Power amp build

[Buffle]

I’ve had a box of Hypex parts around and thought I’d put together a one-off power amp build for personal use based on NC500 modules.  This will be a stereo power amp capable of very high performance in terms of sense of resolution and speed, load invariance, and prodigious power delivery.  Fully dimensioned sound.  Given that the object is to use up some of my Hypex parts, that has served as the starting point. 

 

These modules only contain the drive stage, so require a voltage gain stage capable of driving the 1800 ohm load of the drive module.  It is this predrive stage that will determine the character of the sound, with the nCore drive stages capable of delivering accurately on what they are fed with.  The usual approach is to use op amps, whether integrated or discrete, but that would take away most of the design freedom.  I’ve designed two predrive boards: one using JFETs into MOSFETs in a similar but modified cct used by Nelson Pass, and another using JFETs into step-up transformers into more JFETs.  I’m building the former first, for no particular reason, and will build and test the transformer-based board later.

 

 The predrive stage and NC500 drive stage will be all about precision and lightning fast transients, and this power supply is built around that goal.  I’m using a Hypex SMPS for the drive stage.  I’ve tested these thoroughly and they’re very low noise and capable of delivering more current than I’ll be using.  For the predrive stage and drive module Vaux and Vdr I’m using a linear power supply with CRC filtering for the main rails with shunt regulation for the more critical stages.  As I’ve got current to spare I’m using basic but excellent LED shunt regulation applied to the already CRC filtered rails.  This is a brute force method but will deliver excellent low-noise performance.  I need around 100 mA for the positive rail and 30 mA for the negative rail for these critical areas.  For the shunt LEDs I’m using 5 rows of 7 2.1V 100mA for the +14.7V rail and two rows of the same for the -14.7V rail.  In keeping with overspecifying everything, the capacitance for the CRC filter is massive, as will be seen during the build.  Once built I’ll post oscilloscope images of all rails and they should be more than smooth enough to currents far exceeding operational requirements.

 

Other than that there’s a standby cct for the trigger input and that’s about it.  Quite a simple design done to extremes and very throughly.

 

Due to the brightness from the LEDs inside the case I’ve called it Lightus Deus - a play on words to the inside of the amp being ‘light as day’ and the translation from Latin as ‘Lightning God’, which will suit this amp well.

 

So far I’ve bought a pretty standard enclosure from China, seeing as this will be a one-off, and designed and has printed the PCBs.  I’m now waiting for deliveries from Mouser and RS, as well as on the unobtainium Toshiba JFETs.  First tests passes: everything fits as it should!

 

 

[Guest]

I’ve been watching 12voltvids on YouTube lately in marathon mode, this is something I’d like to do one day. 

[mwhouston]

10 hours ago, Buffle said:

I’ve had a box of Hypex parts around and thought I’d put together a one-off power amp build for personal use based on NC500 modules.  This will be a stereo power amp capable of very high performance in terms of sense of resolution and speed, load invariance, and prodigious power delivery.  Fully dimensioned sound.  Given that the object is to use up some of my Hypex parts, that has served as the starting point. 

 

These modules only contain the drive stage, so require a voltage gain stage capable of driving the 1800 ohm load of the drive module.  It is this predrive stage that will determine the character of the sound, with the nCore drive stages capable of delivering accurately on what they are fed with.  The usual approach is to use op amps, whether integrated or discrete, but that would take away most of the design freedom.  I’ve designed two predrive boards: one using JFETs into MOSFETs in a similar but modified cct used by Nelson Pass, and another using JFETs into step-up transformers into more JFETs.  I’m building the former first, for no particular reason, and will build and test the transformer-based board later.

 

 The predrive stage and NC500 drive stage will be all about precision and lightning fast transients, and this power supply is built around that goal.  I’m using a Hypex SMPS for the drive stage.  I’ve tested these thoroughly and they’re very low noise and capable of delivering more current than I’ll be using.  For the predrive stage and drive module Vaux and Vdr I’m using a linear power supply with CRC filtering for the main rails with shunt regulation for the more critical stages.  As I’ve got current to spare I’m using basic but excellent LED shunt regulation applied to the already CRC filtered rails.  This is a brute force method but will deliver excellent low-noise performance.  I need around 100 mA for the positive rail and 30 mA for the negative rail for these critical areas.  For the shunt LEDs I’m using 5 rows of 7 2.1V 100mA for the +14.7V rail and two rows of the same for the -14.7V rail.  In keeping with overspecifying everything, the capacitance for the CRC filter is massive, as will be seen during the build.  Once built I’ll post oscilloscope images of all rails and they should be more than smooth enough to currents far exceeding operational requirements.

