu: Thanks for the tips, but I don't really need something quick and dirty, have that already 🙂. I guess I'll stay with the current solution, trying to refine it.
The impedance of the speaker is maybe three to four ohms, since i use four strips. It's hard to meassure with a standard multimeter since it is not really that accurate in that range.
I don't really know about the impedance of the six strips solution but I think I recall a value of six ohms or so, which sound reasonable.
haiqu: Thanks for the tips, but I don't really need something quick and dirty, have that already 🙂. I guess I'll stay with the current solution, trying to refine it.
check out this thread:
http://www.diyaudio.com/forums/soli...s-paralleled-place-one-output-transistor.html
http://www.diyaudio.com/forums/soli...s-paralleled-place-one-output-transistor.html
- Status
- Not open for further replies.