That is a fantastic post Thorsten, thanks!
I must admit that going fully active is something I did seriously think about, and also one of the reasons why i wanted to keep power and pre sections separate (the other reason being the ease of upgrading preamp in future). At this point though, I lack necessary knowledge to even begin to design such setup. I understand that and active crossover completely replaces the passive one, which means I would have a very happy multitude of things to get right with it, all at the same time: relative speaker sensitivity, phase shift, time alignment, and of course frequency control. As much as I would love to give it a stab, i regret to say that it is almost certainly beyond my ability at the moment...
I take your point about separate PSU not necessarily being an improvement. My line of thought was that an additional 30cm or so of cable is a reasonable trade-off for better shielding (and that is how the original Gaincard was laid out too), but you made me reconsider this. In fact, I've just come up with an idea how to incorporate the PSU section into my original design, while maintaining perfect symmetry AND allowing some heavy shielding to be employed around PSU... Damn, I'm good
Is transformer the only "noisy" part of the PSU, or do diode bridges radiate something undesirable as well?
I still have mixed feelings about bridging/paralleling multiple chips though. I suppose I really need to hear it myself, in order to be able to judge whether I prefer it or not. As I said earlier I would not sacrifice "directness" of sound for anything else.
One thing puzzles me though - why do you advise against silver? Is it not simply superior electrically?
Also, no one answered my previous question - can biasing into class A be done with a "straight" single-chip topology? ie. not bridged/paralleled etc.
Thanks again, everyone, I'm really glad I posted here, you've given me some wonderful ideas already - and I'm beginning to believe that, with your help, I might really be able to create something extraordinary.
I must admit that going fully active is something I did seriously think about, and also one of the reasons why i wanted to keep power and pre sections separate (the other reason being the ease of upgrading preamp in future). At this point though, I lack necessary knowledge to even begin to design such setup. I understand that and active crossover completely replaces the passive one, which means I would have a very happy multitude of things to get right with it, all at the same time: relative speaker sensitivity, phase shift, time alignment, and of course frequency control. As much as I would love to give it a stab, i regret to say that it is almost certainly beyond my ability at the moment...
I take your point about separate PSU not necessarily being an improvement. My line of thought was that an additional 30cm or so of cable is a reasonable trade-off for better shielding (and that is how the original Gaincard was laid out too), but you made me reconsider this. In fact, I've just come up with an idea how to incorporate the PSU section into my original design, while maintaining perfect symmetry AND allowing some heavy shielding to be employed around PSU... Damn, I'm good
Is transformer the only "noisy" part of the PSU, or do diode bridges radiate something undesirable as well?
I still have mixed feelings about bridging/paralleling multiple chips though. I suppose I really need to hear it myself, in order to be able to judge whether I prefer it or not. As I said earlier I would not sacrifice "directness" of sound for anything else.
One thing puzzles me though - why do you advise against silver? Is it not simply superior electrically?
Also, no one answered my previous question - can biasing into class A be done with a "straight" single-chip topology? ie. not bridged/paralleled etc.
Thanks again, everyone, I'm really glad I posted here, you've given me some wonderful ideas already - and I'm beginning to believe that, with your help, I might really be able to create something extraordinary.
And the people who say the opposite don't have reasons? The question about Peter Daniel's reasons is, which they are. The only sure thing about them is they are not objective, because from an electro-acoustic point of view a low capacitance PSU has no advantage. His reasons must be subjective then.
So is it a question of personal taste? Then you need to find out whether you share his taste or not by at least trying a normal or high capacitance PSU for a significant amount of time.
Or is it only a question of marketing, because that low capacitance PSU is one of the things that makes his products different from most others?
Hi,
First, this Speaker was build behind the iron curtain on the wrong side of the Berlin Wall and when I left in some haste and under cover of doing something in early '89 it, documentation and many other interesting pieces of personal, family and of course my audio history where left behind and never recovered...
What I remember on the MFB thing is that over years from around 86 I was messing with bridge schemes to sense the drivers back EMF (what rythmik now does), but I found that when the bridged was correctly set up for best results the heating of the driver coil and the inductance change as the coil travelled caused the bridge to unbalance and for smoke coming out of voice coils, things to go bang and all that stuff. I had a fair few woofers reconed at the time.
