Yep....brute force uber allis.

By this time, I don't even know if my answer is applicable to the problem.
Maybe if I go back 2 or 3 days and read the whole Thread...

 

So lets take a step back if possible...

Is there a good way of simulating this stuff? The switcher is probably too complicated and unnecessary to simulate, but could I just replace that portion with a generated signal that contains X amount of noise (in simulation), then incorporate filter circuitry to get the desired outcome of less than 30mV? How do you guys do this?

 

Is there a good way of simulating this stuff?

No, not really; the problem we're dealing with here is one involving parasitic (i.e., "stray") inductances and capacitances, and radiated EMI. These things are extremely difficult to quantify in the form of "components" with values to give a simulator something to work with, and although I've been using SPICE for upwards of 25 years I wouldn't even want to attempt it.

The switcher is probably too complicated and unnecessary to simulate, but could I just replace that portion with a generated signal that contains X amount of noise (in simulation), then incorporate filter circuitry to get the desired outcome of less than 30mV?

In effect, that's what I did very early in this discussion when I suggested interposing an RC filter between the switching regulator and your LDO, except I chose the R and C values by hand calculation rather than using a simulator. And we saw the result-- it didn't help, because the problem wasn't where I initially thought it might be. So on to the next possible cause.

How do you guys do this?

The same way we've been doing it for the last 40+ posts: break the problem down into pieces and try to isolate the offending portion of the circuit. And keep going until you find it.

 

No, not really; the problem we're dealing with here is one involving parasitic (i.e., "stray") inductances and capacitances, and radiated EMI. These things are extremely difficult to quantify in the form of "components" with values to give a simulator something to work with, and although I've been using SPICE for upwards of 25 years I wouldn't even want to attempt it.


In effect, that's what I did very early in this discussion when I suggested interposing an RC filter between the switching regulator and your LDO, except I chose the R and C values by hand calculation rather than using a simulator. And we saw the result-- it didn't help, because the problem wasn't where I initially thought it might be. So on to the next possible cause.


The same way we've been doing it for the last 40+ posts: break the problem down into pieces and try to isolate the offending portion of the circuit. And keep going until you find it.

Thanks again for this info.

How about from a design standpoint. If you can't really figure it out until you actually get PCBs made and test them.... would it be common practice to put blank power filter parts on all of the signals in the event that you need to try a few things? How do you go about doing th initial design?

 

How about from a design standpoint. If you can't really figure it out until you actually get PCBs made and test them.... would it be common practice to put blank power filter parts on all of the signals in the event that you need to try a few things?

I don't know what common practice is, only my practice. If I were designing something containing a switching regulator-- ESPECIALLY an off-line switching regulator (i.e., one connected directly to the mains, without any intervening transformer for isolation)-- I would proceed VERY cautiously, for the sake of both safety and with concern for EMI. Switching regulators are tricky little critters, at best. At worst, they can be a nightmare. I don't know that I'd initially lay out the board with extra power filter stuff scattered all over the place, but I certainly would lay out the first PCB so that I can make changes without undue heartache.

How do you go about doing th initial design?

First, I'd anticipate initial failure: that is, I'd plan on having to iterate the PCB at least a few times, because with a switcher on board, there are bound to be some problems at first. It would be unrealistic to expect that my first PCB design is going to be the final one.

Second, I would NOT use a micropower LDO as a post-regulator for the switcher output, for reasons that I've already gone into in several previous posts. The MCP1700 is one of the worst choices possible because of its terrible PSRR, transient line regulation and transient load regulation.

Third, I would follow the switching regulator chip's manufacturer's recommended PCB layout EXACTLY to minimize radiated EMI, as well as using a PCB with power and ground planes.

And fourth, I would consider the possibility that some sort of Faraday shielding or even magnetic shielding on the PCB may be necessary.