 

Other than that there’s a standby cct for the trigger input and that’s about it.  Quite a simple design done to extremes and very throughly.

 

Due to the brightness from the LEDs inside the case I’ve called it Lightus Deus - a play on words to the inside of the amp being ‘light as day’ and the translation from Latin as ‘Lightning God’, which will suit this amp well.

 

So far I’ve bought a pretty standard enclosure from China, seeing as this will be a one-off, and designed and has printed the PCBs.  I’m now waiting for deliveries from Mouser and RS, as well as on the unobtainium Toshiba JFETs.  First tests passes: everything fits as it should!

 

 

As some one who has built about ten or so Class Ds I know what your are talking about when you say fast transients, resolution and speed. Mostly I have found this with the more powerfull modules - 150W and up. In all my builds though I have used DC linear PSs and never SMPSs. Most of my amp modules have been Tripath chip based also. I have had great professional and amature comments about my amps and feel the secret is linear supplies.

 

Your  build is to be far more complex than any of mine where mine are really using the amp modules supplied, building a simple PS and enclosing it all in a premade case. Most modules from hifimediy. Currently I'm using a pair of monoblocks built around the hifimediy T3 modules about 150W each. 

 

These drive Altec horns with FaitalPro Italian compression drivers, a 12" woofer and now with Fostex T90A super tweeters. I couldn't be more happier with the sound. The speakers work out to be 92db effcient and with a ton of power on tap, dynamics, bass and sheer mid-range clarity are off the charts. 

 

Just built a 180W Tripath with four inputs and step attenuator for another SNA member based on the T4 modules but setup for 4ohm speakers. I had a chance to spend a week with the amp before shipping it off. Very impressive and very inexpensive when you consider performance and power. I have two more T4 modules on the way to try something different with.

 

Look forward to final images of your build and listening comments. 

[Buffle]

Deliveries from RS and Mouser this morning, and the predrive board is populated with the 36 pin headers for the nCore modules, test points, heat sinks, MOSFETs, caps, resistors, and pots.  Now just waiting on the rare JFETs - quite a few of them so I can find matched IDSS pairs.

 

As amusing as it was to drop in the 300 VA toroidal for the predrive and drive aux rails in the last pic, overkill has its limits.  A nice 100 VA unit has more than enough current capability for this task and so will go in.  A bit of internal jewellery in the form of a water jet cut, anodised, and laser engraved 6061 aluminium cap for the toroidal clamp so that the looks meet the performance.

 

i have a spool of 99.9% pure silver 1 mm solid wire and some silk tubing, so will use that for a lot of the internal wiring, because why not?  A trivial but nice touch.

 

 

[Buffle]

After populating the PSU and drilling in the many standoffs, it’s starting to take form.  The JFETs are scheduled to arrive Wednesday, so if all goes well it’ll start getting power on the bench Wednesday night.

 

Tonight I’ll get the wiring done and main filter caps in.

[Gridmon]

@Buffle that looks superb. I’ve read so much about the differences ncore guys have heard with various discreet op amps. I’m fascinated to see how your JFETs will go.

[Guest]

It’s exciting stuff actually, I must get into this. I’ve got the time. 

[Buffle]

Power supply tests and it’s already living up to its name.  The photos don’t look too bad, but looking down into it is like staring into the fires of hell.  I’m sure the dark spot in my vision will be gone some time soon.

 

 

Edited January 21, 2019 by Buffle

[Buffle]

3 hours ago, Sime V2 said:

It’s exciting stuff actually, I must get into this. I’ve got the time. 

