IRFP240 IRFP9240 design

[Alex R]

Here is the schematic of a little amplifier module I bought very cheaply indeed out of curiosity. It uses switching mosfets without ballast resistors. They sit on a solid fibreglass board, bolted to a heatsink. I am not enough of a designer to critique it. I'd appreciate your views on the design as a suitable replacement for burned-out power amps in guitar amplifiers. If it seems unstable to you, is there anything that might be done to improve it?

[KevinOConnor]

Hi Guys

The schematic is identical to the Hitachi application circuit used to promote their complimentary mosfets. Trace Eliot used the same circuit, as did H&H. Note that there is no protection circuit other than the zeners which protect the gates from over-voltage. Bias is not adjustable but can easily be made so. For lowest-THD, mosfets require local feedback, high feedback factors for the front end, and/or high idle current. The application circuit has none of these but will perform adequately for guitar, bass, or as a sub-woofer amp.

Although the IRFs are derided as "switching" mosfets, they and all other mosfets can be used in linear circuits, including audio power amps. The supposed advantages for audio do not really exist, but mosfets perform excellently in other areas.

Have fun
Kevin O'Connor

[loudthud]

I can't read what the rail voltage is clearly, is it +/- 55V ? A single pair of the TO-220 devices might be a little light for musical instrument use with 4 ohm or lower loads. You might want to double or triple up on the output transistors and spread them out over a large heatsink.

Yah, what Kevin said. The circuit has no thermal feedback so bias will drift as the thing heats up. The Hitachi FETs didn't need much thermal feedback but the IRF parts have a much sharper threshold. You'll tend to get crossover distortion unless the bias current is pretty high.

[KevinOConnor]

Hi Guys

The mosfets are TO-247/TO-3P case devices rated for 150W each. If properly cooled, the design as presented can easily drive an 8R speaker to 100W. The Hitachi app used parallel TO-3 devices rated for 100W each.

As we advise in the TUT-series, when using mosfets in linear applications you should go with larger case styles for easy mounting, and always use a device with a larger die than what "looks" like it will do. The die size is the size of the actual silicon wafer inside. International Rectifier had the die size as part of the part number in the early days, but does not anymore. However, the die size is listed in the detailed spec sheet. It is easier for hobbyists to just go by case size.

The TO-247/TO-3P case mounts easier than a TO-220, requiring just a mica insulator and a bolt, where the TO-220 also needs a shoulder washer to insulate the bolt from the tab.

Manufacturers offer many devices with a full epoxy coating all around, designated as an "F-Pack" or similar. This introduces more thermal resistance, so the device is derated to about a third or quarter of the non-insulated package.

Have fun
Kevin O'Connor

[Giaime]

Bad design, no thermal compensation. Read new Bob Cordell's book, has a good chapter about mosfet amplifier design.

[Alex R]

Thanks for the information guys.

Yes it's +-55v; that's the maximum recommended.

The same firm do a 4-mosfet board too. I was wondering whether they might fit in old Traces or Sessionettes etc, which sometimes have unrepairable and unreplaceable power boards. I might experiment a little...

[J M Fahey]

As shown, it *will* work, sort of, because it's terribly underbiased.
It will crossover distort all day long; just play a clean chord and let it die into buzz/mush.
If you substitute that 120r resistor with a 1K trimmer you may be able to bias them, but then they will runaway and burn on any reasonable load, they demand very good thermal tracking.
The "designer" just took Hitachi's design , as Kevin noticed, and substituted cheap VMOS for the expensive original LMOS, without any further thought.
"Small problem" is that LMOS is turned on around 0.6 to 0.9V, with a smooth start transfer curve (see the Datasheet) and has negative thermal coefficient from very reasonable 100 or 200 mA up, while typical "Hexfet" need almost 4V (my workhorse IRFP250 VMOS typically needs 3.78V) with an abrupt curve and dangerous positive Thermal coefficient.
Although very robust, they *do* need short protection.
In fact, even Hitachi ones really need it too.

[Giaime]

There's a Linear chip that can do automatic and temperature-proof bias for mosfets: check it out
Linear Technology - LT1166 - Power Output Stage Automatic Bias System

read about it in the aforementioned Bob Cordell's book.

[Steve Conner]

+1 on what JM said. That circuit is meant for use with the good old lateral FETs and they have put the new, cheaper VMOS type in it. Result bad performance and risk of thermal runaway as he said.

Big names like Ampeg (and Fahey? ) have used VMOS, but they need emitter ballast resistors and thermal compensation of the bias voltage for reliable operation.

[Alex R]

Thanks JM and Steve. So I could swap in BUZ901/906 and get something stable and properly biased?

[Alex R]

There's a Linear chip that can do automatic and temperature-proof bias for mosfets: check it out
Linear Technology - LT1166 - Power Output Stage Automatic Bias System

read about it in the aforementioned Bob Cordell's book.

So this works by continuously sensing current passed by output devices and supplying correct bias at continuously varying levels? At every point on the signal wave?

[Steve Conner]

Alex, yes. You would essentially have the old Maplin MOSFET amp module. Other lateral FETs are available in the UK, like the ones from Exicon and Alfet. I like them, IMO they are the missing link between transistors and tubes. (the lateral FET concept dates from the 1930s, before the invention of the BJT)

I guess that bias chip does do what you say. Never tried it myself. There are a few others like it, the LME49810 springs to mind, though it's for bipolar power devices.

[Alex R]

Thanks Steve. The boards cost less than 25 quid so it's still a viable option with lateral FET replacements. I'll give it a go - I have a couple of Traces and a Sessionette in here that need new power boards.

[Steve Conner]

Ok, but be careful, I think the lateral ones have their source and drain pins the other way around. If you just solder them in place of the IRFs the magic smoke may come out.

The IRFs might be good enough for rock and roll. But I think the ACD100 ones from Class-d are cheap just now.

[J M Fahey]

Of course, you will need TO247 (or TO218) cased MOSFETs; the BUZ I know are metallic TO3 but of course there should be a plastic version ... or some 2SK/2SJ types.
As far as they are Lateral or Audio types, fine.
I had to work with the switching ones out of sheer practical reality in Argentina, but I was driven crazy by crossover distortion, if biased "safe".
Heavy metal *loud* bass players were no problem at all; the big boxy no_highs, bass heavy "fridges" , either 8x10" or 2x15" masked it , but piano players or PAs used in quiet settings complained of "some annoying buzz in the speakers"
I also had my fair share of VHF instability and oscillation; if the MOS were mounted in a heat sink just inches from the board, I had to make stamp size "floating" mini-boards soldered right at the transistor legs, to put the gate damper resistor just there.
In the long run I used a Vbe multiplier with a TO92 sensor either *into* a hole surrounded by grease or with the flat side epoxied to heat sink or back panel, with a little aluminum clamp over it.
And even so I use them *slightly* underbiased, just to play it safe as in the Peavey tradition.
But you will have none of these problems if you use the right Lateral ones.
Good Luck.
PS: there's even is a new gadget, *both* N and P Mosfets on a 5 leg TO247 case, excellent for a very compact 100W amp.
http://www.semelab.com/pdf/magnatec/...LF08NP20V5.pdf
These guys seem to be injecting new life into this market, good for them.
They *must* have some suitable TO247 regular N and P devices for you.
EDIT: a real world supplyer:
http://uk.farnell.com/jsp/search/bro...=mode+matchany