running
Got my Purifi modules working last night, just a quick set up like this:
Connex 800 SMPS, 65 volt rails (custom), Belleson regulator for the VDR supply
An old Twisted Pair Audio Placid BiPolar supply for the OPAs/modulator
Everything is just sitting there with wires going all over, this is just for the initial testing...
First impression is that the sound is clearly better than my NC-400 stereo amp build (which is very well optimized) Soundstage depth seems to be better. This sound quality is promising enough for me to be excited about getting this put together in a finished build, I am making a dedicated +/- 18 VDC supply for it now, then to figure out the chassis. By then, maybe there will be some more aftermarket gain stages available to try. No interest in the NuTube distortion add on though... but I would love to see someone make a really good discrete buffer for this: Bruno Putzeys admitted (way back during NC-400 development) that his discrete gain stage he used on the NC-400 outperformed anything he could do with IC OPAs... There does seem to be a slightly higher noise floor (ear to drivers) than the NC-400, but still totally inaudible of course (maybe due to haphazard wiring right now). Neither Hypex nor Connex seem to publish any noise spec for their supplies, so it is hard to say if one or the other would result in quieter performance.
Everything worked right away, no weird anomalies or anything, power up is quiet as a mouse, with the Nampon in the "off" position. I measured DC offset ~60 mV cold before I attached speaker cables, anyone else have a measure for their offset?
I do not have much of any heat sinking (the modules are sitting on a piece of Al plate, not fastened) so no higher volume testing now, but they are making almost no heat. The Connex supply runs really cool as well, especially compared to Hypex ones I am familiar with (SMPS 600 and SMPS 1200).
Got my Purifi modules working last night, just a quick set up like this:
Connex 800 SMPS, 65 volt rails (custom), Belleson regulator for the VDR supply
An old Twisted Pair Audio Placid BiPolar supply for the OPAs/modulator
Everything is just sitting there with wires going all over, this is just for the initial testing...
First impression is that the sound is clearly better than my NC-400 stereo amp build (which is very well optimized) Soundstage depth seems to be better. This sound quality is promising enough for me to be excited about getting this put together in a finished build, I am making a dedicated +/- 18 VDC supply for it now, then to figure out the chassis. By then, maybe there will be some more aftermarket gain stages available to try. No interest in the NuTube distortion add on though... but I would love to see someone make a really good discrete buffer for this: Bruno Putzeys admitted (way back during NC-400 development) that his discrete gain stage he used on the NC-400 outperformed anything he could do with IC OPAs... There does seem to be a slightly higher noise floor (ear to drivers) than the NC-400, but still totally inaudible of course (maybe due to haphazard wiring right now). Neither Hypex nor Connex seem to publish any noise spec for their supplies, so it is hard to say if one or the other would result in quieter performance.
Everything worked right away, no weird anomalies or anything, power up is quiet as a mouse, with the Nampon in the "off" position. I measured DC offset ~60 mV cold before I attached speaker cables, anyone else have a measure for their offset?
I do not have much of any heat sinking (the modules are sitting on a piece of Al plate, not fastened) so no higher volume testing now, but they are making almost no heat. The Connex supply runs really cool as well, especially compared to Hypex ones I am familiar with (SMPS 600 and SMPS 1200).
Last edited:
No...
NC-400 is a dual mono build in a single chassis with 2 SMPS 600, basically stock Hypex set up, with wiring changes (Cardas 21.5 AWG golden ratio litz wire for input, and 11.5 AWG golden ration litz wire direct soldered to the board for speaker output). Also use Hi Fi Tuning Supreme Cu fuses in the SMPS 600s. the Purifi is better than NC-400 subjectively without any doubt here...
One very interesting difference is that the output inductors run very hot in the NC-400, they stay cool in the Purifi, perhaps this has to do with the reduction in hysteresis Mr. Putzeys has mentioned?
NC-400 is a dual mono build in a single chassis with 2 SMPS 600, basically stock Hypex set up, with wiring changes (Cardas 21.5 AWG golden ratio litz wire for input, and 11.5 AWG golden ration litz wire direct soldered to the board for speaker output). Also use Hi Fi Tuning Supreme Cu fuses in the SMPS 600s. the Purifi is better than NC-400 subjectively without any doubt here...
