TA-N88 class D amp: replace V-FETs with MOSFETS

The link I provided was directly to the LittleDiode web site, not to a eBay listing. I would suspect that the picture they have used on their web site listing would be a generic TO3 picture, not a picture of the actual component itself.

Anyone who would be considering spending the amount of money that these components are being sold for, would be wise to inquire directly with any seller before placing the order, not just order them blindly with their fingers crossed that they are the real deal. - Chris
 
ChrisMarantz said:
The link I provided was directly to the LittleDiode web site, not to a eBay listing. I would suspect that the picture they have used on their web site listing would be a generic TO3 picture, not a picture of the actual component itself.

Anyone who would be considering spending the amount of money that these components are being sold for, would be wise to inquire directly with any seller before placing the order, not just order them blindly with their fingers crossed that they are the real deal. - Chris
Click to expand...

Yes, I saw the direct link, but also checked their auction page (which I didn't link because it is against the forum rules). However, on that listing they say "A brand-new, unused, unopened, undamaged item in its original packaging (where packaging is applicable)" which seems weird they would have access to a large supply of untouched NOS units but who knows...and they use the same blank TO3 package as image (but as you say, it might not be the actual item and they do say in the title it is made by Toshiba).

Very sound advice on making some inquiries about the component/picture before buying blindly.
 
Last edited:
Nevermind that TA-N88 uses J28/K82, NOT J18/K60. So stop right there. J18/K60 are half maximum current and would be just at the edge of the maximum voltage spec.
MOSFETs are driven by positive gate voltages WRT source for N-type, VFETs with negative, MOSFET maximum Vgs=+-20V typical, VFET -50V typical and in the positive direction they behave as a diode (low impedance). Without drive, VFETs are 'normally on' and look like a 1 ohm resistor S to D. MOSFETs are open (megaohms). All of this has to be taken into account (includes heavy PSU mods) along with the bane of all D class amps, cross-conduction. AND unlike MOSFETs, VFETs are temperature stable and in this particular case eminently complementary.
BTW making even a modern MOSFET in a standard case switch at 'just' 0.5MHz on a single sided PCB the size of the N88's is literally rocket science, and them making two with opposite polarity switch exactly in synch would be rocket science squared.
IF you must, start with lateral MOSFETs due to lower capacitance but it is unlikely you'll get the switching speed and/or efficiency.
Tried it, lost hundreds of hours on it, put it back the way it was and threw away a handful of dead MOSFETs and BJTs.
 
Last edited:
Unfortunately this is a topic that's been debated for hours before and many other people have reached a dead-end. This doesn't change the fact that there are a ton of amplifiers out there with irreplaceable VFETs in them that are now door-stops.

If you want more information it will have also been debated over on diyaudio.com, chances are Nelson Pass may even have offered his opinion at some point.

I ordered a VFET kit from that site that is built around original and tested 2SK82's and 2SJ28s - its a Pass designed circuit and I'll be starting the build soon. Still some kits available in the next release I hear.
 
slimecity said:
Unfortunately this is a topic that's been debated for hours before and many other people have reached a dead-end. This doesn't change the fact that there are a ton of amplifiers out there with irreplaceable VFETs in them that are now door-stops.

If you want more information it will have also been debated over on diyaudio.com, chances are Nelson Pass may even have offered his opinion at some point.

I ordered a VFET kit from that site that is built around original and tested 2SK82's and 2SJ28s - its a Pass designed circuit and I'll be starting the build soon. Still some kits available in the next release I hear.
Click to expand...

Let me know when this happens !

Nashou
 
slimecity said:
If you want more information it will have also been debated over on diyaudio.com, chances are Nelson Pass may even have offered his opinion at some point.
Click to expand...

here or on DIY forum, when it comes to this subject, take Ilimzn's opinion.....;)
 
Gents,

I had a very extensive reply from ilimzn member so he is not object against quoting. For many who lazy enough to read in full: the topic is practically impossible. In theory it can be managed, however will require something similar to the totally new amplifier design. Pls read below (thanks ilimzn):

