Hello,
No more comments ?
Strange?
IMHO, you can use a voltage amplifier, with a BW of 100MHz, in front of your scope.
But, the usefull BW will be the scope BW, or 40 or 60 MHz.
The best measurement tool would be the Spectrum analyzer, as someone said
previouly.
With it, and with its sensitivity you will have the fondamental frequency, harmonics and noise also.
No more comments ?
Strange?
IMHO, you can use a voltage amplifier, with a BW of 100MHz, in front of your scope.
But, the usefull BW will be the scope BW, or 40 or 60 MHz.
The best measurement tool would be the Spectrum analyzer, as someone said
previouly.
With it, and with its sensitivity you will have the fondamental frequency, harmonics and noise also.
Well, I think that the last several comments have nailed it pretty well.
I would still want to build the probe at the link Demian provided.
At work, where we do mostly RF, we have lots of scopes but mostly only use spectrum analyzers (and network analyzers).
I would still want to build the probe at the link Demian provided.
At work, where we do mostly RF, we have lots of scopes but mostly only use spectrum analyzers (and network analyzers).
Hi all,
A very restful weekend now is over and new energy for the coming week is there ...
@1audio: Thanks again for your comments & suggestions, Demian 😉
And then:
To this end: Just before the weekend I added some 0.1 uF PPS capacitors to the PSU lines on the DAC I work on and doing so attenuated the noise signal I had measured with a factor of ~40. So now it is below the level where the FFT works, yet visually there's still some ...
Adding the 0.1 uF capacitors (which I probably should have done from the beginning) also made a significant impact on the sound in terms of better detail, a more stable soundstage, more nuances, a firmer & less bloaty bass, less listening fatigue, and not least, a remarkably more nuanced timbre.
Somewhat surprising as the main peaks I measured were in the 20 - 40 MHz region (depending on sampling frequency). Yet it makes me interested in reducing it even further if I can.
(Do you BTW know of superb decoupling capacitors? I've used Panasonic ECHU's, however, their dissipation factor increases steeply above 1 MHz (0.1 uF). Might there be a technically & audibly better solution?)
@alayn91:
Well, will end now -
Best wishes for your day 😉
Jesper
A very restful weekend now is over and new energy for the coming week is there ...
@1audio: Thanks again for your comments & suggestions, Demian 😉
I'll consider this, however, wonder if it wouldn't be a useful solution to adjust this on the scope instead? With a gain of ~100 and a max output level of ~2 volts this would mean 20 mV on the input of the probe and this is well within the useful range of the scope, i.e. for these levels I don't have to use the probe (more on this below).
Thanks for the tip 😉
No I actually hadn't experimented with the scope but did so this morning. And it seems that if I e.g. feed the scope with an appr. 20 mVpp 2.5 kHz sine wave I am able to see this on the FFT up to roughly using the 2 V/div setting. However, when feeding the same 2.5 kHz sine wave to the scope at a 1 mVpp level the FFT cannot discern it at the 2 mV/div setting. So it seems that the low level resolution is an absolute one ... 🙁 ... On the other hand experimenting with this I happened to push some buttons and found out that even if low level FFTs appear to be not feasible there are many useful adjustments to the FFT function that I didn't know of ... So a beneficial exercise indeed.
Hmmm... might this be possible when buying from the US? When buying from the UK placed part of the company I have to buy 20 .... But will keep it in mind as an alternative ...
Yup, it looks interesting but considering my oscilloscope's low level FFT response I would like some amplification and as far as I can see it's got about 20 dB of attenuation (depending on the capacitor). But - as an aside - I'm impressed with the range of circuits you know of ... 🙂
And then:
A very good question I think.... My point in doing this is to be able to ask the question or "see" the question if I can put it this way. I don't yet know what to do with what I find but I reckon that knowing what to seek a reply to makes potentially answering the question easier.
To this end: Just before the weekend I added some 0.1 uF PPS capacitors to the PSU lines on the DAC I work on and doing so attenuated the noise signal I had measured with a factor of ~40. So now it is below the level where the FFT works, yet visually there's still some ...
Adding the 0.1 uF capacitors (which I probably should have done from the beginning) also made a significant impact on the sound in terms of better detail, a more stable soundstage, more nuances, a firmer & less bloaty bass, less listening fatigue, and not least, a remarkably more nuanced timbre.
Somewhat surprising as the main peaks I measured were in the 20 - 40 MHz region (depending on sampling frequency). Yet it makes me interested in reducing it even further if I can.
(Do you BTW know of superb decoupling capacitors? I've used Panasonic ECHU's, however, their dissipation factor increases steeply above 1 MHz (0.1 uF). Might there be a technically & audibly better solution?)