MoQ for the PCBs means I have some available.  Happy to share the BoM and schematics.  I can even assist on the Hypex side.  I haven’t kept track, but I’d estimate total parts cost would be around the 2.5k mark - assuming you can source the Toshiba or Linear Systems rerelease JFETs.

 

I’d call it intermediate level for DIY audio.  Nothing overly complicated other than adjustment, for which a FFT analyser would be best but oscilloscope would suffice.  If that all sounds a bit much then the other predrive board I’ve designed doesn’t even require adjustment.  I think they’ll both perform superbly based on the sims of the first one and elegance of the one I’ve yet to build, but that one will be a higher build cost due to the price of the high performance transformers and larger number of unobtainium JFETs that’ll require matching.

 

If anyone is interested, feel free to PM me.

[Gridmon]

O'god my fingers twitch'n. But then I remember I'm a software guy with no time, who will burn himself... still.

 

[Buffle]

I managed to squeeze in a couple more hours on the amp this morning and have just now, following sine wave tests and tuning, played the first music through it (Beethoven’s Sonata No.14 followed by Kidswaste Underwater.  Still a lot more testing to do, as I couldn’t resist stopping to do some early listening, but so far very compelling.  A very vivid sense of depth and superb detail without sounding clinical.

[Mendes]

13 minutes ago, Buffle said:

I managed to squeeze in a couple more hours on the amp this morning and have just now, following sine wave tests and tuning, played the first music through it (Beethoven’s Sonata No.14 followed by Kidswaste Underwater.  Still a lot more testing to do, as I couldn’t resist stopping to do some early listening, but so far very compelling.  A very vivid sense of depth and superb detail without sounding clinical.

How about some Ben Webster live at The Renaissance.........?

 

 

[Buffle]

14 minutes ago, Mendes said:

How about some Ben Webster live at The Renaissance.........?

 

 

Oooh, I forgot about that one!  I’ve packed up my workbench for tonight and put the amp away until I can work on it again on the weekend, but once I’ve finished it I’ll bring out Ben Webster for sure!

 

I’ve been rotating a few amps and a few speakers through my main system lately, both my own and others’ designs.  Most recently this includes bringing in some YG Carmels until I get time next year to build my ultimate speakers - the YGs are very good.  Out of the other brands and my own class A solid state and tube amps, I’ve a feeling this new amp will be taking up permanent residence.  It’s turning out to be superb.

[Mendes]

1 minute ago, Buffle said:

Oooh, I forgot about that one!  I’ve packed up my workbench for tonight and put the amp away until I can work on it again on the weekend, but once I’ve finished it I’ll bring out Ben Webster for sure!

 

I’ve been rotating a few amps and a few speakers through my main system lately, both my own and others’ designs.  Most recently this includes bringing in some YG Carmels until I get time next year to build my ultimate speakers - the YGs are very good.  Out of the other brands and my own class A solid state and tube amps, I’ve a feeling this new amp will be taking up permanent residence.  It’s turning out to be superb.

Very exciting, I am staying tuned.

 

How is the 211 SET going?

 

We must catch up, I still get lots of love for those wonderful speakers....

[Buffle]

11 minutes ago, Mendes said:

Very exciting, I am staying tuned.

 

How is the 211 SET going?

 

We must catch up, I still get lots of love for those wonderful speakers....

I must admit that I’ve not been doing much with tubes lately and haven’t finished the 211 SET.  I’ve been amusing myself with SS class A ideas.

 

We must catch up for sure.  I was recently down South in Manjimup and picked up some rather special reds.  There’s a particular 2000 Shiraz that you might enjoy.  I just need to find the time, as my consulting has me booked heavily until at least the end of 2019, but a night of good music and good wine is just what I need to remain sane  

[Buffle]

Due to travel for work commitments it looks like I won’t be getting to finish this until the later half of Feb, but I did get a little bit more done last night.  Four 22,000 uF capacitors fitted for the predrive rails.  That’s now just under 100,000 uF in capacitance for the predrive stage.  A bit over the top, but I had four nice 22,000 uF 40 V Kemet caps on hand just begging to be used for something like this.  

 

More res pics soon, but at the moment I’ve got test leads everywhere to ensure everything is performing as expected, so the amplifier is having a bit of a “does my butt look big in this?” moment.

