Just picked up an Adcom GFA-5800 (broken)

ben62670 · October 2009

Is the 5800 an NP design? I agree with the matching of the transistors as well as the resistors. It is not an easy under taking for the average DIY guy. It is one more piece of equipment. I paid a little over $100 for my beta tester used.

Systems · October 2009

ben62670 wrote: »

Is the 5800 an NP design?

Yes

I agree with the matching of the transistors .... is not an easy under taking for the average DIY guy.

And can get costly if you need a large sample group to get your desired matches.

disneyjoe7 · October 2009

Are the Mosfets shorter or otherwise bad? Do we know at this point.

heiney9 · October 2009

ben62670 wrote: »

Is the 5800 an NP design? I agree with the matching of the transistors as well as the resistors. It is not an easy under taking for the average DIY guy. It is one more piece of equipment. I paid a little over $100 for my beta tester used.

Yes, and I was just saying what the absolute best way was. I realize it's difficult for an avg diy'r. I know you can pay extra to have them matched from some suppliers.

mmadden28 · October 2009

Hey guys--

I went through the schematic and tested every part-and removed some from the circuit where parallel resistance factored in just to be sure. Essentially not much more wrong than I originally found--So far all the 16 previously mentioned resistors are open, found a flaky Zener and one other open resistor. All 8 transistors on the affected half are shorted. All fuses are intact. There are a couple other zeners that aren't measuring as expected, but they are beyond the path of the damage and I haven't taken them out of circuit yet--perhaps later tonight.

So at least I now know the extent of the damage. I'll post a pic of the schematic marked up with good and bad parts in a bit. Maybe somebody can figure out what could have caused the failure to begin with based on the path of damage.

I already removed all the bad parts and the good resistors that I'm going to replace anyway. And am cleaning the board now.

I found Nelson Pass's article on testing MOSFETS, and ironically in the article he uses the same transistors that are in the 5800 as samples for the test. The tester will enable me to calulate the Vgs for each transostor under test. I'm building the tester now to test all the other transistors and to get the Vgs baseline.

Nelson Pass commented on my post at DIYAudio and indicates the need to match the replacements--in that post and according to the MOSFET testing article I have to match the Vgs ideally less 0.01v, but no more than 0.1v .

I've determined from the test article and the schematic that the 5800 is a Source Follower or Common Drain design and configured in a parallel layout.

If I understand this correctly, the all the 'N' side transistor Vgs has to match the 'P' side Vgs. Of course since this is a parallel design, essentially all 16 transistors have to have the same Vgs as each other. This includes one N to another N as well as the N to P. This would be easier if the N and P side were individual pairs but since its setup parallel, they ALL have to be the same. Do I understand that correctly?
What would happen if they were not matched?

Systems · October 2009

I suspect his meaning was to match like devices( P-P N-N).Close matching of paralleled transistors will ensure they all draw the same amount of current.You might get lucky and recieve replacements from the same manufacturers date code which should improve the likelihood of them being similar.

disneyjoe7 · October 2009

Hmmm.... My thinking would be all N are matched to each other, where the P are match to each other, and they are all complimentary matched to each other. But as far as a killer match not sure 16 would be turf to do, so that my thinking of the 2.0 ohm resistors.

megasat16 · October 2009

mmadden28 wrote: »

If I understand this correctly, the all the 'N' side transistor Vgs has to match the 'P' side Vgs. Of course since this is a parallel design, essentially all 16 transistors have to have the same Vgs as each other. This includes one N to another N as well as the N to P. This would be easier if the N and P side were individual pairs but since its setup parallel, they ALL have to be the same. Do I understand that correctly?
What would happen if they were not matched?

You'll need to test each Mosfet before using in circuit. The P Channel Mosfets need to match with each other (gm and Vgs) and N Channel Mosfets need to match with each other. Also, gm and Vgs of P and N channel Mosfets need to match closely. In short, all P and N channel Mosfets need to be closely matched for reliable gain, low DC offset and low distortion.

Most circuits will tolerate unmatched Mosfets or transistors to some extent but your DC offset may become high if it's the case.

From my experiences and from what I learn from others, you'll get about 20 closely matched pairs out of 50 if you buy transistors and MOSFETs from the same batch (date code, etc.)

