Anyone good at fault finding with amplifiers here?

Hi Adrian,

Unfortunately that is a common misconception, and under many conditions, quite untrue...

Here is a small excerpt from a very good paper in the supplied link:

With the push for faster switching, lower on resistance power MOSFETs, came an unintended consequence similar to, but not seen since the prime of the bipolar transistor, which was the secondary voltage breakdown effect. While MOSFETs are in the charge-carrier dominated region (low Vgs) the MOSFET allows more current to flow as the temperature increases causing a thermal runaway. It was discovered that the SOA curves given by the manufacturers were lacking in giving the region of thermal instability. A review of papers from the automotive industry is described, and recommendations to add the area of thermal instability are included. The four factors that are important in determining the thermal instability are:.

etc, etc....

Link - Mosfet Thermal Runaway

Even the venerable Douglas Self points out the existence of the problem..

FET output devices are not the gift to ears that AudioPhiles seem to believe they are...

They exhibit far greater distortion compared to their bipolar brethren ( anywhere between 10 and 20 times more..), and require much higher bias/standing current to get even close - for a 50 watt single pair output stage it is not uncommon to see bias currents in excess of 150mA to 180mA. Some impressively complex 'correction' circuits have been designed by the Passionate, to try correct this problem, making the resultant amplifier excessively complex. ( If interested search for Error Correction Output Power Stage is based on Bob Cordell’s design, which in turn is based work by prof. Hawksword)

However, my comment on the lack of Source resistors what not aimed at the issue of potential thermal runaway at all..The issue is:

The Gate threshhold voltage variation is embarrassingly wide, and varies even between so called 'matched pairs' , (Also tends to vary with temperature as well, making one chase the 'ideal' operating point up and down the street.) So, it is essential to match ALL fets in the output stage as closely as possible in gate threshold voltage AND transconductance to ensure minimum distortion. But a 'good' match is still not good enough, and the Source resistors helps to cope with that last bit of imbalance, therefore very necessary..

In most designs by credible designers and manufacturers of Audio Amplifiers, if they do a FET output version, you will find Source resistors.

In my book, the only advantage of FET's in the output, is that they more often than not fail open, protecting the speakers a little in a circuit without protection elements..

Joe

Edited By Joseph Noci 1 on 22/08/2019 10:41:25

Joe,

Thanks for that, an interesting read, my excuse is all this came out after I learnt about MOSFETs.

It does infer that the early MOSFETs were stable, but the newer high switching speed ones developed this problem. Neils amp is probably stable as it is quite old, but it could probably benefit from having its heat compounds replaced on the output transistors.

Adrian

Hi Adrian,

I have no doubt his Amp will be thermally stable - I doubt he will use it at full volume with staccato material for hours on end anyway.. The issue is really about distortion, matched pairs ( or matched QUAD in this case) and trying to balance out small transconductance deltas.

I also wonder what the differences between a current date manufacture set of FET's will be, staked up against the old pair in the amp..and how he will match them! Half decent matching should be done with the FET on a heatsink, at a reasonable temp, maybe 35-40degC, and around 1amp or so as the target current, and then selecting FETs with 'identical' gate voltages that give that current..

I suspect Neil will just ignore all this, and fit the FET's! Would be nice if he could measure the distortion before and after..

Joe

Well he will be renewing the thermal compound...

But an extra day or two to wait as I ordered TO3P mounting kits.

No source resistors - the Laney proBass (similar design) uses FOUR pairs of fets in parallel with no balancing.

My FETs are Exicon ones designed for high power audio, not switching, and won't be biased at the low gate voltages which seem to cause problems.

| I have no doubt his Amp will be thermally stable - I doubt he will use it at full volume
| with staccato material for hours on end anyway..

Err.. I thought that was the point

Neil

ECF10N20, ECF10P20, ECF20N20 & ECF20P20

www.exicon.info/products.php

MOSFET secondary breakdown is known as the Spirito effect and occurs primarily at high Vds and low Id. Most modern MOSFET dies are not a single transistor, but contain hundreds or thousands of small MOSFETs in parallel. The effect is dependent upon the actual MOSFETlas it results from changes of threshold with temperature, and how quickly the cells can get rid of heat, which depends upon the transient thermal impedance.

Andrew

P.S. those Exicon FETS appear to be the recomennded equivalent for the original 2SK135 and 2SJ50.

Helpfully the PCB has 135 and 50 written on it in felt pen which will make putting the right one in the right place easier but I AM going to trace the connections just in case... with electronics, paranoia is good.

Neil

Neil's quote re the suitability of those  FETs..

Freedom from secondary breakdown and thermal runaway make them extremely reliable and remove the need for protection circuitry.

EVERY FET spec from every manufacturer, switching FET or not, makes the same claim, and yet...

What is the gate/source voltage in your amp Neil? What is the standing Drain current? Not easy to measure I suppose.

Joe

Edited By Joseph Noci 1 on 22/08/2019 17:34:41

My radio ham friend reckons that when you run out of heat sink compound you can use copperslip. Anyone got any thoughts?