 

All in all only 1 error, which isn’t bad considering I was away when I did the design and created the PCB artwork directly (rather than from schematic and using a parts library) over a couple of nights and, because I was away for work, not using my usual design software but free software downloaded to my laptop (Designspark PCB).  The artwork for a surface mount bridge rectifier for the standby cct was rotated 90 degrees out, meaning the component could only just be made to fit.  I might get another PSU PCB printed, but for now it’s fine as-is.

[Buffle]

I finally had time to finish this.  I couldn’t resist the urge to take things that little bit further and redesigned the PSU to still use LEDs for voltage reference, but via active BJTs for shunting.  The rails to the predrive stages are now so clean that noise is below the measurement threshold of my instruments.

 

The design is modular, allowing me to try alternate predrive designs, but given how this design is performing I simply don’t have the current inclination to do so.  I’m considering publishing some DIY designs at some stage, so may do so then to allow people to pick and choose.

 

Its crystal clear; detailed yet effortlessly musical and with more power on tap than I need even for highly dynamic material.  Always nice to hear a design in which the subjective outcome aligns with the objective engineering.

 

 

 

 

Edited February 26, 2019 by Buffle

[Buffle]

I had the top off just now to do final adjustments, so took a few more photos.  Glad that I did, as in one of them I noticed that I'd not yet screwed the bottom standoffs down on the HV PSU!  I have also to tidy up the wiring sleeves, which I had pulled back to attach probes.  Oh well.

 

The front is probably a bit over the top, but I needed to cap off the screws holding the front plate on with something and had some branded gimbals in boxes, so on they went.

 

Given the modular nature of the PSU, power distribution and final filter cap PCB, and predrive PCB it's a little more 'wired' than I'd normally prefer, but at least all of the critical path wiring is in solid silver with cotton sleeving (except the signal in wiring, which is silk covered), so has some bling cred.  In any case, it's dead quiet due to very careful trace layout and the very over-the-top PSU.  

 

Although I've not tried it, I'm confident that I could plug it into a distribution board also running an angle grinder and still have no noise - 'audiophile' power leads simply couldn't make a difference, as PSU rail noise is undetectable on my oscilloscope on the critical rails.  CRCRC with very quiet LED-referenced shunt regulation with well over-spec'd capacitance on the distribution PCB.

 

A keen eye would notice that a couple of the resistors on the PSU have been snipped.  They were there for initial testing only in order to measure that the PSU would be up to supplying higher currents for other predrive PCBs, but would result only in excess heat for the currents required I this implementation.  At some stage I may remove them simply for neatness, or otherwise leave them in place for testing future predrive designs.

 

In terms of THD, whilst measured I don't believe that, assuming it is adequately low, this tells much of how an amplifier will sound subjectively - it is more the relative proportions with increasing harmonic multiples that will do so, as well as the distortion remaining low at very low levels.  What I will say is that in the predrive design and by biasing the MOSFETs in the predrive stage rather high the profile is as intended and the sound superb - all without NFB in the predrive stage (other than some inherent degenerative feedback).  I could get the distortion even lower than it already is through adjustment, but at the expense of the best profile.  Even still, the THD is low.

 

At some stage I'll take some measurements to dispel some myths around SMPSs and hash.  Well, kind of myths.  More that with correct design as per the Hypex unit used 'hash' need not be a problem.  I've taken these before with these modules, but didn't capture images.  With no problematic hash injection into the ac supply side and clean DC lines to in excess of the required current there simply isn't a reason for a conventional design to be able to perform any better, and without good regulation or PSRR they can perform worse (which is not to say that conventional supplies cannot be as well - they can and this is demonstrated in the toroidal and shunt regulation used for the more critical rails).  Probably the only downside is that SMPSs can require higher/specific expertise to repair, which is not a concern for me.  In any case, it would be a simple matter to replace this supply with an equivalent output conventional or switch-mode unit - it needs only supply between +/- 65 to 90 volts (depending on required power output) and plenty of current.

 

At some point I will fully document the design sufficient for DIYers to build their own - although they will need to source the Hypex NC500 OEM modules, SMPS, and unobtainium JFETs.  I do have a few PCBs handy due to the MOQ.

