The ultimate DIY? A Stax SRM-T2!

[Craig Sawyers]

i have 2 6dj8's that

went massively noisy. One happened at cj. But i am using old stock tubes

pulled from tek scopes. The two tubes that went bad, i put on the side

and recently put one of them back in, and sure enough it was definitely

toast. But it still tests good on the tube tester.

Interesting - I was planning on doing precisely the same, having got a ton of Mullard 6DJ8's from an old and sad 535A. They all measure right on the money, so it will be interesting to see if I get a similar noise problem.

The T2 uses the input tubes in a way that i have never seen before, and

i'll never likely see it again.

Perhaps I'm missing something, but it seems a differential pair cascode with constant current loads

[kevin gilmore]

Perhaps I'm missing something, but it seems a differential pair cascode with constant current loads

Well yes, but show me another audio product that uses this scheme. Then again, there would be virtually

no reason most audio products would need to do something like this. Old tube Tek scopes did this in a few

places mixing solid state and tubes.

[kevin gilmore]

I think that andy's remaining problem is that the entire chassis is not fully grounded together in one

piece, and a couple of star washers should fix that problem. But it does make music.

[Craig Sawyers]

Well yes, but show me another audio product that uses this scheme. Then again, there would be virtually

no reason most audio products would need to do something like this. Old tube Tek scopes did this in a few

places mixing solid state and tubes.

I know it is FET's rather than tubes, but Erno Borbely used this scheme in many of his products, detailed in "JFETS: THE NEW FRONTIERS,PART 2", Audio Electronics 6/99, 16-20. He did not use cc loads on the top drains, but has a cc load on the long tailed sources. I downloaded it originally from Erno's site, but he has now shut down the business and taken his site down.

[kevin gilmore]

Erno was one of the masters for sure.

here is that article

http://www.tkhifi.com/div/Erno_Borbely_fet_articel_2.pdf

and part 1

http://www.tkhifi.com/div/Erno_Borbely_fet_articel_1.pdf

[Craig Sawyers]

With a bit of luck my batches of C3675's will arrive tomorrow.

To prepare for that, I've pulled all the suspect 3675's and checked them on the curve tracer. Two from the batteries were shorted. All the others on the heatsink survived, but had beta between 6 and 15 and had premature breakdown at 550V to 650V. All but one of the eight K216's were also dead. So the heatsinks now only have four high quality C3675's, the A1486's and the J79's. Still need to pull the dead 100V zeners and dead LED's. I hope that has extricated all the demons. All the board mounted 3675's are high quality ones.

[Inu]

I have counterfeit K216. I will take a photo later.

[livewire]

Wow! I feel for you guys.

I hope y'all get to the bottom of this slippery slope soon.

[Craig Sawyers]

I have counterfeit K216. I will take a photo later.

Jeez - is there no end to the junk semiconductor hell we seem to be going through on this build? Even the brand name seems to be counterfeited judging by Justin's Sanyo-stamped C3675's that have a beta of 4 (spec = 30 minimum). I've just decided to pull the four J79's too and verify them on the curve tracer. Tomorrow.

In 35+ years of electronics design and build I've only ever had one out of spec device until now - and that was a 74S-something in which the transition times were out of spec and was generating a few nanoseconds duration glitch. In 1982.

[Inu]

For the ultimate DIY? A Curve Tracer!

[justin]

For the ultimate DIY? A Curve Tracer!

and it must be all tubes

[Craig Sawyers]

and it must be all tubes

<grin>. The Tek 575 transistor curve tracer has precisely two (germanium) power transistors in the step amplifier. Every other active device is a tube. Not totally suitable for the transistors we're concerned with - maximum base step is 0.2V (so not high enough for the FETs) and the basic instrument will only do 200V sweep. The mod-122C version goes to 400V sweep, which is still not enough for the high voltage bipolars in the T2.

The only two golden oldie Tek tracers that will do the job on T2 semis this are the 577 and (if you have deep pockets - which you need for tackling the T2 in the first place) the 576. I'm using a 577 for my measurements.

[struts]

With a bit of luck my batches of C3675's will arrive tomorrow.

Craig, Were any of your 'suspect' C3675s from Dalbani? Where have you reordered from? If Dalbani has been supplying duff silicon I'm in trouble, sourced nearly all silicon from them...

[livewire]

Craig, Were any of your 'suspect' C3675s from Dalbani? Where have you reordered from? If Dalbani has been supplying duff silicon I'm in trouble, sourced nearly all silicon from them...

o.O! See post #1754 - Re: Dalbani

[Craig Sawyers]

Craig, Were any of your 'suspect' C3675s from Dalbani? Where have you reordered from? If Dalbani has been supplying duff silicon I'm in trouble, sourced nearly all silicon from them...

Yes - all of the duff C3675's were from Dalbani. A couple of years ago I bought a batch from them that are perfect - and luckily the two in my BH, the four in the T2 PSU, and all the board mounted ones on the T2 amp are from that batch. But the additional 34 that I ordered from them in the Summer were Chinese iSC brand junk - and all but four of the heatsink mounted ones were from that batch. Dalbani are replacing them without question with Sanyo ones, but I've chased my tail for quite a while tracking down the root cause of my woes, and had to buy quite a few more K216's and 100V zeners in the process.

All the other silicon from Dalbani has been just fine. Well, not quite - some spare (Sanyo) J79's had the leads hanging on by a thread where they enter the body of the device. But again these were replaced with warp speed.

If you haven't soldered yours in yet, shove them in the post and I'll verify them on the tracer for you (Sweden to UK is not too far).

