Svt3 Pro Mosfets

[Jammer56]

Hello All,

I've come across a SVT 3 Pro with a roasted Mosfet output section...possibly other components in the output stage as well.

My question is where can I find a legitimate set of "matched" IRFP240 and IRFP9240's
(4) each?

I know that some buy in qty. and match them themselves, but I don't have the rig, time, or money for a mass purchase.

I did see a matched set offered on eBay from across the pond for $50 plus shipping, but I have to think that there is a reputable source here in the states.

Any input is appreciated.

Thanks!

[Mark Black]

What I have done is to order a complete set (plus a few, and hoping to get all from the same batch) from any reputable supplier for each polarity.

Then it's a matter of trying them out in-circuit & measuring voltage drop across the ballast resistors to see if there are any real oddballs.

Too bad Loud Technologies/Ampeg quit offering graded replacements (last I knew anyway).

[J M Fahey]

FWIW I use lots of MosFets (although in my case I settled on all Nch IRFP250) and have found that "good family" MosFets (I use Fairchild or IRF) bought from a reputable supplier such as Mouser or Digikey are incredibly uniform.

Only precaution is to buy them "all together" so they come from the same batch.

At first, I bought 1 or 2 "sticks" of 30 units, measured them all at once and stuck little labels to each one ... until I found it was unneeded, in any given batch they measured between a few mV from each other.

Test is real fast, I kludged a socket to avoid soldering (not only because of the time but because heating changes properties), or you can use 3 small crocodile clips:



So measure them for peace of mind, but you'll see they are very uniform.

And you'll get far more precision measuring with a 1K "ballast resistor" than through typical 0.47 to 0.1 ones

[Jammer56]

Safety question

Thanks for the input.

One final question regarding the SVT3 Pro:
Since the output section was fried, I don't want repeat the event.
Is there a safe zone with the trimmer pot to start with before power up?
I don't want pull too much until I get a chance to adjust the pot.
Maybe balance the resistance in the center?

Thanks again!

[Mick Bailey]

You should use the lightbulb trick first with no load. Check for DC on the output. Switch off and then connect the load. Check DC levels again. In any case, if the bulb stays bright, then switch off immediately. You need to fix the entire section in one go before powering directly off the mains, else it will go pop again.

[Mark Black]

Great post Juan!

@Jammer56 on a further note:

I usually find a lot of the "flame proof" (or "fusible" if you like) 47 ohm resistors would open up (with no outward sign) so check all those along with the usual diodes & driver transistor testing in the output section.

[loudthud]

If you hunt around on diyAudio Link: diyAudio there is probably someone selling matched sets. Nelson Pass uses lots of those parts and his amps are quite popular on that forum. Note that if/when you join that forum your first five posts will be reviewed by a moderator before they actually show up as a post.

[km6xz]

Matching is not really needed with modern power semiconductors, they are much closer in tolerance than tubes or capacitors now. It was not always the case, in fact back in the late 50s, 60s and until the early 70s tubes were tighter in tolerance than transistors.
In fact if you are using conventional passive components in the rebuild, the IRFP240 and IRFP9240's will be closer to all published spec than those passives.
Find out what caused the problem and correct that and install the new devices and glue components(double check the FP resistors, if stressed they are just as likely to become intermittent as open entirely). Burn in the amp properly at 1/3-1/2 power sine wave (where the greatest dissipation occurs) or full rail to rail if using music for burn-in.

They are simple and stable amps mostly due to the closeness of parameters of those power devices. In the not too distant past, in-coming parts testing and measurement added more to the total cost of components than buying them in the first place for manufacturers building quality equipment. Now, it is really only needed for tubes which have gone downhill in quality over the last 40 years.

[diydidi]

Matching is not really needed with modern power semiconductors, they are much closer in tolerance than tubes or capacitors now. It was not always the case, in fact back in the late 50s, 60s and until the early 70s tubes were tighter in tolerance than transistors.
In fact if you are using conventional passive components in the rebuild, the IRFP240 and IRFP9240's will be closer to all published spec than those passives.
Find out what caused the problem and correct that and install the new devices and glue components(double check the FP resistors, if stressed they are just as likely to become intermittent as open entirely). Burn in the amp properly at 1/3-1/2 power sine wave (where the greatest dissipation occurs) or full rail to rail if using music for burn-in.

They are simple and stable amps mostly due to the closeness of parameters of those power devices. In the not too distant past, in-coming parts testing and measurement added more to the total cost of components than buying them in the first place for manufacturers building quality equipment. Now, it is really only needed for tubes which have gone downhill in quality over the last 40 years.

Sorry to resurrect an old thread.
Are you saying that matching of BJT's and Mosfet's are no longer really necessary?

[J M Fahey]

Guess km6xz thinks about the same, but personal experience is that MosFets from the same batch are always incredibly close in specs.

Can´t vouch for different batches or worse, different brands.