All that said, we really, really, REALLY need to know what happens when the nRF24L01 is replaced with a passive dummy load resistor. I've said this repeatedly already, because I'm suspicious that a lot of the high-frequency "hash" noise you're seeing on the 3.3V line may be due to RF emitted by the module, either radiated or conducted. We've already determined that the spikes were being generated by the switcher, since they disappeared when we tried powering this thing from a battery plus a 7805; now we need to track down the source of the hash. PLEASE DO THIS.

 

I don't know what common practice is, only my practice. If I were designing something containing a switching regulator-- ESPECIALLY an off-line switching regulator (i.e., one connected directly to the mains, without any intervening transformer for isolation)-- I would proceed VERY cautiously, for the sake of both safety and with concern for EMI. Switching regulators are tricky little critters, at best. At worst, they can be a nightmare. I don't know that I'd initially lay out the board with extra power filter stuff scattered all over the place, but I certainly would lay out the first PCB so that I can make changes without undue heartache.


First, I'd anticipate initial failure: that is, I'd plan on having to iterate the PCB at least a few times, because with a switcher on board, there are bound to be some problems at first. It would be unrealistic to expect that my first PCB design is going to be the final one.

Second, I would NOT use a micropower LDO as a post-regulator for the switcher output, for reasons that I've already gone into in several previous posts. The MCP1700 is one of the worst choices possible because of its terrible PSRR, transient line regulation and transient load regulation.

Third, I would follow the switching regulator chip's manufacturer's recommended PCB layout EXACTLY to minimize radiated EMI, as well as using a PCB with power and ground planes.

And fourth, I would consider the possibility that some sort of Faraday shielding or even magnetic shielding on the PCB may be necessary.

All that said, we really, really, REALLY need to know what happens when the nRF24L01 is replaced with a passive dummy load resistor. I've said this repeatedly already, because I'm suspicious that a lot of the high-frequency "hash" noise you're seeing on the 3.3V line may be due to RF emitted by the module, either radiated or conducted. We've already determined that the spikes were being generated by the switcher, since they disappeared when we tried powering this thing from a battery plus a 7805; now we need to track down the source of the hash. PLEASE DO THIS.

Thanks again for these design tips. I am researching different linear voltages as we speak so I can ditch this one. I just want to find a known reliable one, with a high PSRR, and I didn't want to use an adjustable one for lower component count.

I removed the RF sensor, and soldered in a 150 ohm resistor on the RF sensors power/gnd pins to simulate a load of approximately 22mA. I was getting approximately 100mV ripple, and then I turned on the average setting on the scope, and it gave me a 30-40mV ripple at 100mHz.

 

I removed the RF sensor, and soldered in a 150 ohm resistor on the RF sensors power/gnd pins to simulate a load of approximately 22mA. I was getting approximately 100mV ripple, and then I turned on the average setting on the scope, and it gave me a 30-40mV ripple at 100mHz.

I'm stumped, then; I'm all out of ideas.

I'll repeat a suggestion I made earlier: find a local EE with experience in these things and have him get hands-on with your device. I think we've about exhausted the possibilities of remote diagnosis-- at least, remote diagnosis by me, anyway.

 

Did you ever scope groun

Thanks again for these design tips. I am researching different linear voltages as we speak so I can ditch this one. I just want to find a known reliable one, with a high PSRR, and I didn't want to use an adjustable one for lower component count.

I removed the RF sensor, and soldered in a 150 ohm resistor on the RF sensors power/gnd pins to simulate a load of approximately 22mA. I was getting approximately 100mV ripple, and then I turned on the average setting on the scope, and it gave me a 30-40mV ripple at 100mHz.

Did you ever scope ground right next to the probe ground clip?

 

Did you ever scope groun

Did you ever scope ground right next to the probe ground clip?

I did yes, completely forgot to report that. Scoping ground showed 10mV tops, it was MUCH thinner than everything else.

 

I did yes, completely forgot to report that. Scoping ground showed 10mV tops, it was MUCH thinner than everything else.

Well, that's a good thing.
Give this a read.
http://cds.linear.com/docs/en/application-note/an101f.pdf
Maybe you could sync the scope so we can see the frequency of the hash?
Then scope the input to the switcher, the linear and the output while I look at the data sheets.

