The ultimate DIY? A Stax SRM-T2!

[Kerry]

I'm sure the A version of the BC556 should do just fine.

Just a note on the thermal grease.  I've had issues with some types being conductive at 500+ volts - it's been years so I don't remember what I used.  Make sure it doesn't get between the Aavid washer and the screw / ceramic insulator.  

Looking good so far

[starcat]

Thanks, Kerry - good point.

I've used for the PSU EKWB EK-TIM Ectotherm I had on pile from high-performance water cooled graphics cards and ordered some more Arctic MX-4 for the amp itself. Both are stated to be free of metal particle and non-conductive.

[starcat]

What are you guys using for the 5M resistor on the Bias output to the Stax jack? Xicon 273 series doesn't have anything above 2.2M and KOA has only 4.7M in the MF1/2 series with 0.5w 50ppm and 700V?  

Edited March 18, 2021 by starcat

[hhobeika]

3 minutes ago, starcat said:

What are you guys using for the 5M resistor on the Bias output to the Stax jack? Xicon 273 series doesn't have anything above 2.2M and KOA has only 4.7M in the MF1/2 series with 0.5w 50ppm and 700V?  

I used Vishay VR37000004704FR500 Metal Film 1/2W 4.7M 1% in my KGSSHV.

Mouser P/N: 594-VR37000004704FR5

[starcat]

Thanks very much. The Vishay VR37000004704FR500 looks like being 200ppm vs the KOA MF1/2CC4701F 50ppm, if that really matters on the output. 

Edited March 18, 2021 by starcat

[hhobeika]

5 minutes ago, starcat said:

Thanks very much. The Vishay VR37000004704FR500 looks like being 200ppm vs the KOA MF1/2CC4701F 50ppm, if that really matters on the output. 

I don't think the temperature coefficient will matter in this part of the circuit but I guess both resistors would work just fine considering their voltage rating.

[starcat]

That KOA is 4.7K, LoL. Will get the Vishay VR37000004704FR500. Thanks for the heads up!  

[hhobeika]

Just now, starcat said:

That KOA is 4.7K, LoL. Will get the Vishay VR37000004704FR500. Thanks for the heads up!  

Good catch.  Haven't noticed that!

[Pars]

A couple of old T2 BOMs I have use a 71-CMF60-5M for the 5M resistor.

[JoaMat]

This was not meant to happen. Modifying the original DIY T2. But now it’s done. I couldn’t resist to change the output current sources.

Replaced two 2SA1486 with single 2SA1968 and removed four 150K resistors. I put a small heat sink on the transistors. They will dissipate half a watt or so.

 2SA1968 is a 900V device, but I’ve reduced high voltages to +/-400V so that should be fine. By the way the 2SC3675 is also  a 900V device.

[starcat]

Me preparing the wiring and waiting for some Xicon 300R and finally solder the C3675 as well (still need to test).

Started with the amp board as well...

Edited March 30, 2021 by starcat

[G600]

Reminds me good memories!

[jamesmking]

Longish term modern DIY T2 reliability.

I estimate its been used around 8+ hours (and more often than not 12 hours) a day almost every day since I built it.

yesterday I took it apart for cleaning. I noticed all four 140K resistor strings connected to the voltage reference in the virtual batteries where somewhat discoloured.... hmmm... on further testing I found the D24 led on both channels did not light up at all not even dimly. :-(. but the amp seemed to work and sounded good.

(on the left is a brand new resistor from the same pack as the 140Ks I removed from the virtual batteries.)

I checked also psu voltages - good. I measured the voltage across the D24 led- about 1.4V. When the amp was originally built these leds definitely glowed as strongly as the others. All the leds where tested before soldering in and were from the same bag I even hand matched them just for overkill. 

The D24 leds are part of the circuit that provides base voltage and current to transistors that provide feed into the virtual batteries. I checked the virtual batteries - all were 741V on startup and the adjustment pots could adjust the output. So the batteries look good despite the slightly cooked resistors. I removed the voltage references and tested them in a glden reference LV - all good.