In a nutshell: in a servicing job where I´m replacing, 2 or 3 out of, say, 8 devices (same polarity, same rail of course) I would either pull the good ones to check VGSON@11mA (what my tester tests) and select same voltage among the new ones (if available) *or* plain replace all 8 for peace of mind, specially if it´s a critical amp such as one used in a live PA or DJ setup, where lost revenue may be much higher than parts cost.

Now if I´m building new stuff or replacing *all* devices, yes, I´d check anyway but I´m quite confident about uniformity.

[diydidi]

I have fixed quite a couple of Montarbo W440 speakers, replacing all the fets without matching, and just making sure the Batch numbers were the same.
And, thank God, they haven't come back yet. So, I had it that this was good enough.
Lately I received a couple of SVT 3Pro's and 4's for repair. I did my usual 'same batch' exercise on output stage. When it got to biasing I consulted the schematic. It says: adjust bias pot for 25mV across each source resistor.
So I turn bias up slowly, get 25 over one resistor, check across the others and what do I find.. One is up at 35mV another 21mV and the lowest was 17 something.
This got me thinking; Have I got it all wrong???
Maybe as km6xz stated, this spread is OK...
Any thoughts???

[MarkusBass]

For SVT3Pro and 4Pro the MOSFETs have to be matched. With "one batch" approach you will always get the results like you have now. The problem is that MOSFETs from one batch have different Vgs(off) voltage. I usually have to buy 10 MOSFETs to select 4 matching.
But...
Even the MOSFETs installed by Ampeg differ sometimes by 100 per cent (eg. 14 mV and 28 mV) and such amps work without any problems. When I match MOSFETs I get about 20 per cent accuracy and I prefer to do the matching. I'm confident that such amp will perform better than an amp without matched transistors.
And I suggest that you match the MOSFETs for SVT3Pro no matter what people say.

Mark

[J M Fahey]

I have fixed quite a couple of Montarbo W440 speakers, replacing all the fets without matching, and just making sure the Batch numbers were the same.
And, thank God, they haven't come back yet. So, I had it that this was good enough.
Lately I received a couple of SVT 3Pro's and 4's for repair. I did my usual 'same batch' exercise on output stage. When it got to biasing I consulted the schematic. It says: adjust bias pot for 25mV across each source resistor.
So I turn bias up slowly, get 25 over one resistor, check across the others and what do I find.. One is up at 35mV another 21mV and the lowest was 17 something.
This got me thinking; Have I got it all wrong???
Maybe as km6xz stated, this spread is OK...
Any thoughts???

Think of it this way:
> you are actually seeing 35-17=18 mV difference between turn-on voltages.
> scary 2:1 current differences ... until you realize that means 17/.47=36mA difference ... which still looks bad.

Let´s put it in perspective:

what really matters and which might kill some transistors is 2:1 difference in current at high power.

let´s calculate that.

just quoting from memory, too sleepy to go get the actual schematic but just as an order of magnitude analysis, I seem to remember we have roughly +/-70V rails, 4 ohm loads, 6 MosFets (3 per side) and 0.47 ohm ballast resistors.
If mistaken, just repeat calculations with proper values.

Supposing pèrfect parts:
70V/4r=17.5A peak current
being 3 devices: 5.8A each
which through 0.47 ohm resistors will drop: 5.8 * .47=2.7V = 2700 mV

those extra 2700mV which we added as necessary to drive these MosFets will swamp/crush the puny 17 mV difference

The ballast resistors end up being the main factors in determining MosFet high currents and AFAIK they come in 10% tolerance or thereabouts

so, as far as the MosFets are from same batch and you have reasonable ballast resistors, you are perfectly safe.

Now, in designs (there are a few like that out there, generally mindless "conversions" from Lateral Mosfet original schematics) without ballast resistors, life will become "interesting" , such as Evel Knievel´s:



NOTE: I am not saying that MosFets do not need to be matched, far from it.

What I *do* say is that modern manufacturing is incredibly precise and they usually come pre-matched close enough from the Factory !!!!!

Brave New World !!!

[diydidi]

All good and well. BUT It doesn't make life any easier. How then do I bias the amp properly if i'm not able to bias for an average of 25mV over each ballast resistor.
OR, should I still bias for Average regardless of the highest and lowest reading?
Taking my example above 17+21+22+25/4= 23.75
Any other way of biasing recommended?
I tried biasing with 4ohm load and 1khz signal for little crossover notch visible. But this was too cold: got about 1mv on some ballast resistors..??

[MarkusBass]

All good and well. BUTHow then do I bias the amp properly if I'm not able to bias for an average of 25mV over each ballast resistor.

You seem to forget about the second option: match the transistors. I do it and I get voltages from 22 to 27 mV - then the problem that you mention simply does not exist . Ampeg does not match the MOSFETs but they get preselected transistors from transistors manufacturers. The transistors are marked with a colour dot (Vgs(off) voltage range). Of course the transistors are from one batch but they still do it. If this were not needed, Ampeg wouldn't order preselected transistors (because this is additional cost).

Mark