 

I've also designed a subwoofer amp to match it.  Not strictly a subwoofer amp, as it will perform across the full audio bandwidth, but basically a simplified (single PCB) version of this amp with a miniDSP PCB built in for adjustment purposes.  Given that I named the current amp Lightus Deus, which is latin for Lightning God, I've named the other amplifier Tonabit Deus - Thunder God.  Just as 'Lightus Deus' made a nice play on words with 'Light as Day', I kinda thing Tonabit Deus works along the lines of Tone and Bit, given the DSP adjustment built in.  I'm currently running my four subwoofers with the prototype of the Tonabit Deus, so will document the build of the final version.

 

 

Edited February 26, 2019 by Buffle

[mwhouston]

On 22/01/2019 at 12:28 AM, Buffle said:

MoQ for the PCBs means I have some available.  Happy to share the BoM and schematics.  I can even assist on the Hypex side.  I haven’t kept track, but I’d estimate total parts cost would be around the 2.5k mark - assuming you can source the Toshiba or Linear Systems rerelease JFETs.

 

I’d call it intermediate level for DIY audio.  Nothing overly complicated other than adjustment, for which a FFT analyser would be best but oscilloscope would suffice.  If that all sounds a bit much then the other predrive board I’ve designed doesn’t even require adjustment.  I think they’ll both perform superbly based on the sims of the first one and elegance of the one I’ve yet to build, but that one will be a higher build cost due to the price of the high performance transformers and larger number of unobtainium JFETs that’ll require matching.

 

If anyone is interested, feel free to PM me.

He is a 180W Tripath amp I finished a few days back. Brd. is hifimediy T4 v1.3 and a 500VAC 30+30 Tranni. Can’t get any easier than this. Sounds absolutely stunning with real deep bass. 

 

About to do another but going to do a few upgrades to the brd. first.

[Buffle]

42 minutes ago, mwhouston said:

He is a 180W Tripath amp I finished a few days back. Brd. is hifimediy T4 v1.3 and a 500VAC 30+30 Tranni. Can’t get any easier than this. Sounds absolutely stunning with real deep bass. 

 

About to do another but going to do a few upgrades to the brd. first.

If you're interested in taking a leap into a fair less easy but very worthwhile project, let me know.  In this case (Lightus Deus) the class D stage is providing mostly the current output, with a small amount of voltage gain, with the predrive circuit providing the input buffering and lion's share of the voltage gain.  It's very much into the high end in terms of design and performance.

 

While both the Tripath modules and the nCore output stages are class D, that's about where the similarity ends.  Of course, what I can't speak to is which you'd prefer, but the topology - other than being class D, is markedly different.  As mentioned, these nCore output stages are merely that, requiring a dedicated input/VAS.  Most (nearly all) of the designs on the market incorporating these output stages use opamps for VAS with these stages, the reason for which is a mystery to me other than to say that they are easy to design and implement and offer very low measured distortion.  The intention for this particular amplifier was to apply my expertise and experience to create what I believe to be the ultimate achievable without requiring large numbers of unobtainium parts and thereby making the design essentially impossible to replicate.

 

My only reservation in making the design available to others ties back to my initial post.  I don't have a particularly large ego in designing non-commercial circuits and in this case saw an opportunity to draw from a design by Nelson Pass, being his BA3 preamp design.  It allowed me to capture much of what I'd originally sought out to design without requiring the aforementioned large number of hard to find parts that would have required ordering in large volume in order to adequately match.  Pass' design is an elegant approach that, with some modification, incorporated the design elements that I wanted with only one pair of these parts per channel.  My circuit is therefore an evolution of that design, rather than purely original (not that such a thing frequently exists in any product in any case).  For this reason I would only provide the PCBs to those wanting to use them for DIY and not for product sale.  Of course, there is also quite a bit of original work in my design too.

 

A word of warning, though.  While the PCBs are not expensive the overall build is not cheap.  This is a high-end design and has a BoM to match.  Being DIY it will still be far cheaper than buying a manufactured and marketed product, but expect a good $3k (at least) in parts by the time it's finished.  As far as performance is concerned, I'm fortunate to be in a position to have rather a few amplifiers go through my system, both reasonably priced and hugely expensive and both my own and others' designs/products.  This one will be staying firmly put as my reference.  

