Subtle fault

[ruffrecords]

If I have learnt one thing in my fairly long career in electronics it is that there really is no such thing as a subtle fault and if there is they are extremely rare. It has been a very useful maxim that has allowed me to find faults that have stumped others. But recently I very nearly met my nemesis.

The Marks 3 tube mixer is based on the EZTubeMixer design but the mechanical arrangements have changed dramatically so that the module width is halved from 2.8 inches to 1.4 inches. The tubes are now mounted vertically, the edge connector pin out has been changed completely to better distribute power and provide more signal IO and although the amplifier circuits are identical, the PCB layouts are completely new. In addition I have reduced the HT voltage and done away with heater elevation so I can use the 12V supply also as a utility supply.

So even though the circuit is proven, you still expect some teething problems with the new implementation and that is exactly what I got. I have returned to this problem on and off for nearly 5 months. It is characterised by oscillation at high gains, excessive noise and intermittent signal level changes. I changed PCB layout to add new stopper resistors very close the the tube grids, I added small caps across feedback resistors to reduce the chance of oscillation.  None nothing seemed to work.

Then a few weeks ago I discovered a problem with the switched mode power supply used for the heaters - it turns out it does not like a load of less than 10% of its nominal output - at less than 10% the dc output is still OK but it has lots of interference on it and this gets into the output and shows up as an oscillation at high gain. Plugging in enough boards to reach 10% of the SMPS rated output cured that problem. So I started retesting the new designs. The oscillations were gone but the intermittent signals and poor noise performance persisted - they were 10dB noisier than the EZTubeMixer originals. I kept plugging away trying different things and going down all sorts of false trails. I put it aside for a week or so and then returned to it but I discovered the oscillations returned at very high gains 70dB. Now I am thinking that subtle PCB layout issues may the the cause and  was toying with the idea of new PCB layouts.

Then today, on a separate project, I was making some cables that connect a mic input XLR to a 3 pin Molex connector that plugs into the motherboard into which the new boards plug. To get hot cold and screen into the right positions on the Molex I used as a reference and existing short 3 pin Molex the 3 pin Molex cable that I was using to connect the mic signal across the PCB from its input to the mic transformer. I had made nine of these cables when I decided to have a break. By chance I saw another 3 pin Molex to 3 pin Molex connector I had made for another project some time ago and I notice hot, cold and screen were wired differently on these compared to the reference cable.

To cut a long story short, the reference cable was from an old EZTubeMixer board. For the new boards I had updated the Molex connector footprint so its silk screen showed the right way to fit the Molex connector. This cue was absent on the original EZTubeMixer PCBs. I soon discovered the EZTubeMixer board I had borrowed the reference cable from  had its Molex in backwards and the cable had been wired to correct this. So the reference I was using had screen and hot swapped!!!!!. I quickly made up a correct cable and started retesting boards. To my delight all the intermittent problems disappeared along with oscillations at high gain and the noise dropped down to the expected levels.

So despite nearly losing the faith, I am still convinced that 99.9% of all faults are not subtle.

I shall sleep a lot easier tonight.

Cheers

Ian

 

[emrr]

Wow that’s a good one.  Happens to all of us. 

 

[PRR]

So feeding the Screen into the high-gain amplifier INput is bad?

Or using generic connectors where all pins look the same make it too easy to lose track of what connects to what?

 

[Bo Deadly]

I just solved something that was subtle and annoying me for a while too. I made a mic pre with a THAT 1646 output. I found that when connected to an unbalanced load, it would generate major high frequency noise. You could actually record with it but and not hear it but if I looked at the output with a scope, it was clearly visible. I used it with a balanced load just fine for a couple months. But I could never plug it into an unbalanced load for fear that it would burn out the 1646 or damage something downstream.

So I stared at the circuit for a while and then it hit me. The 1646 is powered with a single supply and biased with vitrual ground. So it has large coupling caps on each output. The XLR out has a little board with RF protection which includes a small chip inductor. So when pin 3 is grounded by the unbalanced load, I had output -> large cap -> inductor -> ground. Meaning it was a resonant circuit. Oops. Using an inductor was a little pedantic anyway. I think build out resistors are pretty standard now. Replacing the inductors with simple 100R build-out resistors fixed the problem 100%.

 

[ruffrecords]

So feeding the Screen into the high-gain amplifier INput is bad?

Or using generic connectors where all pins look the same make it too easy to lose track of what connects to what?

Not standardising and documenting pin out and following through with unambiguous PCB footprints is the root problem. I have a standard and  proper footprint but I must document them, print it, laminate it and nail it to the wall of the workshop so I do not repeat this mistake.

Cheers

Ian