@alayn91:
Hi & thanks you also for commenting ... I may have used an incorrect word for this: I assume the "noise" I measure is a consequence of the pulses generated by the DAC when converting the digital data stream to an output signal.
Well, will end now -
Best wishes for your day 😉
Jesper
@marce:
Would it be something like this? See the vias in the brownish square area compared with the similar layout in the next stage.
Greetings,
Jesper
Would it be something like this? See the vias in the brownish square area compared with the similar layout in the next stage.
Greetings,
Jesper
On a spectrrum analyzer, to get the best low-level resolution and the lowest noise floor, you need to narrow the resolution bandwidth that you are using. Maybe the FFT is similar and you could try lowering the frequency span or bandwidth and get better low-level resolution.
What kind of probe, or input, are you using? If there is a large resistance, somewhere in the signal path (either series or shunt), you might even be running into basic resolution problems with the thermal noise.
One of the important reasons to care about the noise power that is generated by a high-value resistor is that it represents the limit for the smallest voltage we can resolve across the resistor, when observing a certain bandwidth.
See the seventh slide, at:
http://rfic.eecs.berkeley.edu/~nikne...pdf/lect11.pdf
For a 10k resistor, for example, the noise voltage is 13 μV (rms), for only a 1 MHz bandwidth (there's a typo on the referenced slide). And that's only the thermal noise, with or without any current flowing.
In general, at 20 deg C,
v_noise = (1.272 x 10e-10)(√(RB))
where R is resistance in Ohms and B is bandwidth in Hertz.
For large resistances and large bandwidths, the best-possible resolution gets pretty bad, pretty quickly.
For HF decoupling, what usually matters most are: 1) inductance: SMALL physical cap size/lead spacing and 2) inductance: SHORT distance between IC pin and cap (and ground).
Caps don't usually have much inductance to speak of, except due to their physical size or lead spacing (basically the same as the self-inductance of a similar-length conductor). So you would typically want to use the largest capacitance value available in the smallest SMD component size that will reach from IC pin to the ground plane or ground, and you would ideally connect the SMD cap directly between the IC power pin (right at the IC body, ideally) and a ground plane that was a fraction of a mm away. But in reality, if you cannot connect the cap within 1 to 2 mm of the IC pin, it will probably not do much good, for the highest frequencies. (On the other hand, if you have power and ground planes you can use lots of decoupling caps in parallel, and then their distance from the IC is not as critical, because the inductance tends toward 1/ncaps times each cap's inductance.)
Hi gootee,
Thanks for your input ... I've tried to follow your link but get a "404" error message. Can I ask you to post it again?
Very interesting but I don't quite get the theoretical basis for it ... To my knowledge inductance in a wire from the IC won't change even if there are many capacitors at some distance. And I wonder why ground & power planes should matter in this context ... Might you have a link to some literature where I can read about this (I'm aware of the inductance challenges at HF and also in theory the benefits of locating capacitors close to the IC)?
Best regards,
Jesper
Thanks for your input ... I've tried to follow your link but get a "404" error message. Can I ask you to post it again?
Very interesting but I don't quite get the theoretical basis for it ... To my knowledge inductance in a wire from the IC won't change even if there are many capacitors at some distance. And I wonder why ground & power planes should matter in this context ... Might you have a link to some literature where I can read about this (I'm aware of the inductance challenges at HF and also in theory the benefits of locating capacitors close to the IC)?
Best regards,
Jesper
Last edited:
Jesper,
Re-post of correct link to paper about resistor noise and minimum resolvable voltage:
http://rfic.eecs.berkeley.edu/~niknejad/ee142_fa05lects/pdf/lect11.pdf
---
Paralleling decoupling capacitor inductances:
(We wouldn't put the decoupling caps at the other end of a wire, anyway. They would need to be _RIGHT AT_ the IC power pins, in that case. But...)
No, the single wire's inductance wouldn't change. You would need to use a separate pair of wires, straight from the power/ground pins of an IC, for EACH decoupling capacitor, in order to get everything in parallel, with as little mutual inductance as possible, to have any hope of getting the 1/ncaps reduction due to paralleling the inductances of n capacitors.
But instead of lots of wire or PCB trace pairs, just use two solid copper planes. They will each have lower inductance, between any two points, than wires, anyway (and if stacked close-enough together will add some helpful HF capacitance).
It turns out that it doesn't quite give a 1/n factor to the resultant inductance. But it gets kind of close; more than halfway there, it looks like.
Here is a Bruce Archambeault paper about it:
http://www.emcs.org/acstrial/newsletters/winter10/DesignTips.pdf
And there is much more good material:
https://www.google.com/#bav=on.2,or...decoupling+capacitors+inductance&sa=N&start=0
Cheers,
Tom
Last edited:
The Henry Ott link about a 3rd of the way down is a good start.
