Multiplexed 7 Segment Noise

[TaylorD93]

Hi All,

I have a WEIR 4000T Bench PSU, 30V 1A, 80s era electronics. It has a 2x ADD3701 DVM Chips (one for Voltage and the other for Current displays) and 2x DS75492 Digit Drivers each drives a 4 Digit Common Cathode 7 segment display.

The original fault with the PSU was dead segments, so i bought some modern 7 segment units, usual 0.56" height digits and wired it up the same. So that the DS75492 is switching the Cathodes of each Digit, while the ADD3701 is switching the anode to the +5V rail of each of the segments to be illuminated. There was an array of 68R resistors in series with each of the 7 segment anodes (between the ADD3701). As the display was very very bright (almost too bright), i bumped these up to 220R and they are now a good intensity, but overall the display now work just fine.

However Im not sure if its due to the change in 7 segment module, but now i am picking up the switching noise on the output of the PSU, typically in order of 20mV Pk to pk. The worst of it is when the digits get switched (approx 625Hz), although there is still ringing on the output when it is changing between which segment to be illuminated (approx 1.6-4kHz) which is typically 12mV Pk2Pk.

The PSU layout of the Display board isnt great, so the VCC line for the ADD3701 travels almost around 2/3 of the perimeter of the board before reaching the driver IC, but fitting additional decoupling caps at each IC (10nF || 100nF) directly on the underside and also fitting wire links to try to shorten the "loop" currents from the 75492 back to the ADD3701, hasnt really helped.

The PSU can be disassembled to an extent, as it is constructed with 2 PCB's with harnesses going between them, one for the Linear regulation, and the other for the Displays. Moving the Measurement/Display PCB away, does significantly reduce the switching noise as nothing is screened, i suspect it is just getting coupled across, however im at a loss as to what else to try without rewiring the entire unit with screened cable. 

Am i chasing my tail trying to resolve this switching noise? I hoped that going to the 220R resistors from 68R would have significantly reduced the loop currents in the circuit, but it is still measurable on the output connections.
I have even gone to the extent of using some sticky back aluminium foil and covering the rear of the 7 segments as well as the wiring from the Display PCB which made little difference. But with it completely disassembled but still wired/connected together, an offcut sheet of 1.0mm Stainless Steel attenuated the ringing nearly completely from the output, so i dont believe it is conducted via the harness, but just radiated from the display PCB. But i would rather find ways to attenuate the noise at source, rather than blocking it elsewhere. I cannot measure the noise actually on the Linear Reg circuit, so i think its filtering is working correctly and i dont think that is amplifying it, but it is just noise coupled to the output.

Does anyone have any other suggestions to reducing the switching noise? Afterall, if i wanted a noisy PSU, i would have just used a Switch Mode unit instead   

Thanks

[geggi1]

How about replacing the original displays with LCD modules. You can get LCD modules from Ebay or other chinese web pages.

[floobydust]

How about replacing the original displays with LCD modules. You can get LCD modules from Ebay or other chinese web pages.

You've never enjoyed the glory of LED panel meters, and made in England no less 

OP, I would suspect your measurement, your scope grounding etc. might be taking you on a goose chase.
The display's power, where does it come from and how is it related to the main output? That's what doesn't make sense, how the two are coupled for the EMI seen.
The ADD3701 datasheet only wants a 1uF cap on the 5V reg output and says "larger makes more noise" so that is a bit odd.
With mux'ed LED displays, large bulk capacitance is needed and given the age of the PSU - how are the capacitors doing? Solid polymer are very good at reducing the display noise you are seeing. I think 10-100nF decoupling caps are not helpful for the low frequency, I also use 100-330uF there.
Last resort would be 10-47nF caps from the display board ground to the chassis ground - but I'm not sure what the isolation voltage is for the power supply, when running it output floating not connected to PE.

[DavidAlfa]

Are you sure the noise wasn't already there?
80s power supply... Those electrolitics have +30years and might need replacing.

If you still think it's the display, disconnect it and re-check the output.

[TaylorD93]

Caps have been replaced with good quality replacements. And yes ive removed the displays and the noise stops.

I did consider maybe fitting some ferrite beads to the 220R resistor legs, or some kind of series inductors to try and snub the switch spike.

[TaylorD93]

How about replacing the original displays with LCD modules. You can get LCD modules from Ebay or other chinese web pages.

You've never enjoyed the glory of LED panel meters, and made in England no less 

OP, I would suspect your measurement, your scope grounding etc. might be taking you on a goose chase.
The display's power, where does it come from and how is it related to the main output? That's what doesn't make sense, how the two are coupled for the EMI seen.
The ADD3701 datasheet only wants a 1uF cap on the 5V reg output and says "larger makes more noise" so that is a bit odd.
With mux'ed LED displays, large bulk capacitance is needed and given the age of the PSU - how are the capacitors doing? Solid polymer are very good at reducing the display noise you are seeing. I think 10-100nF decoupling caps are not helpful for the low frequency, I also use 100-330uF there.
Last resort would be 10-47nF caps from the display board ground to the chassis ground - but I'm not sure what the isolation voltage is for the power supply, when running it output floating not connected to PE.