The discoloured resistors can be explained by the fact they have about 0.423W flowing through them excluding base current draw from the 2 transistors and are on the underside of the board with little airflow - maybe1W would be better or some airflow holes in the pcb, or mount them to the topside of the pcb. So I decided to de-solder the 140K resistors for measuring and the 140K resistors have drifted slightly but all measure within 141.5K. So a little cooked but probably not the case of the problem.

But why are the leds not lighting....

Looking at the circuit diagram the D24 leds should get about 0.5mA current excluding any draw from the bases of the transistors. (560V / (560K *2 resistor string to ground). I used the exact SAME leds for the other positions and it also looks like those leds get about 0.5mA excluding base transistor draw and they glow nice and bright.

I de-soldered the leds and tested the D24 resistor string to ground. They spot on resistance and have not been damaged or drifted and so should provide 0.5mA current draw through the led.

However the desoldered D24 leds would not glow at all with diode check on my brymen BM869s multimeter but do pass current and are not open or shorted. I tried my keithley  2015 in diode check mode (it has adjustable test current) and got absolutely no glow at 10 or 100micro amp but did get strong glow at 1mA - same  situation on both D24 leds. I tested my remaining leds from the same bag and they glow on the brymen admittedly fairly dimly, and glow on the keithley at 100 micro amp again dimly but visibly. (I did test all of them on the brymen before soldering into the amp). So  the leds in D24 have degraded over time . I used the same leds in the other places in the amp and they are all fine.

This would explain why the batteries are working but the leds are off....

So the plan is to do some more testing on the leds and replace them with another model. But what's causing the leds to degrade in brightness on the -560V rail when the others have not? any ideas?

regards and and I apologise for the long winded post

James

 

Update:

I replaced the D24 leds... they now light brightly with a voltage drop across them of about 1.7V.

I replaced the resistor strings with the same resistors but now placed on the top of the pcb, raised from the pcb by about 1cm. After a ten minutes with no lid on the amp the resistors reach 65C on the top and 71C on the sides. So its easy to see how on the underside of the pcb with the lid on they could reach temperatures which could cause discolouration.

 

 

Edited April 30, 2021 by jamesmking

[JoaMat]

50 minutes ago, jamesmking said:

But what's causing the leds to degrade in brightness on the -560V rail when the others have not? any ideas?

A less contagious and unknown mutation?

50 minutes ago, jamesmking said:

I apologise for the long winded post

No need to. Informative posts are always appreciated.

[simmconn]

The components will age, LEDs in particular, and accelerated under elevated temperatures. I would check the LED’s I/V curve on the curve tracer before popping it on a circuit. They could differ quite a bit in characteristics even if looked similar on the outside. If a circuit only designed less than 1ma for the LED, that’s likely starving it. I’d make sure the operating current is sufficiently past the “knee” area of the curve.

[jamesmking]

16 hours ago, simmconn said:

The components will age, LEDs in particular, and accelerated under elevated temperatures. I would check the LED’s I/V curve on the curve tracer before popping it on a circuit. They could differ quite a bit in characteristics even if looked similar on the outside. If a circuit only designed less than 1ma for the LED, that’s likely starving it. I’d make sure the operating current is sufficiently past the “knee” area of the curve.

I think it might well be thermal. About a month ago I replaced the groove tube el34s with modern production branded mullards and the bases of these run far hotter than the groove tubes so that could also have contributed more heat conduction to the leds... I did match all 6 leds in the amp by curve tracing on a peak dca75. The other 4 leds still work well and have very similar voltage drops meanwhile both the D24 leds (which are positioned close to the El34s) degraded. I did not raise any of the leds off the pcb and I think there was excessive heat conduction from the pcb into the underside of the led. This time i'm using leds from the same batch but have raised them off the pcb and time well tell if they degrade.

I carefully de-soldered the degraded leds and curve traced them. As expected their characteristics had considerably changed compared to the rest of the leds in the bag.