What I eventually implemented successfully was the classic German "electret microphone glued to the cone" scheme covered here a number of times. Using a chipamp was really incidental. The origin for me was a mid 80's publication of the west german "Funkschau" magazine. My own publication for the east german DIY electronics mag was rejected because I was considered politically unreliable and referenced western sources..
Well, much water under the bridges across Elbe, Oder and Spree since those years.
Ciao T
Dear Thorsten,
Thanks for sharing. I am currently doing a same project with a electred Mic. Do you feel to open a new topic with me where we can discuss about it? Because I am sticked with questions, and I was never able to find schematics anywhere on the Net.
I am really sorry Uncle_Leon, we get it to a new topic. Forgive me
With kind regards,
Bas
That is a fantastic post Thorsten, thanks!
I must admit that going fully active is something I did seriously think about, and also one of the reasons why i wanted to keep power and pre sections separate (the other reason being the ease of upgrading preamp in future). At this point though, I lack necessary knowledge to even begin to design such setup. I understand that and active crossover completely replaces the passive one, which means I would have a very happy multitude of things to get right with it, all at the same time: relative speaker sensitivity, phase shift, time alignment, and of course frequency control. As much as I would love to give it a stab, i regret to say that it is almost certainly beyond my ability at the moment...
I take your point about separate PSU not necessarily being an improvement. My line of thought was that an additional 30cm or so of cable is a reasonable trade-off for better shielding (and that is how the original Gaincard was laid out too), but you made me reconsider this. In fact, I've just come up with an idea how to incorporate the PSU section into my original design, while maintaining perfect symmetry AND allowing some heavy shielding to be employed around PSU... Damn, I'm good
Is transformer the only "noisy" part of the PSU, or do diode bridges radiate something undesirable as well?
I still have mixed feelings about bridging/paralleling multiple chips though. I suppose I really need to hear it myself, in order to be able to judge whether I prefer it or not. As I said earlier I would not sacrifice "directness" of sound for anything else.
One thing puzzles me though - why do you advise against silver? Is it not simply superior electrically?
Also, no one answered my previous question - can biasing into class A be done with a "straight" single-chip topology? ie. not bridged/paralleled etc.
Thanks again, everyone, I'm really glad I posted here, you've given me some wonderful ideas already - and I'm beginning to believe that, with your help, I might really be able to create something extraordinary.
Dear Leon,
Then I strongly suggest again you give my nested feedback schematic a try with help from us. Because I heard the difference between the standard simplistic gain-clone and the one with the Nested Feedback, and I can guarantee you that the implementation as showed in my earlier post sounds day and night better in all aspects. Even though the circuit looks more complicated (and we want less is more) keep in mind that the chip is basically a slave of the input opamp, and the end result will be as good as the input opamp is.
With kind regards,
Bas
Maybe Dr. Linkwitz can help?
Indeed they do. ~100 nF capacitors directly on the rectifier pins help snub their switching noise up.
No. In a solid state amplifier you bias the output stage into class A so that both the positive and the negative power transistors remain switched on even when no signal or only a low signal is present. The bias current passes through both transistors.
The trick to bias a chipamp into class A is to give it a DC offset, so one of the output transistors (e. g. the positive one) remains switched on. The other does not, which means you need a second chipamp with opposite bias to provide the other half of the signal (the negative). Then you connect the two amps in parallel and hope that the bias current is equally high so that it passes from the positively biased amp to the negatively biased one. When the currents are not perfectly matched the difference will pass through the speaker, which is bad. Even if it does not destroy the speaker, it will probably produce more distortions than the class A bias will reduce. One also has to be alert to thermal drift which could affect the two bias settings in an unfavourabe way during operation. If you want a class A amplifier Pass DIY: Home of DIY Audio, Amplifiers, Preamps and Speakers is the place to go.
(edit) Well you could bias the chip into SE class A. That would mean you give it a DC offset that is half of one part of the supply voltage. That would be a quite ridiculous arrangement, where you provide a split power supply, one half of which is used to keep the corresponding transistor from turning on, you get a DC voltage at the output that forces you to use a capacitor in series with the load and you only use half of the output stage.
