I learned a lot by building this. Read : "I made a lot of mistakes."
Finished module |
Once again, I patiently captured the layout in DIYLC. I used Yves' own layouts to guide me.
I chose to make the two PCB version, one on top of the other, as I was sure one PCB would be too large for the Eurorack format.
I added two reverse polarity protection diodes. And I chose Schottky diodes this time, to limit the voltage drop.
Anyway, I reproduced the mistake of having the power connector horizontally and not vertically. Certainly due to the fact it was oriented that way in Yves' layouts.
Here are the layout I used.
The layout of the two boards |
Work-in-progress |
I quickly diagnosed that the oscillator did not oscillate.
Board inspection found nothing. Without an oscilloscope, I was blind. You cannot debug by ear a non audible LFO.
I finally bought a used Tektronik oscilloscope. It confirmed the diagnostic, but still, no clue.
Oh, what's this ? An unconnected pin on R5 ?
Soldered. First problem fixed.
Oscillator debugging. First success. |
It uses the pink noise output as reference. And nothing on the pink noise output as well.
A wrong connection this time, due a misleading layout : because of the bad layout drawing, it was not clear which pin of the Q4 transistor to connect R20. Of course, it was clear to me at layout time but I chose the wrong one at solder time.
Now white, pink and random outputs seemed to work. More on this later on.
The sample&hold did not to hold.
Not holding properly |
I reproduced the circuit on a breadboard with spare parts, so that I could better debug it and learn how it works.
Part of the circuit on a breadboard |
I switched capacitors, changed resistors, exchanged transistors to not avail.
Out of idea, I gave up and asked Google about it.
Finally, the solution was given by Yves Usson himself. In a forum he detailed why it might not work :
Here are the possible causes for a holding fault :The last three could be excluded as the circuit behaved properly with another (about 1Vpp) input signal and I had the same behavior after having reproduced the circuit on the breadboard. So dead components were out of the question.
- too high white noise signal (>8Vpp)
- a dead BF245C (Q5)
- a dead TL072 (U2)
- a dead diode (D3)
Noise source was effectively way too high : about 20Vpp.
In fact, noise saturated both output opamps so R34 and R27 should have been lowered. I didn't want to lower my noise sources as much as 8Vpp, so I chose a compromise.
I chose R34 and R27 as 220k, both because hasty measure and crude calculation would give me about 16Vpp and I had some in my drawers.
As the circuit seems to be sensitive to the input voltage, I decided to add a passive attenuator on the input.
More debugging |
It took some time and effort but the journey was worth it.