[struts]

Yes - all of the duff C3675's were from Dalbani. A couple of years ago I bought a batch from them that are perfect - and luckily the two in my BH, the four in the T2 PSU, and all the board mounted ones on the T2 amp are from that batch. But the additional 34 that I ordered from them in the Summer were Chinese iSC brand junk - and all but four of the heatsink mounted ones were from that batch. Dalbani are replacing them without question with Sanyo ones, but I've chased my tail for quite a while tracking down the root cause of my woes, and had to buy quite a few more K216's and 100V zeners in the process.

All the other silicon from Dalbani has been just fine. Well, not quite - some spare (Sanyo) J79's had the leads hanging on by a thread where they enter the body of the device. But again these were replaced with warp speed.

If you haven't soldered yours in yet, shove them in the post and I'll verify them on the tracer for you (Sweden to UK is not too far).

That's an extremely kind offer Craig. Everything is soldered in place but I haven't started testing as I haven't got the umbilicals made up yet. Sounds like for the sake of a couple more weeks it would be wise to pull the 3675s and send them over to you. Rather that than risk silicon genocide and a protracted and complex troubleshoot. Please could you PM me your address?

[Inu]

I have counterfeit K216. I will take a photo later.

Uploaded.

Picasa Web Albums - inuponken - T2

[Craig Sawyers]

Uploaded.

Picasa Web Albums - inuponken - T2

Wow. If you look at the pics of the C3675 I posted a few days ago you will see the same differences in the lead frame - the bad ones all have a short supporting section near the transistor body.

The good news is that I've just received 41 off C3675 that are all measuring right on the money.

[Craig Sawyers]

Some C3675 measurements on a batch of 41 received from Vintage Parts.

I got these through the post yesterday, and have measured the current gain at 20V, 5mA and 750V, 5mA.

Minimum: 53/65 @ 20V/750V

Maximum: 74/96 @ 20V/750V

Majority: from 60/73 to 67/80 @ 20V/750V

So why not just pick the super-high gain ones?

Well, the gain enhancement at 750V has two components. First, the collector characteristics have a gradual slope, measured at 10^-6 (ie an output resistance of a megohm). That would enhance the collector current for a given fixed base current from 5mA to 5.75mA going from 20V to 750V - ie 15%. But the gain ratios measure around 22%. That is because the curves have an additional component at high voltage as the transistor approaches breakdown at >900V - a slight upward curve starting around 500V. Actual VBR(CEO) breakdown is over 900V (measured at 1100V with base open circuit, against spec of >900V)

However, the very high gain transistors have a gain enhancement of 30%, which implies that their breakdown voltage is somewhat lower than their lower gain brethren.

So for the batteries, where the C3675 is stressed the most, I'll be picking eight transistors with a high voltage current gain of around 75. That will put the base current at around 67uA at Ic=5mA, 740V, well within the capability of the battery drive circuit.

That is the logic, at least....

If the replacement Dalbani transistors arrive today, I'll measure all those too, just to make sure they are pukka.

[struts]

I have counterfeit K216. I will take a photo later.

Inu, How did you determine they were counterfeit? Just by the measurements or are there any other telltale signs?

I noted Craigs comments about the lead frames above, I wonder if this is a common characteristic of counterfeit parts? If so, I wonder why? Seems silly to go to all the trouble of producing such exact copies and then making such an easily identifiable mistake..

[kevin gilmore]

So here is my comment on the "fake" parts.

I've been stocking up on lifetime supplies of the semiconductors i may need in the future.

When you get them in large bags, the bags have a label on them that says where they

were diffused, and where they were packaged. The diffused is always japan, and the

packaging is taiwan,korea,china... So the parts could be the real silicon, but just packaged

inside of less than adequate lead frames.

[Craig Sawyers]

The diffused is always japan, and the

packaging is taiwan,korea,china... So the parts could be the real silicon, but just packaged

inside of less than adequate lead frames.

Interesting. So the problem with the iSC ones could be poor packaging stressing the silicon. Mine not only have poor gain, but the collector base breakdown voltage with base open VBR(CEO) is 700V. The good ones are measured 1100V (spec is >900V).

[struts]

That's an extremely kind offer Craig. Everything is soldered in place but I haven't started testing as I haven't got the umbilicals made up yet. Sounds like for the sake of a couple more weeks it would be wise to pull the 3675s and send them over to you. Rather that than risk silicon genocide and a protracted and complex troubleshoot. Please could you PM me your address?

Yup, unfortunately mine look like the suspect ones . I'll see if I can get them desoldered over the weekend and send them off to you, Craig. I hope Dalbani have stocks, I noticed you cleaned out Vintage Parts!

[Craig Sawyers]

Craig. I hope Dalbani have stocks

Well they have allegedly sent me 34 replacements for the crap ones last Monday. They still have not got through the Christmas post backlog yet.

I noticed you cleaned out Vintage Parts!

Guilty as charged

[Craig Sawyers]

Here's a question for Inu and Kevin. Is there any indication of the power Mosfets oscillating? Or bursts of oscillation at certain points in real signal? Reason I mention it, and I should have recalled this from power amp designs, they have a horrid tendency to burst into high frequency self oscillation. There was one generation of Trace Elliott bass guitar amp in which the power mosfets physically exploded through this effect (although at 1kW output there was LOTS of energy driving them).

I spotted that on the K216's on my curve tracer, but the patch leads were around 2 inches long to the test fixture and certainly had an effect.

In the T2 design, there is a 750 ohm resistor in series with the J79 gates (and KG has it tight up to the gate, which is exactly right), which is a classic way to stop this (the old grid stopper idea), and kind of suggests that it was introduced to stop self oscillation of the J79 in the orignal circuit layout. But there is no such protection on the K216's.

So the question is - do we need to do anything? Either a resistor in series with the K216's gate or a ferrite bead on the gate lead?