Last edited:
That is why I like the nested feedback approach. Like Quad's current dumping, it compensate for the class B related crossover distortion, and biasing in class A isn't a must anymore.
With kind regards,
Bas
Hi,
This assumes the speaker designer did get it right, which is not neccesarily the case. Which specific speaker do you have? If you are looking at building a new one, going for an "active" design groundup can be a very good thing.
If you use a programmable digital crossover it is very easy to sort out the crossover and to indeed change it. Pro-audio units like the Behringer DCX2496 (24 Bit & 96KHz digital crossover) need modifications for our kind of purposes, but happily these are mostly subtractive and hence easy to do.
BEHRINGER: DCX2496
The DCX2496 has six outputs (so 3-way stereo) and allows time alignment, crossovers and equalisation to be applied individually to each output. Thus you can first equalise the drive units pretty much flat for the range of interest and maybe one octave overlap, time align them for the listening position and then use any particular textbook crossover setting as starting point for fine tuning.
Of course, you need means to measure, but an inexpensive external pro-audio sound card with phantom powered microphone inputs and a decent measurement microphone can make this job very easy.
I suspect that unless you have a lot of money in a speaker designed to appeal to ear and measurements together the result will both measure and sound A LOT better than the original speaker.
I did something like that in the late 90's with a PMC 3-Way studio monitor that was already active, but with an analog crossover. With new better amp's (Bryston originally) and fully time-aligned and equalised flat the improvement was stunning. We used a digital X-Over that was 20-Bit/48KHz from BSS at the time, heavily modified.
Maybe at the moment, but is not THAT HARD to change.
I would not use 1pcs diode bridges, but discrete diodes. To be honest, I do not like the diodes commonly used in Gainclones.
Mine uses 40CPQ100 Dual Schottky diodes. These are common cathode, just like old tube rectifiers, so you need 3pcs to make one bridge, in my case 6pcs for the whole stereo amp.
You can wrap a torroidal mains transformer with Mu-Metal strips cut from these kind of sheets here:
Ultraperm MUMetal Shielding Sheet 0.15mm, 31.5cm x 19.5cm 80% nickel | Diy HiFi Supply
I was rather sceptical about this myself, but it works well. I just wind a (set of) strip(s) around the outside of the torroid. It cuts hum rather appreciably.
Diodes themselves are not really noisy, but the switching and rectification current spikes can be radiated by the wiring from the bridges and to them, tightly twisting the wires that carry the circulating currents can reduce this majorly.
I tend to position PSU electrolytic capacitors as free "shield" between noisy PSU sections and the rest. This can work out very well.
Okay, I would suggest "balanced amplifiers" if you have a balanced source (like the DAC Chips from the Behringer digital crossover) and then to use LM3875 for MidRange/HighRange or for FullRange Amplifiers, likely operated at no more than +/-24V.
I would not recommend the Bridge/Parallel design shown by Nat Semi for full range of Mf/Hf Amplifiers. But they are pretty good for easy/quick Amp's for "heavy lifting" in the bass.
Silver wiring/cables have a "sound" of it's own. This sound quality rarely goes well with solid state systems.
Sure, so what? You are listening to the sound, right, not measuring resistance, right?
I did answer it. I commented it helps the treble to be a little smoother. It can be done with any chipamp.
Ciao T
This assumes the speaker designer did get it right, which is not neccesarily the case. Which specific speaker do you have? If you are looking at building a new one, going for an "active" design groundup can be a very good thing.
If you use a programmable digital crossover it is very easy to sort out the crossover and to indeed change it. Pro-audio units like the Behringer DCX2496 (24 Bit & 96KHz digital crossover) need modifications for our kind of purposes, but happily these are mostly subtractive and hence easy to do.
BEHRINGER: DCX2496
An externally hosted image should be here but it was not working when we last tested it.
The DCX2496 has six outputs (so 3-way stereo) and allows time alignment, crossovers and equalisation to be applied individually to each output. Thus you can first equalise the drive units pretty much flat for the range of interest and maybe one octave overlap, time align them for the listening position and then use any particular textbook crossover setting as starting point for fine tuning.