The D24 led on the left channel which is closest to the EL34 had the most change from new and the lowest voltage drop in circuit 1.4V in circuit, the righthand channel D24 is a little further from the EL34 and in circuit had 1.5V drop and is a little closer to a new led which has a drop of about 1.7V in circuit. To put the change into context I measured multiple leds from the bag and the curves almost superimposed over each other. So the drift in the D24 leds is significant.

Edited April 30, 2021 by jamesmking

[simmconn]

Nice! Looks like around 2ma would give the least impact to the circuit due to LED aging 😄. Joking aside, the LED part you used doesn’t seem to have a sharp curve. If I’m not mistaken, the LED in the circuit here is expected to act like a zener diode with low dynamic impedance. Maybe it’s time to pick a different P/N.

[starcat]

I still have not soldered the LEDs to the mainboard, so any ideas for new P/N's are highly welcome 🙂 

[starcat]

On 4/27/2021 at 10:31 AM, jamesmking said:

I noticed all four 140K resistor strings connected to the voltage reference in the virtual batteries where somewhat discoloured.... 

James, were the discoloured resistors KOA or Xicon? 

[Kerry]

For D24, most people used LTL-4213 on the original builds.

In the active batteries, I use a 7V reference and set the current down to .00093mA (3x 260k resistors).  That puts about .23W on each resistor. 

You’d need to adjust the sum of R31, R32 to 14.7k

 If you’re planning on doing this, I can double check the values. 

[simmconn]

14 minutes ago, Kerry said:

For D24, most people used LTL-4213 on the original builds.

In the active batteries, I use a 7V reference and set the current down to .00093mA (3x 260k resistors).  That puts about .23W on each resistor. 

You’d need to adjust the sum of R31, R32 to 14.7k

 If you’re planning on doing this, I can double check the values. 

Typo? Based on your calculations the current should be 0.93mA

[Kerry]

Yes. Typo. Food coma  

[jamesmking]

9 hours ago, simmconn said:

Nice! Looks like around 2ma would give the least impact to the circuit due to LED aging 😄. Joking aside, the LED part you used doesn’t seem to have a sharp curve. If I’m not mistaken, the LED in the circuit here is expected to act like a zener diode with low dynamic impedance. Maybe it’s time to pick a different P/N.

the fresh led has a sharper curve than the used ones, I am zoomed in on the curve, If we look at the full curve the LtL-4221nlc is sharper compared to the HLMP-1700.  The TLLR4401 looks promising but is also not quite as sharp. 

I almost feel like putting 2 pin sockets in the D24 position and trying different les to discover which sound best 🙂

6 hours ago, starcat said:

James, were the discoloured resistors KOA or Xicon? 

Xicon 273 series. Although discoloured they are not burnt. The drift from new was less 140.5K -> 141.5K so not actually that significant. I did place them on the underside of the pcb. This time I'm placing on the top side and I think that will be enough. Although 0.42W through a 0.5W part is still a little close to the limits for my tastes. I have modified my gerbers to include 4mm airflow holes under each pcb. I think one day I will build a second T2 with some modifications built in.

Edited May 2, 2021 by jamesmking

[Craig Sawyers]

The rated power for a resistor is in free air. So to push 0.42W into an 0.5W rating, it absolutely needs to be spaced from the board to allow convective air flow. If tight to the board, there is a hot area where it contacts the board and long term will lead to resistor failure.

In fairness, even Tektronix have been guilty of that, with the board discolored under some resistors.

Quad was less successful in the 405 power amp, where a 3.3k resistor feeding a 15V zener in underrated. In due course it fails open, putting one output to rail. The ridiculous crowbar then fires and short circuits the output. The power transistors and drivers then instantly fry. Then the crowbar itself blows up. Then the fuses blow.

[G600]

33 minutes ago, Craig Sawyers said:

Quad was less successful in the 405 power amp, where a 3.3k resistor feeding a 15V zener in underrated. In due course it fails open, putting one output to rail. The ridiculous crowbar then fires and short circuits the output. The power transistors and drivers then instantly fry. Then the crowbar itself blows up. Then the fuses blow.

Wow, what a cascade of horrors!