"Well, I am sorry to have to dis-abuse you of some of your assumptions.
One is that 'switching is just switching' - nothing could be further from the truth, or we would have TA-N88 class D amp analogs in the very early 80s, using MOSFETs. yet the first viable class D full range amps (not subwoofer amps) appeared in the late 90s, and were not even close to the N88. The reason is that the VFET is still in the very top of switching devices, and in fact it's original (military grant financed) purpose, was to be a switch used in radar applications, namely to create the outgoing radar pulse. It should be noted that while the radar is emitting, it is blind, so the closer and quicker you want your radar to measure, the shorter the outgoing pulse has to be, which is of extreme importance when you are trying to detect and intercept missiles traveling over the speed of sound. At this point, one might think how a triode characteristic fits into this, and the answer is that while it is largely incidental, it help greatly with turn on and turn off transients, which is why the TA-N88 can operate at 0.5MHz being a completely discrete construction and with no special measures to counter cross-conduction. This sort of thing is at the very limit of what can be done with today's MOSFETs, even the latest generation - most modern class D amps operate between 200 and 400kHz and use special driver chips to drive discrete MOSFETs, often custom made, which include various compensations for voltages and temperature - one more plus for the VFET, it is extremely stable with temperature. N88 comparable operation is still only possible inside integrated chips or very special constructions such as some of Sony's S-master series amps. To make a long story short, the reason why conversion is extremely difficult mostly comes down to the VFETs input capacitance which is at least an order of magnitude lower than most MOSFETs. It is simply impossible to find a MOSFET with a breakdown voltage of over 200V, and 10A drain current with Cgs of 150pF. At best you will get about 3x that using SiC MOSFETs for silly money, regular ones will get you in the 5-8x ballpark. Katerals will be around 3-4x the capacitance but too slow and too high Rdson.
So, in naswer to your question: it IS possible to do a conversion but it comes down to essentially building a new amp. The bias power supply in the PLPS has to be completely rewired for lower voltage and opposite polarity, and the driver circuits must be carefully rebuilt, with adjustments for actual MOSFETs used. I attempted such a thing 10 years ago and could never get it sufficiently reliable or high quality to make it a viable project and ended up relegating that unit (which came with dead VFETs) to a parts unit, after hundreds of hours spent on it. This also implies really good measuring equipment - without some sort of storage scope which can capture real time events in the 100MHz realm, AND without really good low capacitance probes, you are dead in the water.
So what were the biggest problems? First, the aforementioned MOSFET gate capacitances. This is somewhat mitigated by the fact that less voltage is needed (about 15V) but the switch on-off behavior is so nonlinear due to stored charges, it is very difficult to set up the drivers to prevent cross conduction if you want good quality. I used IRFP340 and IRFP9240 (2 pairs each) to get approximately the same behavior, but this proved in fact to be impossible with existing circuitry topology because unlike the Sony VFETs, which are almost perfect complementaries, the MOSFETs are not even close. The transient behavior as they go from on to off and back, is different and non-complementary, and further, depends on temperature, which becomes a mechanism of destruction if cross-conduction happens. For a class D amp it is IMPERATIVE not only for performance but for reliability that the complementary sides switch on and off precisely in synch. If one side is not switched off when the other goes on, cross-conduction happens and while both are on the power supply is shorted through the output stage. If one side switches off and the other has a delay before switching on, a 'gap' occurs, during which the output filter is undamped and will produce resonant voltages far in excess of normal. The same happens when cross-conduction stops but to a lesser degree. I am not even going to mention how many MOSFETs were killed due to that. So I learned the hard way why most modern class D amps use only N-ch MOSFETs, it is the only way to get some control over this problem. This is made worse by the fact that everyone using only N-ch has now produced 4 generations of steadily improving N-ch MOSFETs while P-ch MOSFETs became extremely rare and technologically stuck in the past. My last ditch attempt at a conversion was using EXICON 10N25 and 10P25 lateral MOSFETs which have a lower input capacitance (Still, two pairs get to 3-4x the original 2 VFET pairs) and they simply proved to be too slow. They are also not completely complementary but more so that standard HEXFETs, and they have a different pinout (fortunately this proved to be an asset to some degree), and have a negative tempco, which resulted in less cross conduction as temperature rises. And it DOES rise as Rdson (effective) is quite a bit more than with the VFETs (TA-N88 uses some tricks here). The net result is distortion, somewhat like under-biasing a class AB amp. The best I could get to is around 200kHz with this setup and that just didn't sound good no matter what (and no surprise really)."
 
"My advice would be to sell the N88 if you don't like it. Even amongst modern class D amps, it is a huge achievement, far, far before it's time.
Some J18/K60, J28/K82, and even NEC J20/K70 are available from China, the usual Buyer Beware caveat applies. A group buy could help here, also someone who owns a curve tracer.
Also, there are a number of J28 (P-ch) types available from a reputable seller, and there is no reason not to build an amp only with P-types, in fact using Nelsons transformer coupled ideas, it is not that difficult, and someone should nudge him a bit toward that path. Also, any single ended design, such as the SIT amp can easily be transposed from an N-type part to P-type by using reverse polarity on all power supplies and polarized caps. It would be the pinnacle of lazy to pull out N-ch VFETs out of working amplifiers only to avoid a bit of thinking on how to use the ones readily available.