 

Interesting. The switcher is very fast 500Khz. The linear doesn't do anything above 100Khz.
So maybe something like this after the switcher:
But lets see all the signals first.

 

Interesting. The switcher is very fast 500Khz. The linear doesn't do anything above 100Khz.
So maybe something like this after the switcher:
But lets see all the signals first.

Hi Ronv, this is interesting. Would you mind explaining the curves a bit more in the image? For some reason I imagined a 500Khz signal on one side, and on the other side a flat line.

@OBW0549 , you have mentioned the LM317 in the past.... what about the TLV70233? I was trying to find a voltage regulator that has good PSRR and one that doesn't require so many extra components. The TLV70233 seemed to fit the bill? Or how about the LP5907 - this one seems even better yet?

 

@OBW0549 , you have mentioned the LM317 in the past.... what about the TLV70233? I was trying to find a voltage regulator that has good PSRR and one that doesn't require so many extra components. The TLV70233 seemed to fit the bill? Or how about the LP5907 - this one seems even better yet?

Based on the curves of PSRR vs. frequency for those two parts, I'd say either one would be a major improvement over the MCP1700. Certainly, they'd be worth trying.

 

Hi Ronv, this is interesting. Would you mind explaining the curves a bit more in the image? For some reason I imagined a 500Khz signal on one side, and on the other side a flat line.

@OBW0549 , you have mentioned the LM317 in the past.... what about the TLV70233? I was trying to find a voltage regulator that has good PSRR and one that doesn't require so many extra components. The TLV70233 seemed to fit the bill? Or how about the LP5907 - this one seems even better yet?

You have it right. That is the frequency response curve so you can see how much attenuation there is.
Here is one with a 500KHz square wave in and the output after the filter. The bottom shows a step in current so you can see the response to that. It is low enough in frequency to regulate out I think.

 

One other thing. Since the frequency is pretty high the capacitors need to be ceramic not electrolytic to get the best results.

 

So I have one more question if you guys don't mind? (related to this subject)

If you have a somewhat simple circuit that is powered off a typical power adapter (say 12VDC power adapter), is there a generic filter you typically put on all of your designs at the power input of a power adapter?

 

If you have a somewhat simple circuit that is powered off a typical power adapter (say 12VDC power adapter), is there a generic filter you typically put on all of your designs at the power input of a power adapter?

Your question is a bit vague. If by "at the power input of a power adapter" you mean the power input to the adapter, i.e., where it plugs into the mains outlet, no I don't. If you mean the output from the adapter, where it plugs into my device, the answer is "sometimes." I don't have a standard or generic filter design I use when I do use a filter, but when I do it's usually something similar to what @ronv showed in post #54.

 

Your question is a bit vague. If by "at the power input of a power adapter" you mean the power input to the adapter, i.e., where it plugs into the mains outlet, no I don't. If you mean the output from the adapter, where it plugs into my device, the answer is "sometimes." I don't have a standard or generic filter design I use when I do use a filter, but when I do it's usually something similar to what @ronv showed in post #54.

Ok yeah I didn't word that very well. Where the output from the adapter plugs into the PCB.

 

You have it right. That is the frequency response curve so you can see how much attenuation there is.
Here is one with a 500KHz square wave in and the output after the filter. The bottom shows a step in current so you can see the response to that. It is low enough in frequency to regulate out I think.

I was playing around with this and the scope today. I can see the peaks drop in amplitude with the filter. Is it still acceptable for me to add bulk capacitors to this power as well? When I tried a 1000uF just to experiment, the frequency dropped down really low... so it was a long slow moving wave at this point.

 

You have it right. That is the frequency response curve so you can see how much attenuation there is.
Here is one with a 500KHz square wave in and the output after the filter. The bottom shows a step in current so you can see the response to that. It is low enough in frequency to regulate out I think.

How would I calculate one of these for a 330KHz switcher instead? Can I also still use bulk capacitors like usual or does this need to be factored in differently?