Of course, you need means to measure, but an inexpensive external pro-audio sound card with phantom powered microphone inputs and a decent measurement microphone can make this job very easy.
I suspect that unless you have a lot of money in a speaker designed to appeal to ear and measurements together the result will both measure and sound A LOT better than the original speaker.
I did something like that in the late 90's with a PMC 3-Way studio monitor that was already active, but with an analog crossover. With new better amp's (Bryston originally) and fully time-aligned and equalised flat the improvement was stunning. We used a digital X-Over that was 20-Bit/48KHz from BSS at the time, heavily modified.
Maybe at the moment, but is not THAT HARD to change.
I would not use 1pcs diode bridges, but discrete diodes. To be honest, I do not like the diodes commonly used in Gainclones.
Mine uses 40CPQ100 Dual Schottky diodes. These are common cathode, just like old tube rectifiers, so you need 3pcs to make one bridge, in my case 6pcs for the whole stereo amp.
You can wrap a torroidal mains transformer with Mu-Metal strips cut from these kind of sheets here:
Ultraperm MUMetal Shielding Sheet 0.15mm, 31.5cm x 19.5cm 80% nickel | Diy HiFi Supply
I was rather sceptical about this myself, but it works well. I just wind a (set of) strip(s) around the outside of the torroid. It cuts hum rather appreciably.
Diodes themselves are not really noisy, but the switching and rectification current spikes can be radiated by the wiring from the bridges and to them, tightly twisting the wires that carry the circulating currents can reduce this majorly.
I tend to position PSU electrolytic capacitors as free "shield" between noisy PSU sections and the rest. This can work out very well.
Okay, I would suggest "balanced amplifiers" if you have a balanced source (like the DAC Chips from the Behringer digital crossover) and then to use LM3875 for MidRange/HighRange or for FullRange Amplifiers, likely operated at no more than +/-24V.
I would not recommend the Bridge/Parallel design shown by Nat Semi for full range of Mf/Hf Amplifiers. But they are pretty good for easy/quick Amp's for "heavy lifting" in the bass.
Silver wiring/cables have a "sound" of it's own. This sound quality rarely goes well with solid state systems.
Sure, so what? You are listening to the sound, right, not measuring resistance, right?
I did answer it. I commented it helps the treble to be a little smoother. It can be done with any chipamp.
Ciao T
Hi,
Do you read German? I have some scan's I can send you, which I got in preparation for a revial of this old idea. I could e-mail them, PM me.
Ciao T
Do you read German? I have some scan's I can send you, which I got in preparation for a revial of this old idea. I could e-mail them, PM me.
Ciao T
Really ?..
But when contradicting someone, paralleling chips seems relevant to you..
Some sort of opportunism...
http://www.diyaudio.com/forums/chip...rival-hi-end-design-advice-2.html#post2145510
I designed a 6x LM3886 bridge/parallel circuit for a commercial active speaker for a heavy subwoofer. I can assure you it has the balls and the current to blow the paintings from your wall...
As long you give it a loads of heatsink, because things get really hot, and you don't want the Thermal Protection spoils the fun. We choosed the chips for their reliablity, no need for bias adjustment in mass production and for it's relative compact size. If you know the parameters of the woofer, you can careful adjust the power supply voltages to get the max out of the design without stress them out.With kind regards,
Bas
Last edited:
Hi,
Hardly opportunism. Merely correcting your misrepresentation claimint it was not possible to get high output current with chipamps.
Note, it is possible to design a Bridge/parallel chip-amp that subjectively gives up nothing in clarity etc. to a single chip, but has much more power and load compatibility.
However, the method to do so has not been documented by anyone and I am not going to document it so any KopyKat from Kanton can bring out a kit and make money of it. Let them copy Nat Semi's app-note.
Ciao T
Hardly opportunism. Merely correcting your misrepresentation claimint it was not possible to get high output current with chipamps.
Note, it is possible to design a Bridge/parallel chip-amp that subjectively gives up nothing in clarity etc. to a single chip, but has much more power and load compatibility.
However, the method to do so has not been documented by anyone and I am not going to document it so any KopyKat from Kanton can bring out a kit and make money of it. Let them copy Nat Semi's app-note.