The Display board has its own isolated supply, but does have a common reference with the linear regulator so that its A to D can measure the Current and Voltage outputs.

There is a single 220uF near one of the 75492 IC's, but nothing "local" to the ADD3701 or the 2nd 75492. Hence why i fitted the 10nF || 100nF across their VCC to DGND pins. I have a copy of the schematic somewhere, i printed it off but cant find the PDF to share. Ill see what i can do.

[floobydust]

I wouldn't try cutting noise down at the display board, other than adding +5V rail capacitors (across ADD3701 +5V pin to 75492 GND) - I would look at the wiring to the board and add ferrite beads perhaps on the ground and experiment there. The board must be grounded at the (-) output post to eliminate wiring voltage drops, but that would add noise.
As a kid, I used to put AM radios near mux'ed LED displays and the RF they made sounded so cool. Even pocket calculators of the era. So there are MHz harmonics spewing from the displays.

[Kleinstein]

The current for the LEDs is quite large and a linear supply would mainly take up the more lower frequencies. So 100 nF+10 nF are not really helping much. It would need more capacitance, more like 100 µF to really make a difference.

Ideally the display should have a separate reference: some the dual slope converters use a hold cap to invert the reference and this can produce quite some spikes on the reference. The minimum would than be some low pass filtering in the reference.
The other reason for a separate reference is to at least see if the reference is drifting.  A drifting supply is not good, but a supply that dirfts and suggests that it is stable is worse. So an independent refenerence is a good thing.

[MikeK]

Keep the 10nF || 100nF, though.  Just add the 100uF.

[Someone]

OP, I would suspect your measurement, your scope grounding etc. might be taking you on a goose chase.

Sounds like a possible explanation, the ground clip of a normal scope probe combined with its high impedance makes for a large voltage/current gain. Power supplies are usually better measured with low impedance probing. Usual check for gooses [sic], try measuring the ground in the same position!

[TaylorD93]

OP, I would suspect your measurement, your scope grounding etc. might be taking you on a goose chase.

Sounds like a possible explanation, the ground clip of a normal scope probe combined with its high impedance makes for a large voltage/current gain. Power supplies are usually better measured with low impedance probing. Usual check for gooses [sic], try measuring the ground in the same position!

Admittedly i am using the short wire from the probe to the 0V output, so not quite as good as i could.

Would it be better to do a differential measurement instead? two probes A - B ?

[inse]

Is the noise also present when you apply a load to the PSU?
If so, the regulation loop is succeptible to disturbance - that's the only explanation I could think of.

[TaylorD93]

Is the noise also present when you apply a load to the PSU?
If so, the regulation loop is succeptible to disturbance - that's the only explanation I could think of.

Yes it is still there, the amplitude of the noise does not vary with load either.

Changing Voltage makes no difference.
Varying the resistive load (tried different values from 10R to 500R) matchs the same noise figure as with no load at all.

I cannot remember if the noise was still present when the PSU enters current limit mode.

[inse]

You could investigate whether the noise is still present with the sense lines of the meters disconnected, but the meters still running.
This will (hopefully) give more evidence how the noise is coupled into the PSU.

[Terry Bites]

You might want to slow the driver output edges. How about some fine wire mesh in front of the LEDS for a real '70s look!
Ever popular: A CM choke to keep the switching noise in a local loop.
I've heard of RFI nightmare scenarios from big LED displays in sports arenas- goodbye AM.

[inse]

Why would anybody want to listen to AM radio during a sports match? :-)

[SeanB]

Strangely enough in the USA AM radio is still big, with many listeners on it, and had better coverage than FM, especially in the rural areas. Remember that Europe will have many of it's constituent countries fit into US states, and still not actually inconvenience people much, and that distances in the USA mean flying is the only option, otherwise it is a 2 to 3 day drive there, if you have at least 2 drivers, so you do not have to stop over for sleep.

[floobydust]

LED muxing can be done wrong in firmware, if you have overlap between digit drivers i.e. turn on the next digit then turn off the present digit, or don't have a delay waiting for the digit driver to shut off before turning on the next digit, it can increase EMI. You have to use a scope to make sure there is no overlap.
OP's ADD3701 already has deadtime about 50usec but not sure if it's enough. The datasheet mentions design for 300mA transients which seems out to lunch for a single LED digit being on.
Have to see the PSU's schematic or some pics to understand what the problem is. The capacitance between PE and output (-) can be anything, and important is the LED displays power wrt the output (capacitance) as well. Some PSU's use a dedicated winding just for the panel meters.