The problem I'm talking about is rarely noticed since few people bother to test op amp input stages with wildly mismatched impedances. It doesn't happen very often in reality, except when you're using a passive volume control. Also, note that I tested with 8Vrms, 10kHz and 3kohm only to get about -100dB THD. Test conditions closer to reality would yield distortion contributions well below that of the amplifier.
Ncore was announced just a month ago so no actual end-user products are available that use Ncore (except the new batch of Grimm LS1, but I take it they had some advance notice).
Ncore was announced just a month ago so no actual end-user products are available that use Ncore (except the new batch of Grimm LS1, but I take it they had some advance notice).
Bruno very kindly PM'ed me with a response to this. Bottom line, Hypex-powered amps are far more widespread than I realized. Great news.
Shocker: He keeps his maniacal Grimm twin on speed dial!
Hi Bruno, some silly question came to mind...
Like the UCD´s I guess the NC will have similar topology having a speaker (cold) connection on the board which is connected to the ground connection on the board, no?
If so, would there be any possible gains in improving this connection e.g. by ultra thick copper traces/thick copper wire links to lower the resistance between speaker and PS ground?
Are there any differences between the current UCD models regarding the resistance from speaker cold to PS connection on the boards?
cheers,
Like the UCD´s I guess the NC will have similar topology having a speaker (cold) connection on the board which is connected to the ground connection on the board, no?
If so, would there be any possible gains in improving this connection e.g. by ultra thick copper traces/thick copper wire links to lower the resistance between speaker and PS ground?
Are there any differences between the current UCD models regarding the resistance from speaker cold to PS connection on the boards?
cheers,
Like the UcD module, the NC has two feedback traces coming directly off the connector so any resistance between the negative terminal and circuit ground is canceled by the feedback loop. This is why the negative speaker terminal is labeled "LS-", not "ground". In our schematics, the differential feedback nets are not labeled "feedback" and "ground" but "feedback-pos" and "feedback-neg" and they're treated absolutely equally. In fact, the whole modulator and feedback circuit is differential. That doesn't mean the signals are symmetrical with respect to ground, but that ground potential is no part of the equation. Only the difference between signal pairs counts, whether or not any of them is held at ground potential at some point. I'll probably write an article explaining this style of circuit design for Linear Audio. I think many people would be helped by an explanation of how to eliminate ground from your signal path altogether even whilst using it to return current.
For the NC400 I'm having custom screw tags made, shaped like a "П" with a screw through the top and a square nut inside so the speaker wire is squeezed between the nut and the underside of the П. You can attach 2 wires, one to the left of the screw, one to the right.
Of course both ends of the П are soldered into the PCB. One end goes to the amplifier output (or ground for the other terminal), feedback is taken from the other end of the П. What this means is that you get perfect 4-point sensing at the place where the speaker wire makes galvanic contact. For something like Ncore this makes sense. Output impedance being below 3 milliohms at 20kHz, a few centimetres of wire has more inductance than this. Of course you'd wonder why be so picky about the connection if the speaker cable will dominate anyway. The answer is biwiring. A lot is being said about sense or nonsense of biwiring but what it is about is that loudspeakers are nonlinear. If the woofer draws a distorted current, biwiring can prevent this current from causing distortion in the voltage going to the tweeter. What it takes is an amplifier with output impedance much lower than that of the speaker cable. Running two separate cables between the amp and speaker will make a difference. Since the amplifier feedback senses exactly where the speaker cable is connected, you can just connect two sets at the same point and be certain that the common impedance between the HF/LF current loops is just the amplifier's output impedance. If you're fanatic you could make sure that the wires of the two cables are inserted on alternating sides of the screw. That way the 4mm of resistance between the 2 points where the wires contact the terminal gets canceled too
For the NC400 I'm having custom screw tags made, shaped like a "П" with a screw through the top and a square nut inside so the speaker wire is squeezed between the nut and the underside of the П. You can attach 2 wires, one to the left of the screw, one to the right.