One very interesting difference is that the output inductors run very hot in the NC-400, they stay cool in the Purifi, perhaps this has to do with the reduction in hysteresis Mr. Putzeys has mentioned?
Pardon my momentary diversion of the thread. Since my business reputation is on the line here, I feel compelled to respond.
I do recall that conversation. I do NOT recall attacking you personally in it, however. I do recall disagreeing with you, though, and I do recall pointing out that your tendency to communicate with LARGE pictures that completely trash the flow of the thread, font sizes and colours does nothing to strengthen your argument. I think it was around then, you decided to leave the conversation (and future conversations). I do recall that I found your reaction a bit melodramatic, but since I didn't find your contributions to the thread to be that insightful - and many of them to be factually wrong - I chose to shrug and move on. The increase in thread SNR following our transgression was welcome as well. I'm a bit surprised that - years later - you're still upset and expecting an apology. Here you go (note that I resized the image so it wouldn't blow up the thread):
I do agree that I tend to be pretty direct in my communication. I also tend to be fact-based and science-based. That ruffles some people's feathers - including yours apparently. I have no plans to change.
I do not recall feeling personally attacked by you. Trust me! You don't have that kind of power over me. I just don't take everything you say as gospel.
I'm not sure why you're hung up on the slew rate needed. The math is pretty simple: SR = 2*pi*Vpeak*f, where SR is the slew rate needed to accurately reproduce frequency, f, to a peak amplitude of Vpeak.
So to reproduce 40 V peak at 20 kHz (so 100 W into 8 Ω at 20 kHz), you'd need a slew rate of: SR = 2*pi*40*20000 = 5.026 MV/s (also known as 5.026 V/us).
You can find the peak voltage needed for a particular power level as: Vpeak = sqrt(2*R*P), where Vpeak is the peak amplitude, R is the load impedance, P is the output power, and sqrt() is square root.
Now you can argue that the human hearing extends beyond 20 kHz and you can redo the math for other power levels, but the fundamental math does not change. You can read a bit about it on the Wikipedia page for Power Bandwidth. Should that not convince you, I suggest deriving the math yourself. It's pretty simple to do. All you have to do is to take the derivative of the sine wave at the zero crossing (that's where the derivative is the largest).
I also recommend engineering texts on the topic, such as Sergio Franco, "Design with Operational Amplifiers and Analog Integrated Circuits". There's a 4th edition now. I also seem to recall that you can find a full 3rd edition as a .pdf online if you search hard enough.
Tom
M.Sc.EE; B.Sc.EE (Hons); BA (Psych, Hons)
I do recall that conversation. I do NOT recall attacking you personally in it, however. I do recall disagreeing with you, though, and I do recall pointing out that your tendency to communicate with LARGE pictures that completely trash the flow of the thread, font sizes and colours does nothing to strengthen your argument. I think it was around then, you decided to leave the conversation (and future conversations). I do recall that I found your reaction a bit melodramatic, but since I didn't find your contributions to the thread to be that insightful - and many of them to be factually wrong - I chose to shrug and move on. The increase in thread SNR following our transgression was welcome as well. I'm a bit surprised that - years later - you're still upset and expecting an apology. Here you go (note that I resized the image so it wouldn't blow up the thread):
I do agree that I tend to be pretty direct in my communication. I also tend to be fact-based and science-based. That ruffles some people's feathers - including yours apparently. I have no plans to change.
I do not recall feeling personally attacked by you. Trust me! You don't have that kind of power over me. I just don't take everything you say as gospel.
I'm not sure why you're hung up on the slew rate needed. The math is pretty simple: SR = 2*pi*Vpeak*f, where SR is the slew rate needed to accurately reproduce frequency, f, to a peak amplitude of Vpeak.
So to reproduce 40 V peak at 20 kHz (so 100 W into 8 Ω at 20 kHz), you'd need a slew rate of: SR = 2*pi*40*20000 = 5.026 MV/s (also known as 5.026 V/us).