[mwhouston]

23 hours ago, Buffle said:

If you're interested in taking a leap into a fair less easy but very worthwhile project, let me know.  In this case (Lightus Deus) the class D stage is providing mostly the current output, with a small amount of voltage gain, with the predrive circuit providing the input buffering and lion's share of the voltage gain.  It's very much into the high end in terms of design and performance.

 

While both the Tripath modules and the nCore output stages are class D, that's about where the similarity ends.  Of course, what I can't speak to is which you'd prefer, but the topology - other than being class D, is markedly different.  As mentioned, these nCore output stages are merely that, requiring a dedicated input/VAS.  Most (nearly all) of the designs on the market incorporating these output stages use opamps for VAS with these stages, the reason for which is a mystery to me other than to say that they are easy to design and implement and offer very low measured distortion.  The intention for this particular amplifier was to apply my expertise and experience to create what I believe to be the ultimate achievable without requiring large numbers of unobtainium parts and thereby making the design essentially impossible to replicate.

 

My only reservation in making the design available to others ties back to my initial post.  I don't have a particularly large ego in designing non-commercial circuits and in this case saw an opportunity to draw from a design by Nelson Pass, being his BA3 preamp design.  It allowed me to capture much of what I'd originally sought out to design without requiring the aforementioned large number of hard to find parts that would have required ordering in large volume in order to adequately match.  Pass' design is an elegant approach that, with some modification, incorporated the design elements that I wanted with only one pair of these parts per channel.  My circuit is therefore an evolution of that design, rather than purely original (not that such a thing frequently exists in any product in any case).  For this reason I would only provide the PCBs to those wanting to use them for DIY and not for product sale.  Of course, there is also quite a bit of original work in my design too.

 

A word of warning, though.  While the PCBs are not expensive the overall build is not cheap.  This is a high-end design and has a BoM to match.  Being DIY it will still be far cheaper than buying a manufactured and marketed product, but expect a good $3k (at least) in parts by the time it's finished.  As far as performance is concerned, I'm fortunate to be in a position to have rather a few amplifiers go through my system, both reasonably priced and hugely expensive and both my own and others' designs/products.  This one will be staying firmly put as my reference.  

It’s true you can listen to 10 class D amps and they will all sound different. Clock stability and more particularly where and how they take their NF really influences the sound.  That’s where Tripath made inroads. I understand they take one NF point in the switching domain where others stayed with the analogue domain. 

 

Not it sure both have felt the Tripath modules have a little more warmth than others.  

[Buffle]

38 minutes ago, mwhouston said:

It’s true you can listen to 10 class D amps and they will all sound different. Clock stability and more particularly where and how they take their NF really influences the sound.  That’s where Tripath made inroads. I understand they take one NF point in the switching domain where others stayed with the analogue domain. 

 

Not it sure both have felt the Tripath modules have a little more warmth than others.  

That's one difference, for sure.  In the case of Hypex' nCore (and also UcD), a clock as such isn't implemented.  Instead, they are self-oscillating.  A particular strength of Hypex nCore (and, again, UcD) is that the feedback loop is taken from after the output filter, meaning that they are load invariant.  This or the use of a higher switching frequency are, imo, the dominant approaches currently being used in terms of providing truly high-end performance - although I'd add that they are necessary (in comparison with existing solutions) but not in themselves sufficient for high performance across a wide range of speaker pairings.  As with all topologies, it is the overall design that influences that.  An example would be that things like the self-oscillation were improved between the UcD and nCore Hypex technologies, with the outcome quite an improvement in performance and nCore being truly high-end.

[davewantsmoore]

Quote

feedback loop is taken from after the output filter, meaning that they are load invariant.  This or the use of a higher switching frequency are, imo, the dominant approaches currently being used in terms of providing truly high-end performance - although I'd add that they are necessary (in comparison with existing solutions) but not in themselves sufficient for high performance

There's a thread somewhere here, where it's repeated endlessly that the only route to acceptable performance is sky-high switching frequencies.   Where were you?!?!?      The irony is that those same poster(s) categorically disapprove of high-speed DA conversion.... and don't seem to see the parallels.