Of course both ends of the П are soldered into the PCB. One end goes to the amplifier output (or ground for the other terminal), feedback is taken from the other end of the П. What this means is that you get perfect 4-point sensing at the place where the speaker wire makes galvanic contact. For something like Ncore this makes sense. Output impedance being below 3 milliohms at 20kHz, a few centimetres of wire has more inductance than this. Of course you'd wonder why be so picky about the connection if the speaker cable will dominate anyway. The answer is biwiring. A lot is being said about sense or nonsense of biwiring but what it is about is that loudspeakers are nonlinear. If the woofer draws a distorted current, biwiring can prevent this current from causing distortion in the voltage going to the tweeter. What it takes is an amplifier with output impedance much lower than that of the speaker cable. Running two separate cables between the amp and speaker will make a difference. Since the amplifier feedback senses exactly where the speaker cable is connected, you can just connect two sets at the same point and be certain that the common impedance between the HF/LF current loops is just the amplifier's output impedance. If you're fanatic you could make sure that the wires of the two cables are inserted on alternating sides of the screw. That way the 4mm of resistance between the 2 points where the wires contact the terminal gets canceled too
Since you are pushing the envelope with your feedback connection, you could allow a split feedback and do a separate sense line like four-wire sensors (PRT's etc). you could run a second set of speaker leads (small diameter would be fine) and take your feedback point at the speaker terminals or even the driver itself in the case of active crossovers.
Sheldon
Interesting with a biwiring connection which is designed into the terminal and its feedback connection -especially as NC400 is to be a Hi End module. The only problem with biwring is that you normally alter the electrical properties of the cable as you for instance double the capacitance. This means that biwiring has to be thought into the the design of the cable itself to achieve the intended balance between inductance and capacitance hence tonal and dynamic balance of the sound with two separate runs of cable. Even though Nordost cables seems to be designed for biwiring (often four packs of individually isolated conductors next to each other) they do sound quite different according to which "pack" of conductors that goes to HF and LF. So although biwiring has some technical pros, be careful of the cons that usually follows. The special connection of NC400 together with the very low output impedance and load independent operation may change the game somewhat though. (Bruno, do you have any experiences on the matter of singlewiring/biwiring in relation to how UCD/NC compares to other amplifier types?)
... I would probably still advocate for biamping over biwiring due to the above explanation; so add more NC´s ;-)
cheers,
... I would probably still advocate for biamping over biwiring due to the above explanation; so add more NC´s ;-)
cheers,
Hello Bruno
A couple of questions , will you be developing a new power supply for the NC400 or will the SMPS400A400 do the job.
Secondly with the discrete input stage why did you want to use the MAT14 is it do with the needing some thermal tracking for the third transistor which is used to generate a current reference to improve its tracking for compensating the base current.
A couple of questions , will you be developing a new power supply for the NC400 or will the SMPS400A400 do the job.
Secondly with the discrete input stage why did you want to use the MAT14 is it do with the needing some thermal tracking for the third transistor which is used to generate a current reference to improve its tracking for compensating the base current.
Biwiring: Use separate cables, not cables that bunch both runs together. You want to avoid coupling between the runs, not promote it. You'll find that with an amp like Ncore, biwiring sounds almost like biamping. Bad for my business to tell you this, I know...
The SMPS400 will do the job, except for maximum power (which is 600W/2ohm on the NC400). We'll make a special SMPS version for the Ncore with all the right connectors but essentially it'll be a beefed up SMPS400.
The MAT14 was indeed for bias current cancellation. However, I forgot that this circuit will add a nonlinear capacitance to the input as well so I decided to drop it. That in addition to the fact that you can't get non-monolithic matched quads. Instead I'll use matched duals which you can get as separate dice, and rearrange them so that the input currents of the two noninverting inputs are matched. Then I can cancel the error with a pot. It's not that critical anyway.
The SMPS400 will do the job, except for maximum power (which is 600W/2ohm on the NC400). We'll make a special SMPS version for the Ncore with all the right connectors but essentially it'll be a beefed up SMPS400.
The MAT14 was indeed for bias current cancellation. However, I forgot that this circuit will add a nonlinear capacitance to the input as well so I decided to drop it. That in addition to the fact that you can't get non-monolithic matched quads. Instead I'll use matched duals which you can get as separate dice, and rearrange them so that the input currents of the two noninverting inputs are matched. Then I can cancel the error with a pot. It's not that critical anyway.