You can find the peak voltage needed for a particular power level as: Vpeak = sqrt(2*R*P), where Vpeak is the peak amplitude, R is the load impedance, P is the output power, and sqrt() is square root.
Now you can argue that the human hearing extends beyond 20 kHz and you can redo the math for other power levels, but the fundamental math does not change. You can read a bit about it on the Wikipedia page for Power Bandwidth. Should that not convince you, I suggest deriving the math yourself. It's pretty simple to do. All you have to do is to take the derivative of the sine wave at the zero crossing (that's where the derivative is the largest).
I also recommend engineering texts on the topic, such as Sergio Franco, "Design with Operational Amplifiers and Analog Integrated Circuits". There's a 4th edition now. I also seem to recall that you can find a full 3rd edition as a .pdf online if you search hard enough.
Tom
M.Sc.EE; B.Sc.EE (Hons); BA (Psych, Hons)
Last edited:
Just...
To try something different. In some ways the Connex appears to be a more advanced design, but I am no expert in SMPS, it certainly appears to be more efficient than Hypex, considering how cool it runs. Connex claims lower noise levels for their approach, but does not provide any specs so...
To me SMPS design is basically alchemy, so I would love to hear from any members well versed in SMPS design to hear what they think about the various design approaches.
Perhaps I need to pick up a Hypex SMPS 1200 and compare as well...
And then there is dual mono with two supplies-lots of things to try...
I do like the idea of using a custom linear for the input stage and modulator, for here the efficiency of a SMPS does not really matter much to overall power use. And I am fairly competent in low level linear power supplies...
To try something different. In some ways the Connex appears to be a more advanced design, but I am no expert in SMPS, it certainly appears to be more efficient than Hypex, considering how cool it runs. Connex claims lower noise levels for their approach, but does not provide any specs so...
To me SMPS design is basically alchemy, so I would love to hear from any members well versed in SMPS design to hear what they think about the various design approaches.
Perhaps I need to pick up a Hypex SMPS 1200 and compare as well...
And then there is dual mono with two supplies-lots of things to try...
I do like the idea of using a custom linear for the input stage and modulator, for here the efficiency of a SMPS does not really matter much to overall power use. And I am fairly competent in low level linear power supplies...
The Connex is regulated were the Hypex is not. I wonder if this is what you hear. I have been thinking about trying regulated smps for my nC400 monos. (Also use one smps600 per channel)
I did read a thread where the claim was that Connex had more noise than Hypex, but can't find it.
When it comes to lower wattage aspects, I think the Smooth Like Butter and Salas Shunt Regulator are a good option. I also did mention in the [Hypex NCore NC500 build] thread that I am considering a linear supply - but have not calculated the cap bank size yet. The Hypex 1200a400 can deliver 25A (25 joules) at max peak, so that could be a target.
- EDIT-
I am not super sensitive to a lower efficiency.
When it comes to lower wattage aspects, I think the Smooth Like Butter and Salas Shunt Regulator are a good option. I also did mention in the [Hypex NCore NC500 build] thread that I am considering a linear supply - but have not calculated the cap bank size yet. The Hypex 1200a400 can deliver 25A (25 joules) at max peak, so that could be a target.
- EDIT-
I am not super sensitive to a lower efficiency.
Higher order means here more than one loop (feedback or feedforward) or loops that feed back/forward to more than one place, as explained in the famous The F-word article by BP.
I think our listening impressions are similar. The NC400 and a good NC500 build are supposed to be quite similar. Soundstage is bigger but also more forward, without losing depth in what is "behind" the speakers. But it is the eerie naturalness of the treble that hit me. Almost scary, as if I had removed components from the system...
That makes me even more eager to build my 4th order with Be tweeter at the top and listen via this amplifier.
This is really useful information! Your feedback is making me build one.
I will like the take the same route and pair it with a Salas UBiB for the +/-18VDC supplies.
The Connex part number to use is SMPS800RE or is it the 800RS?
TIA.