Just think about this: the completely discrete circuits in the N88 switch over from the + to - power supply (so a total 160V span) in only 80ns, and that is at full power (With substantial current). This is the equivalent of 2kV/us slew rate, with very little jitter (this would be heard as noise in the output), and with a current slew rate of 82A/us. Getting this done with regular MOSFETs is a BIG problem, even though in theory the MOSFET is as fast as you can drive it. But at this sort of current change, the inductance of the pins themselves can produce spikes larger than the MOSFET driving voltages.
VFETs are still produced for the military (at extreme prices :( ) and now are being migrated to SiC, as it seems that topology is the simplest to make on SiC. MOSFETs and IGBTs have pushed VFETs out of all but the most demanding power switching applications, where extreme speed and voltages are needed, as VFETs are also capable as a semiconductor structure, of voltages of 2-4kV. Some IGBTs can go to about 3kV, and MOSFETs are limited to around 1.2-1.6kV. Before MOSFETs and IGBTs were capable of over 1kV voltages, VFETs (SITs) were used in those applications, mostly made by TOKIN, however, production has been discontinued about 10 years ago. As far as I know, the only commercial SITs produced are made in the ex-USSR, but this is for kV and kA apps like motor regulation on electric locomotives."
 
I can see J20/K70 for sale at Ebay, not very expensive. Looks promising. However, the only spec-sheet I found states Vds max = 100V, while J28/J70 is 170V. Rails in TA-N88 are +/-78V. Just 32V reserve - is it OK?
 
If you want to fix a TA-N88, there is actually no other VFET you can use except for J28/K82, If you have a working N88, keep yjose VFETs in it.
However, you could use J20/K70 for a more classical construction of an audio amp. These constructions, including ones by Nelson Pass and others on the diyaudio.com forum can use less specifically targeted VFETs than the J28/K82.

It should be noted that there are two very different constructions of J28/K82, although to my knowledge, N88 power amps only use the second one, which is optimized for that use.
The original J28/K82 were designed as 'larger', i.e. higher current and voltage J18/K60 successors. These were presented by Sony at a JAES convention for use in class AB output stages of power amplifiers. The JAES papers even show a micro-photograph of the silicon die, approximately 4x4mm in size. These were produced by the same process used for the J18/K60, and were graded using the same rank system as the J18/K60.
That being said, every N88 I have seen uses rank 3 VFETs (usually the complete code is xx33 where xx is a pair of letters coding the batch number or production series). These are internally a double-die component, there are actually two smaller (about 3x3mm) VFET chips connected in parallel. At first glance they look like J18/K60 chips but they are actually different (the die attachment points are not the same, for example). These were optimized for work in the N88 and the production of the older single die was stopped somewhere in the early stages of N88 commercialization, as fra as I know, they were never used in production or sold except as samples. One area of optimization is that they can take far higher voltages than the original spec shows, over 200V.

The Sony J28/K82 VFETs have one characteristic which is useful for simple transformer coupled input amps, and that is low capacitance. However, they are not the best at linearity at low voltages and high currents - because they do not work in linear mode under those conditions. They also have certain characteristics which make them good in complementary output stages but NOT at any current or voltage. There is an 'ideal' range where their non-linearity cancels out between the P and N part, but it is far outside what the simple constructions offer.

@arky18 - in complementary output stages (including class D), each transistor must be able to withstand the complete span from -V to +V power supply, in the case of the N88, at least 156V, under all conditions. In class D, every transistor always sees the full span in every cycle of the modulating PWM waveform by definition - as a switch, either it is closed to and connects the common output to one power supply (so voltage across it is very small), or it is open, when it sees the full span because the other side is closed, connecting the common output to the other power supply. For class AB stages this sort of situation only happens at clipping (full power).
 
Just reading this interesting thread and I started wondering how do you make a basic check if v-fet transistor is bad or good.
 
miromustang said:
how do you make a basic check if v-fet transistor is bad or good.
Click to expand...

Use a multimeter that has a diode test function. You should measure a junction voltage between GS and GD. Next switch the meter to ohms. You should measure a low resistance, 2 ohms or less, between D and S; the resistance reading will depend on the V-FET you're looking at.
 
Vint Age said:
Use a multimeter that has a diode test function. You should measure a junction voltage between GS and GD. Next switch the meter to ohms. You should measure a low resistance, 2 ohms or less, between D and S; the resistance reading will depend on the V-FET you're looking at.
Click to expand...
Thank you very much. Just got a TA-N9B and one channel was dead but it seems that speaker relay is intermittent. It is time to get it sorted out before something bad happens.
 
ta-n9 are 400w mono beasts and there's no "b" version - just rack mount 3U huge black boxes. Have two waiting for restoration... I suppose miromustang mistyped a TA-N7B unit.
 
ilimzn said:
These constructions, including ones by Nelson Pass and others on the diyaudio.com forum can use less specifically targeted VFETs than the J28/K82.
Click to expand...

For one I'd highly recommend the "L'amp" build over on diyaudio - an embarrassingly simple circuit running a VFET per channel (I'm using 2 x Sony 2SJ28) - runs in Class A mode, maybe 10 watts per channel....best amp I've ever heard, so much so that I'm selling my previous favourite, Pioneer M22.

One criticism of attempting to rebuild the old Sony VFET amps is that their circuitry is somewhat dated or complicated (Im not totally clear on this point, but it makes sense), so you are better off doing a new proven build, that is simple if possible, and that way you have less "clutter" getting in the way of that lovely VFET sound.
 
You must log in or register to reply here.
Back
Top Bottom