Last edited:
bunching cable runs is usually a bad idea as it further adds capacitive coupling but although spaced apart two cable runs still doubles the capacitive coupling like paralleling two capacitors... From my experience this can really screw the balance and performance of a cable, but I don´t know whether that holds true for all amps or the whether the audible effect is reduced by high feedback and low output resistance...
Almost like biwiring, that statement sure helps to sell the first modules
Almost like biwiring, that statement sure helps to sell the first modules
The capacitance in the Ncore module's output filter is 2uF. This is effectively strapped directly across the output terminals. The amplifier wouldn't notice even if you attached 200nF worth of cable. But even for more normal amplifiers, you'll find that the only cable parameter that produces easily measurable differences at the speaker end is inductance. Likewise, speaker cables do not couple capacitively, they couple inductively. Keeping them apart is the easiest way to prevent them from undoing the benefit of biwiring. Another method is the one used in cat5 cables: use different pitches of twist.
My experiences with all sorts of cable fidling is based on ICEpower, arcam and Naim amps so they use quite different approaches concerning output "filter" and on all slight differences in capacitive coupling in the speaker cables were audible and a great concern to get right. Quite similar results actually although the stark differences in the amps...
I haven´t done the same experiements on UCD´s and of course neither the NC with its 2uF! coupling between the outputs -that is A LOT BTW! -I´ll guess the feedback scheme must be really effective to get the output stable with that load Isn´t that cap magnitudes larger than on the UCD´s??
Well I´ll guess as usual there is difference between what is measurable and audible...
BTW CAT5 can actually be made into rather good speaker cable, but IME one separate run for each pole is necessary to get a decent sound...
I haven´t done the same experiements on UCD´s and of course neither the NC with its 2uF! coupling between the outputs -that is A LOT BTW! -I´ll guess the feedback scheme must be really effective to get the output stable with that load
Isn´t that cap magnitudes larger than on the UCD´s?? Well I´ll guess as usual there is difference between what is measurable and audible...
BTW CAT5 can actually be made into rather good speaker cable, but IME one separate run for each pole is necessary to get a decent sound...
UcD180 to 400: 680nF. UcD700: 1uF. UcD2k: 2uF. The cap doubles as one of the poles of the feedback loop function so it doesn't get in the way of stability. Quite the contrary. I've always been quite pleased with that idea
Never tried using cat5 like you described. I've always parallelled the pairs (all whites=-, all colours=+). I don't know how big the effect would be but I feel suspicious of loudspeaker currents running round loops where the wires can move.
Never tried using cat5 like you described. I've always parallelled the pairs (all whites=-, all colours=+). I don't know how big the effect would be but I feel suspicious of loudspeaker currents running round loops where the wires can move.
Hello.
I have been folloving this thread for a while with great interest and as it seems I'm just missing one aspect before adding at least one pair too my christmas list. Do you have any idea how much they will set me back ? The plan so far is 2x smps1200 + 2x NC. I'm not expecting an exact number but some idea would be nice.
Best regards. Bengt Dahlberg
I have been folloving this thread for a while with great interest and as it seems I'm just missing one aspect before adding at least one pair too my christmas list. Do you have any idea how much they will set me back ? The plan so far is 2x smps1200 + 2x NC. I'm not expecting an exact number but some idea would be nice.
Best regards. Bengt Dahlberg
Maybe I schould clarify that I'm not looking for the cheapest solution possible, I'm quite aware that good quality comes at a price and that's just as it schould be.
Cat out of the bag?
Hi Bruno,
So in theory you could supply us hardcore DIY types with Ncore modulators to retro-fit our existing amps?
Hi Bruno,
So in theory you could supply us hardcore DIY types with Ncore modulators to retro-fit our existing amps?
Price: the internal price discussion is still ongoing.
Retrofit modulator: That would be kinda cool and it's technically feasible but how many of those would we sell? I don't think it's an economical proposition for us to do this. Besides, there are other improvements (most notably idle losses and open-loop distortion) that retrofitted modules wouldn't benefit from.
Retrofit modulator: That would be kinda cool and it's technically feasible but how many of those would we sell? I don't think it's an economical proposition for us to do this. Besides, there are other improvements (most notably idle losses and open-loop distortion) that retrofitted modules wouldn't benefit from.
- Status
- Not open for further replies.