PSU deglitching

[Walter66]

Hello,
my goal is to deglitch a power supply for tube electronics.

The output shows glitches, which seem to be of higher frequency than the random DC output noise.
Is there any possiblity to add or varying the circuit to avoid those glitches?
Maybe to improve the output DC in Signal-to-noise ratio, less hum, too?
The circuit draws approx 40mA of current and the voltage Uce of the darlington IC is 15V.

Another strange thing is, approx one or two minutes after power on, there is a loud humming coming from the speakers (I use it as a preamp PSU). When loudest, it become very fast quiet again, like it was before. Maybe it has something to do with the warmup time of the preamp tubes?
How to avoid this? Will tube rectification make it better?
Many thanks for your ideas.

 

[ruffrecords]

|Glitches can be caused by switch off transients in the bridge rectifier. Try adding a 100nF film capacitor directly across the ac input to the bridge.

If you are only drawing 40mA at 300V, you really do not need to use two large inductors. You can use a simple CRC circuit prior to your capacitance multiplier and you can use much larger C values. 220uF will give you a lot lower ripple.

It is also perhaps a good idea to add a shunt resistor to the 20uF connected to the 2SD1409 emitter so it always has a dc path to ground if if no load is present.

Cheers

Ian

 

[Walter66]

Thanks, Ian. 100K as a shunt resistor for 3mA idle or higher current?

 

[ccaudle]

The output shows glitches, which seem to be of higher frequency than the random DC output noise.

Are the glitches periodic, occurring at regular intervals?

Without seeing your wiring to the 'scope I would also guess that you can pick up external interference on the scope probes. Whenever you are attempting to measure or diagnose noise it is very important to use good instrumentation wiring practice.

 

[Walter66]

Are the glitches periodic, occurring at regular intervals?

Without seeing your wiring to the 'scope I would also guess that you can pick up external interference on the scope probes. Whenever you are attempting to measure or diagnose noise it is very important to use good instrumentation wiring practice.

As can be seen in the "Random noise" picture, the spikes aren't periodic but chaotic. Had to widen the time scale to be able to see that they have a swinging voltage character added to the random noise. The breadboard wiring isn't perfect, of course, and there were even more problems when it was more imperfect. But I don't think their existence just results from bad wiring or too long wires.

 

[ruffrecords]

Thanks, Ian. 100K as a shunt resistor for 3mA idle or higher current?

100K would be fine. You could probably go higher, say 220K. Just need to watch the dissipation and choose an appropriately rated part.

Cheers

Ian

 

[Walrus]

Am I reading that scope picture correctly, and the scale is 2mV per square?
If so, the noise signal is around 3-4 mV P-P.
On a 300 Volt supply, isn't that something like -97dB?

As ccaudle says, that could be just random hash from modern lighting, wifi etc.

 

[Cqwet Dbdfte]

Some rectifier diodes are noisier than others, change to something better, like UF4007, and/or bypass each diode with ceramic caps and/or 100k 1W resistors.
Fast transients can excite resonances in transformers leakage inductance.
LC is not needed for low current, as previously stated, use wire wound resistors for added inductance. Low ESR ceramics in the 0.1uF range helps the electrolytics which are typically useless at high frequency.

 

[merlin]

I'm wondering if you've partially damaged the base of the transistor every time you switch the supply off, creating popcorn noise in the transistor. I'd put a protection diode from base to collector, and try a new transistor, just to see if the noise changes.

 

[Cqwet Dbdfte]

Hello,
my goal is to deglitch a power supply for tube electronics.

The output shows glitches, which seem to be of higher frequency than the random DC output noise.
Is there any possiblity to add or varying the circuit to avoid those glitches?
Maybe to improve the output DC in Signal-to-noise ratio, less hum, too?
The circuit draws approx 40mA of current and the voltage Uce of the darlington IC is 15V.

Another strange thing is, approx one or two minutes after power on, there is a loud humming coming from the speakers (I use it as a preamp PSU). When loudest, it become very fast quiet again, like it was before. Maybe it has something to do with the warmup time of the preamp tubes?
How to avoid this? Will tube rectification make it better?
Many thanks for your ideas.

The circuit is just a LC filter with a capacitance multiplier, provides no regualtion. Inductors are likely big, and a simple RC filter should be enough. If hum is getting thru, some other problem exist. Big chokes can also pick up unwanted magnetic fields.
Using a regulated supply is smaller and works better, broadband noise sub 500uV p-p.

 

[Walter66]

The circuit is just a LC filter with a capacitance multiplier, provides no regualtion. Inductors are likely big, and a simple RC filter should be enough. If hum is getting thru, some other problem exist. Big chokes can also pick up unwanted magnetic fields.
Using a regulated supply is smaller and works better, broadband noise sub 500uV p-p.

Once I've tried this Maida regulator with LM317 and it was technically near perfect, but didn't sound as good as this cap multiplier circuit.
Maida regulator had literally no hum.

Thought about trying the Maida with LT1085, but if it ends sounding like the 317 version it wouldn't make any sense to me.

Did I do anything wrong? It sounded like a wrong dimensioned regulator circuit. Electronically processed, flat sounding, no dynamics. Could be the wrong regulator IC? What can be tuned to bring that circuit to life? Output was 300VDC, 40mA.

 

[Walter66]

I'm wondering if you've partially damaged the base of the transistor every time you switch the supply off, creating popcorn noise in the transistor. I'd put a protection diode from base to collector, and try a new transistor, just to see if the noise changes.