Systems · October 2009

megasat16 wrote: »

You'll need to test each Mosfet before using in circuit. The P Channel Mosfets need to match with each other (gm and Vgs) and N Channel Mosfets need to match with each other. Also, gm and Vgs of P and N channel Mosfets need to match closely. In short, all P and N channel Mosfets need to be closely matched for reliable gain, low DC offset and low distortion.

Matching between parallel devices will help in the long term relability aswell.Offset might be an issue with unmatched P-N devices but most offset in power amps is the result of poor matching in the input stage. Besides his original N channel IRF 244's are date coded 95' so may be difficult to get the new 9240's to match.

mmadden28 · October 2009

Interesting...The schematic shows at least 3 transistors (TO92) that are missing from the actual build. There are leads soldered in, but nothing else. Same thing on the working Left channel--I suppose it was a design change at some point. Also the schematic shows the N channels should be IRF240's as opposed to IRF244's. I have to look up the specs yet to see the difference, but I would have expected that if using IRF244s, that they should have used IRF9244's as the compliment in the P stage as opposed to the IRF9240's. Its the same for both channels.
Everything else appears true to the schematic.

mmadden28 · October 2009

Vishay makes a version of the IRF240--a package of 20 for $1380.00 :eek: and with a 58 day lead time. :rolleyes:

heiney9 · October 2009

mmadden28 wrote: »

Interesting...The schematic shows at least 3 transistors (TO92) that are missing from the actual build. There are leads soldered in, but nothing else. Same thing on the working Left channel--I suppose it was a design change at some point.

Everything else appears true to the schematic.

Ha Ha, so it seems not only Polk made changes on the fly and never bothered to print a new schematic. lol

Nelson is the man, it's great he is so helpful and even has a decent memory about all his designs....He's answered many of my questions is the past as well.

H9

megasat16 · October 2009

GV#27 wrote: »

Matching between parallel devices will help in the long term relability aswell.Offset might be an issue with unmatched P-N devices but most offset in power amps is the result of poor matching in the input stage. Besides his original N channel IRF 244's are date coded 95' so may be difficult to get the new 9240's to match.

Yep, it'll be hard to get new devices matched to the old ones. It's cheaper and more reliable to remove all old trannies and use the matched new ones. He can sell some of the old ones and recover a few $$.

mmadden28 wrote: »

Vishay makes a version of the IRF240--a package of 20 for $1380.00 :eek: and with a 58 day lead time. :rolleyes:

It's much much cheaper buying a working mint condition Adcom GFA-5802 than buying 20 IRF240s.

Systems · October 2009

megasat16 wrote: »

It's much much cheaper buying a working mint condition Adcom GFA-5802 than buying 20 IRF240s.

No kidding.

Systems · October 2009

mmadden28 wrote: »

Also the schematic shows the N channels should be IRF240's as opposed to IRF244's. I have to look up the specs yet to see the difference, but I would have expected that if using IRF244s, that they should have used IRF9244's as the compliment in the P stage as opposed to the IRF9240's. Its the same for both channels.
Everything else appears true to the schematic.

Maybe it was a supply thing?There was a big mosfet shortage back in the mid 90's. I remember as I was trying to source some back then for my ZEN projects.Many IRF HEXFET's are interchagable given they have sufficient current and voltage ratings for the circuit they will be used in.
I can't see in the pic but are 244's used in the other channel aswell?

mmadden28 · October 2009

GV#27 wrote: »

I can't see in the pic but are 244's used in the other channel aswell?

That Vishay price is insane, but thankfully the IR versions will only set me back about $200 for 16 Q's. Still a chunk of change but I think I'm still ahead value wise. I hope they can pick matched ones--I'll have to call them once I'm ready to order--I still have to get the resistors too.

Yes 244s were used in both channels instead of the 240s.

Systems · October 2009

mmadden28 wrote: »

Yes 244s were used in both channels instead of the 240s.

As I mentioned previous,it may have been a supply thing and the 244's voltage and current rating's were such that it was a suitable substitude.

mmadden28 · October 2009

Here's some pics of the prepared board--all cleaned up and ready for new parts

All contact points (including the fuses) have been cleaned with DeOxit and treated with DeOxit Gold as well. :cool:

Now while I'm waiting to figure out what to do about sourcing the replacements and then waiting for them to arrive-I'll give the rest of this beast a bath.

ben62670 · October 2009

Very nice. If you have problems matching them I can help;)

mmadden28 · October 2009

Here's the schematic. (sorry it was two pages an I don't have the right editing tools on this PC)

Everything circled in Red was a bad part (Resistors OPEN, Transistors SHORTED).