Now that's an interesting idea! Why would it conduct heat? because it's loaded with copper particles? I just got my tin of the stuff out stirred it up and stuffed the ohm meter probes in - nothing at all! not even megohms! Put some between two sheets of copper clad PCB, with strips of thin plastic (maybe 0.1mm?) interspersed to prevent the copper layers from touching, connected to the ohm meter, and pressed the copper sides together squeezing the copperslip down to a very thin layer. I hoped to force metallic particle contact - still nothing!

So, are the particles actually metallic copper? The colour certainly indicates so. Is it the lubricant that so effectively coats the metal particles that they are electrically insulated? Or is my stuff all oxidised and now an insulator? Still copper colour though.

Joe

I won't lose sleep worry about it

Had a google and found this. **LINK** Looks as tho almost anything is better than nothing, and genuine thermal grease is not electrically conducting

Yep, I know the typical thermal compounds/greases are insulators - was just wondering why copperslip is non-conductive...

Anyhow, rather than greases which are really messy when you repair a lot of amps..I have for many years been using a sheet material - the on I use is called ChoTherm - comes in various thicknesses sheets up to around 400mm x 400mm. Just cut to shape and size and fit between the component and heat sink. Heat conductivity performance tables provided with each, and they rival or surpass thermal grease performance. I have even used some that are a 'sheet' of pliable 'jelly' - does not really flow, but used to fit between tops of components and a combined heat sink / lid - cut a sheet section and place on top of the parts and fit/press the lid on - squeezing the stuff into all the crevices and spaces. It works very well, conducts heat nicely, and if you need to work on the parts, just lift the lid and peel the stuff out - it all comes out in one piece and is re-usable. Made by RayChem.

Guess I'm drifting off topic here, but seeing as Neil's amp is all fixed, I suppose we can..

Joe

Finishing up MEW 285 while waiting for Farnell to deliver my mounting kits!

I have a mono camera, very small, just webcam sized chip, that I'm planning to convert to peltier cooling. As supplied the aluminium case has a 'column' that the chip is mounted on, thermal contact is a blob of grey putty that has the appearance of being 'loaded' with something, but i know not what.

What is in thermpath type compounds? Zinc oxide, titanium dioxide or...

Neil

Neil,

There are quite a few different types. Apart from the silicon rubber type that come pre cut for the package, there are the mastic types and paste types. Each can have a different thermal particles. Ranging from Diamond dust to Zinc Oxide.

Where you dont need electrical insolation e.g. for a CPU you can just use the mastic pads or paste. The commonest OEM type is a grey mastic, that on the first few thermal cycles softens and squidges out.

The best for us mortals is Silver paste, having said that I would happily use white paste too, just not on my gaming rig where I am trying to dump 120W through 1" sq.

The secret is to have the two surfaces flat and smooth. No scratching the old stuff off with a blade. Then put it on very thinly, just a little to thick for engineers blue.

Danger of silver paste is it can conduct, so be very clean when removing old.

If in doubt Artic is a good brand to buy.

If you need electrical isolation e.g. for a TO3 package, you need to use a solid insulation i.e the rubber (or olden days mica) Depending on the type you may also need to use paste too, I have never used paste on the silicon ones.

After a few thermal cycles if you dont have a sprung loaded clip check the bolts are tight.

Adrian

I have Farnell paste, I just wondered what it was as it looks suspiciously like a thick version of nappy rash cream...

My mica washers came this afternoon.

Apparently mica is used for making optical etalons, because the two faces are (supposedly) perfectly parallel as they follow molecular cleavage plains.

Neil

All now working and tested.

Blimey! It is much louder without distortion - I suppose that just getting the extra ~8 volts of headroom before distortion sets in equates to about a 15% increase in voltage but over a 30% increase in power, but then allow for the fact that the asymmetrically driven speakers would have been adding distortion as well and the reversed speaker would have been taking away bass, it's hardly surprising it plays louder. Also I'm sure the sound is better, with a better bottom end (it deals very well with my five-string which has a low B - about 31Hz - you need good speakers to reproduce the fundamental without a lot of bass boost).

So, I'm a very happy bunny. My understanding of how these amps works is much better now, and also I've discovered the benefits of really looking below the surface - I had noticed a few months ago that it was clipping the top of the waveform but not the bottom and didn't register that this was the sign of a fault.

Thanks to everyone who contributed ideas and thoughts to this thread, I've learned a lot from you (and also Agedhorse in 'another place' who actually worked on these beasts).

Also thanks to Horovitz and Hill who explain thermal runaway in FETs in language I could understand - after the primer of this thread I could actually understand what they were on about in detail.

Neil

By far and away the best general book on real world electronics there is. I bought Edition 1 when it came out in 1980.Later some barsteward half inched it from my desk at work. The company replaced it with the then current Edition 2. I treated myself to a copy of Edition 3 a couple of years back.

Andrew

I have 2 and 3, they are quite different, I miss all the examples, especially the 'spot the deliberate mistake' tests! You have to buy a second book for them now

If only they weren't so cumbersome, or I would read it in the bath!

Neil