It seems that those glitches are periodic, 35us means approx 30 kHz frequency. The Darlington still works, it has Uce=15V. I've killed some of them, they go into shorting mode. The circuit that I copied from don't have an additional protection diode and also works with approx 250V output. Its a commercial preamp. Certainly some kind of the same high voltage Darlington with integrated protection diode.

30 kHz, does this means something? Maybe PSU switching frequencies on the line supply? There is a Schaffner EMC filter on the AC input.

 

[Walter66]

Am I reading that scope picture correctly, and the scale is 2mV per square?
If so, the noise signal is around 3-4 mV P-P.
On a 300 Volt supply, isn't that something like -97dB?

As ccaudle says, that could be just random hash from modern lighting, wifi etc.

A good PSU should have 100dB of noise suppression at least. This one has less and the glitches are problematic. They should be eliminated.

 

[Walter66]

Are the glitches periodic, occurring at regular intervals?

Without seeing your wiring to the 'scope I would also guess that you can pick up external interference on the scope probes. Whenever you are attempting to measure or diagnose noise it is very important to use good instrumentation wiring practice.

It seems they are periodic now, 30 kHz.

 

[Cqwet Dbdfte]

Once I've tried this Maida regulator with LM317 and it was technically near perfect, but didn't sound as good as this cap multiplier circuit.
Maida regulator had literally no hum.

Thought about trying the Maida with LT1085, but if it ends sounding like the 317 version it wouldn't make any sense to me.

Did I do anything wrong? It sounded like a wrong dimensioned regulator circuit. Electronically processed, flat sounding, no dynamics. Could be the wrong regulator IC? What can be tuned to bring that circuit to life? Output was 300VDC, 40mA.

A power supply should only supply clean (DC) power. The FET+LT1085/LT317A circuit that was in LT's application manual has been in use since the late 80's, (maida what?).
Those LT parts have much more gain bandwidth than the vanilla LM317, and should provide 20dB + lower noise.
A clean low noise DC supply provides no noise, and have low output impedance that resists load induced variations, vs. a mushy RC or LC filter.
Of course resistors can be added to the output if you want some kind of sound effects generator that supports intermodulation from amplification stages.
Are load induced voltage variations in the power supply really a good thing?

 

[Cqwet Dbdfte]

Mike Maida at National Semiconductor gets credit for the original HV boost of the LM317 application 44 years ago, March 1980!
R7 is to negate the low ESR of the C2 cap. A higher ESR cap would not need it.
HV FET's were not much of thing back then, so he used a Darlington BJT.

Type Designator: NSD134
Material of Transistor: Si
Polarity: NPN
Maximum Collector Power Dissipation (Pc): 1.75 W
Maximum Collector-Base Voltage |Vcb|: 300 V
Maximum Collector-Emitter Voltage |Vce|: 300 V
Maximum Emitter-Base Voltage |Veb|: 7 V
Maximum Collector Current |Ic max|: 0.1 A
Max. Operating Junction Temperature (Tj): 150 °C
Collector Capacitance (Cc): 3 pF
Forward Current Transfer Ratio (hFE), MIN: 60
Noise Figure, dB: -
Package: TO202

LB-47 High Voltage Adjustable Power Supplies​

 

[Admiral-dk]

A VERY Cheap and efficient (not power wise) way of getting rid off those Pesky Peaks !!!!

Simply adding a single resistor in series with the Rectifier. It can be placed between the Transformer and Rectifier or Capacitor and Rectifier.
If you want it to sound like a GZ34 - use a 47R 10-11W resistor.
If Tube-Rectifier Sound isn't on the menu - then down in the 0.5-5 Ohm range - usually 2-5W resistor. In Tube Power Amps I usually place a 1.5 Ohm 4W.

This trick isn't nearly as useful in Solid-state Power Amp designs - but still works nicely for PreAmps + it helps filtering other kinds of EMC from the outside coming In through the Power Entry ....

Per

 

[Cqwet Dbdfte]

I especially like those old hollow tube wire wound power resistors that when mounted on a steel screw get added inductance.
Adding resistance+inductance in series with transformer leads will improve filtering. I see no point in having a single diode PN junction between AC winding and power ground, unless you are drawing gobs of current.
Switching power supplies add common mode chokes on inputs along with capacitance to reduce conducted emissions, and reception, to meet legal levels.

 

[Walter66]

A VERY Cheap and efficient (not power wise) way of getting rid off those Pesky Peaks !!!!

Simply adding a single resistor in series with the Rectifier. It can be placed between the Transformer and Rectifier or Capacitor and Rectifier.
If you want it to sound like a GZ34 - use a 47R 10-11W resistor.
If Tube-Rectifier Sound isn't on the menu - then down in the 0.5-5 Ohm range - usually 2-5W resistor. In Tube Power Amps I usually place a 1.5 Ohm 4W.

This trick isn't nearly as useful in Solid-state Power Amp designs - but still works nicely for PreAmps + it helps filtering other kinds of EMC from the outside coming In through the Power Entry ....

Per

Thanks for the hint, Per. Tried it with a 100 Ohm resistor and it improved a bit. But the spikes are still there. Will build it with an EZ81 and a smart construction, maybe the breadboard is responsible for the noise.

 

[Cqwet Dbdfte]

Thanks for the hint, Per. Tried it with a 100 Ohm resistor and it improved a bit. But the spikes are still there. Will build it with an EZ81 and a smart construction, maybe the breadboard is responsible for the noise.

Look at the o'scope output with probe signal and ground shorted together touching ground in the amplifier.