Ciao T
No doubt that it can work, although i would prefer buffering
the output with high power Bjts.
Seems simpler to implement.
I once tested this kind of configuration, and it worked
well enough for the purpose, i;e , driving a low impedance
basses box.
cheers,
wahab
But then still you need to bias, and that adds more time for mass production.
With kind regards,
Bas
Hi,
Okay, but you did not exactly say it like that, did you?
Shall we call it even stevens then?
BTW, I am still puzzled by your comments on the sonics of chipamp's in the same context. Would you mind setting out the conditions and context? It may help others to avoid getting equally poor results if we can identify what went wrong.
For fun, I just designed a special "current dumper" power booster using a Nat Semi Chip, intended to be run from a high impedance tap of SET Amp. This "booster" is of course unity gain (yes, it is possible to make a unity gain follower from an LM38XX).
Switching this in and out (it is switchable) shows no subjectively detectable difference, except the tighter bass (lower Zout) and extra power. This suggests that a correctly applied LM38XX is essentially transparent sonically speaking.
Ciao T
Okay, but you did not exactly say it like that, did you?
Shall we call it even stevens then?
BTW, I am still puzzled by your comments on the sonics of chipamp's in the same context. Would you mind setting out the conditions and context? It may help others to avoid getting equally poor results if we can identify what went wrong.
For fun, I just designed a special "current dumper" power booster using a Nat Semi Chip, intended to be run from a high impedance tap of SET Amp. This "booster" is of course unity gain (yes, it is possible to make a unity gain follower from an LM38XX).
Switching this in and out (it is switchable) shows no subjectively detectable difference, except the tighter bass (lower Zout) and extra power. This suggests that a correctly applied LM38XX is essentially transparent sonically speaking.
Ciao T
Hi,
Not neccesarily. You could add a significant current sink to a chipamp to make the output Class A for normal operation (say first watt) and use a complementary pair of "current dumper" transistors inside the feedback loop. Simply make the "dumpers" turn on at around 80% of the class A bias current (I assume you know how to).
But then you might as well buy the Nat Semi Driver Chip and some thermal-track transistors and make an amp from that. You may have to adjust bias (but generally you can do without or you can "servo" the adjustment).
But that is a wholly different kettle of fish and once we are that far we can build even more interesting stuff. It will not really comprehensively better chipamp based stuff though, so for easy DIY "quality" amps the good old Gainclone (I cannot remember who coined the term first, I do not think it was me) takes a fair bit of beating and even Apogee Scintilla's can be driven well by bridged/parallel versions with better power supplies.
Ciao T
Not neccesarily. You could add a significant current sink to a chipamp to make the output Class A for normal operation (say first watt) and use a complementary pair of "current dumper" transistors inside the feedback loop. Simply make the "dumpers" turn on at around 80% of the class A bias current (I assume you know how to).
But then you might as well buy the Nat Semi Driver Chip and some thermal-track transistors and make an amp from that. You may have to adjust bias (but generally you can do without or you can "servo" the adjustment).
But that is a wholly different kettle of fish and once we are that far we can build even more interesting stuff. It will not really comprehensively better chipamp based stuff though, so for easy DIY "quality" amps the good old Gainclone (I cannot remember who coined the term first, I do not think it was me) takes a fair bit of beating and even Apogee Scintilla's can be driven well by bridged/parallel versions with better power supplies.
Ciao T
Well, no need of bias, since the chip will provide all the output
current under 2A, the added transistors will start conducting
as soon as this current is not enough to feed the load..
Undoubtly, there will be some slight distorsion added,
but the high negative feedback ratio will easily get rid
of it..
regards,
wb
Not wishin to be polemical, but i was pointing the single circuit
specifications.
That said, i can only aknowledge that for the money, the chip
has very good caracteristics.
I woudn t go at unity gain, since the internal compensation
is implemented to make the chip unconditionnaly stable for
the minimum gain specified by NS.
Below this value, it can be stable, but only conditionnaly,
which is never good for the sonics.
I am glad you mention this. A few post back, someone could't believe a difference between Z-out of 0.01 Ohm or 4uOhm made any obvious difference in sound
With kind regards,
Bas
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.