 

 

I haven't read every post in this thread .... can you show the difference in performance between your design and the VAS in the Hypex nCore OEM designs?    It sounds superior?!

[Buffle]

3 hours ago, davewantsmoore said:

There's a thread somewhere here, where it's repeated endlessly that the only route to acceptable performance is sky-high switching frequencies.   Where were you?!?!?      The irony is that those same poster(s) categorically disapprove of high-speed DA conversion.... and don't seem to see the parallels.

 

 

I haven't read every post in this thread .... can you show the difference in performance between your design and the VAS in the Hypex nCore OEM designs?    It sounds superior?!

I do recall one or two conversations on here regarding class D and switching frequencies in which some inaccurate myths were being perpetuated, but one that I did provide input to did not seem to go anywhere meaningful and was removed.

 

In regard to the nCore modules, the NC500 (OEM only) module only has modest voltage gain of a little over 12 dB.  It requires an external input drive stage with sufficient voltage gain and drive current for the module's 1.8 kohm input impedance.  Many of the commercial offerings use an op amp for this purpose, whether discrete or integrated, which works quite well and is of course very easy for a lot of people to implement without requiring a huge amount of expertise.  It's a reasonable approach and I'm not dismissing it.  For me, however, this doesn't sit with my ideals in that it is using an all purpose tool when a design specific for the purpose could be design and implemented.  That is, a circuit designed to achieve a singular goal and with just the required amount of gain.   

 

The design I've used treats the nCore output stage much the same as one would a set of output transistors - albeit ones requiring less voltage gain and more current to drive.  In essence it is not complicated at all: a simple JFET pair input driving MOSFETS in complimentary class A configuration.  This is supplied by a power supply designed for theoretically optimum transient response, being shunt regulators.  With CRCRC, LED-referenced shunt regulation, and very overspecified capacitance the result is very, very low noise rails.  The same PSU design is used for the sensitive NC500 15V +/- rails.  It's a somewhat simple design, implemented very thoroughly.

 

I got a bit carried away in my initial design for the pre-drive stage, quickly realising that it was going to prohibitive to build.  As I started considering more reasonable constraints it occurred to me that what I was arriving at was similar to Nelson Pass' BA3 preamp, so I set aside my own design and decided to look at instead adapting his circuit to the task.  The adaptation required wasn't huge, which is why I consider my design an evolution rather then revolution.  I'll later test another couple of predrive designs that I came up with which would more accurately be labelled as the latter, hence the modular build style.  These other designs will better satisfy me ego, given that they are original, but I'm finding the performance of this design so remarkable that I'll likely put this off.

 

Is it 'better' than the discrete and opamp variations used by Hypex or others?  I'm a bit biased here, but will say that the result is truly amongst the best amplifiers I've heard.  Crystalline, three-dimensional transparency,  seemingly unlimited drive power, and yet not etched or clinical.  While often said, combined with the YG Carmel speakers I currently have in my room the sound truly approaches an open window to music - just effortlessly transparent and with lifelike cymbal splashes and djembe drums sounding just how I've heard them so many times in real life in South Africa. 

 

In comparison, the opamp versions I've heard sounded comparatively two-dimensional, seeming detailed and yet somehow relatively two-dimensional.  What I cannot explain is why, given the measured performance of the op amps typically used.  As an engineer, everything tells me that the performance should be perceptually near identical, and yet the difference is large.  However, also as an engineer I've applied the sum of my current knowledge in designing based on previous experience, inclinations, and testing over the years in designing it for what I expect should be superior performance.  As both an engineer and a psychologist I should dismiss my subjective analysis somewhat, as I haven't done DB tests, which I value and which would support a difference and that one is clearly preferred.  Let's just say that I am as confident as I can be that the difference would be readily picked and the result highly statistically significant over a large number of samples.

Edited February 27, 2019 by Buffle

[Buffle]

BTW, I've also built and tested rather a few of my own designs over the years that I expected would perform well subjectively but just didn't, so I'm not too biased.  Experience is based at least as much around discovering what doesn't work well as what does.