Circuit oddities are circled in Orange:
Q9, 10, and 12 (TO92) were all missing--looks to be a deliberate design change. nothing changed with the traces or signal path, just left open. However there are small leads soldered in the holes which you can see in the second pic in post #49. Not sure if the transistors were cut out later or what. They are missing from both channels.
C7 and C6 were also missing but replaced with a jumper.
Also as previously noted, The schematic calls for IRF240s, when IRF244s were actually used (in both channels).
It should also be noted that R14 (top middle) is a resistor for the ERROR LED. While it appeared fried (couldn't read any color bands), it tested fine-its also outside the main path of damage (and outside the signal path) but will be getting replaced anyway.

I haven't tested any of the other transistors out of circuit, but none of them were shorted. All other resitors and zeners are good. I had to remove a few from the circuit to test them due to parallel resistance factors.

So based on the failed part path--any ideas as to what caused the failure?

ben62670 · October 2009

Someone touched the negative to positive speaker wire;)
Once one goes it runs down the line very fast. I know...

mmadden28 · October 2009

ben62670 wrote: »

Very nice. If you have problems matching them I can help;)

Thanks Ben.
I'm building the test circuit from Nelson Pass's MOSFET testing article, so that I can determine the Vgs of the IRF244s that are still good. If I can match with just 8 I'll save about $100.

Here's my protoboard so far--I need to add a few switches for when I make a permanent build out of it and of course use higher wattage parts but...I'll just move the wires manually for now.

mmadden28 · October 2009

ben62670 wrote: »

Someone touched the negative to positive speaker wire;)
Once one goes it runs down the line very fast. I know...

Without popping the fuse first?
The fuses are rated 15A, I think the Transistors are rated 20A IIRC.
And why only half the channel? Why not all 16 Qs? This is one of those parts of the Audio Amp theory I lack understanding in.

ben62670 · October 2009

By the time the fuse goes it is usually too late. Hlaf a channel blew on one of my Adcom's too:(

megasat16 · October 2009

mmadden28 wrote: »

Without popping the fuse first?
The fuses are rated 15A, I think the Transistors are rated 20A IIRC.
And why only half the channel? Why not all 16 Qs? This is one of those parts of the Audio Amp theory I lack understanding in.

Well, the previous fuse should have been blown when the trannies are shorted and resistors are opened. I bet the current fuse in the amp is a replacement. The replacement fuse doesn't blow coz the resistors are blown and effectively the shorted MOSFETs are floating and no path for current to flow.

mmadden28 · October 2009

megasat16 wrote: »

Well, the previous fuse should have been blown when the trannies are shorted and resistors are opened. I bet the current fuse in the amp is a replacement. The replacement fuse doesn't blow coz the resistors are blown and effectively the shorted MOSFETs are floating and no path for current to flow.

I'm sure that's possible. I'll have to see if the fuses are easily found--All 4 output fuses are the same--15A Opaque Gray colored body.
I'm guessing that the original owner probably didn't take the time to find the correct fuse-(although he did Dynamat the heck out of it, he didn't keep it clean) unless he took it to a repair shop first where they replaced them then gave him the bad news. The box does have a return address label from a repair shop. Maybe I'll call them up and ask what kind of history they have on it and while I'm at it see if they have replacement parts

mmadden28 · October 2009

Any thoughts on RCD (RESISTORS COILS DELAYLINES) brand for wirewound resistors? Specifically to replace the 1ohm's.

It's the only brand (in the specs I need) at Newark which is one source that has both N and P FETS that I would need.

Systems · October 2009

Yes to the RCD's.

disneyjoe7 · October 2009

Ok the cost to fix this amp....

Sorry I been an electronic repair person, your amp IMHO is NFR "Not Feasible to Repair"

But with all those MOSFET I think it would work like this, and it would be really cheap. Remove every other MOSFET on the left side, also remove resistors. Put them on the other side, with all those MOSFET I would think the amp would be the same watt output. What 5800 amp is now a 2900 amp

Systems · October 2009

IMO it is not necessary to replace the 244,s if they are good, I would just buy a few extra 9240's and install the 8 closest matches.

Just picked up an Adcom GFA-5800 (broken)

Comments