What you are trying to do is reduce the overall inductance of the power delivery system.
When a signal switches the capacitors in the PDS act as a bucket brigade providing the charge required to switch. The 1st capacitance is the on chip capacitance, followed by: the power plane capacitance, the smallest decoupling caps, the nest size up and so on.
Vias down to planes are the best option and give the lowest overall impedance.
What you are trying to do is reduce the overall inductance of the power delivery system.
When a signal switches the capacitors in the PDS act as a bucket brigade providing the charge required to switch. The 1st capacitance is the on chip capacitance, followed by: the power plane capacitance, the smallest decoupling caps, the nest size up and so on.
Vias down to planes are the best option and give the lowest overall impedance.
Hi marce & Tom,
Thank you both for your inputs and Tom for the links. I'll look into it in the days to come and hope to learn more ...
'Best from over here (Denmark) 😉
Jesper
Thank you both for your inputs and Tom for the links. I'll look into it in the days to come and hope to learn more ...
'Best from over here (Denmark) 😉
Jesper
Hi Jesper,
after this week I am holiday, so should have more time. A couple of years ago I did some quite extensive signal integrity simulations and did a lot of work modelling scope probes to add to the stimulation, when I get a second or too I want to revisit the info as there may be some useful info for you....
I also have all my PIA (power system integrity adviser, SIV for power) notes from the course I did in Munich last year, that again may have more info.
Marc
after this week I am holiday, so should have more time. A couple of years ago I did some quite extensive signal integrity simulations and did a lot of work modelling scope probes to add to the stimulation, when I get a second or too I want to revisit the info as there may be some useful info for you....
I also have all my PIA (power system integrity adviser, SIV for power) notes from the course I did in Munich last year, that again may have more info.
Marc
Hi Marc,
Sounds interesting and I'd be pleased to learn hear more about what you find. However - please also feel free to be on vacation if you should find that this is what you would like to be ...
Best regards ;-)
Jesper
Sounds interesting and I'd be pleased to learn hear more about what you find. However - please also feel free to be on vacation if you should find that this is what you would like to be ...
Best regards ;-)
Jesper
Hi All,
I've been experimenting a bit with this HF add-on amplifier (AOA) and have now come up with a quite simple circuitry that appears to "enlarge" the look into the digital noise from e.g. a PSU when combined with an oscilloscope (which could also have an FFT/spectrum analyzer).
The bandwidth should be a bit higher than 200 MHz although I'm unable to measure this in practice. It appears to be stable. Also, I'd like to add that I haven't measured distortion etc. - the circuitry is meant to give some extra mainly qualitative information about the noise of a PSU.
FYI I've attached an image of the circuitry & also a few photos of the probe's noise level (image 1) as well as what it picks up measuring a digital PSU (image 2) compared with the oscilloscope's highest resolution (image 3).
If anyone is interested I can also post an image of the PCB layout. The PCB size is about 40 * 16 mm incl. a BNC connector.
Cheers,
Jesper
I've been experimenting a bit with this HF add-on amplifier (AOA) and have now come up with a quite simple circuitry that appears to "enlarge" the look into the digital noise from e.g. a PSU when combined with an oscilloscope (which could also have an FFT/spectrum analyzer).
The bandwidth should be a bit higher than 200 MHz although I'm unable to measure this in practice. It appears to be stable. Also, I'd like to add that I haven't measured distortion etc. - the circuitry is meant to give some extra mainly qualitative information about the noise of a PSU.
FYI I've attached an image of the circuitry & also a few photos of the probe's noise level (image 1) as well as what it picks up measuring a digital PSU (image 2) compared with the oscilloscope's highest resolution (image 3).
If anyone is interested I can also post an image of the PCB layout. The PCB size is about 40 * 16 mm incl. a BNC connector.
Cheers,
Jesper
Attachments
Update on HF probe ...
Hi All,
Here's an update from me regarding the HF probe that I've been working on on & off for some time now.
Since my last post I've made another version of the LMH6624 based probe with "on-probe" PSU and (IMHO) a very fine PCB layout (looks good ;-)). Amplification is about 140 times and the BW is a bit above 100 MHz. All in all a quite satisfying solution but also complex and time-consuming to assemble (and occasionally with a bit of oscillation it seems).
Realising that I probably wouldn't be able to easily progress further in this direction I reverted to a suggestion 1audio came up with in the beginning of the thread - a probe based on the RAM-8A+ from minicircuits. Link is here:
https://www.minicircuits.com/WebStore/dashboard.html?model=RAM-8A+
To this end I took the rare liberty of ordering some samples from minicircuits and have now built a probe that appears to work very well and also to be very simple to both make and use.
It's based on minicircuits' suggested schematic & PCB layout yet with minor modifications to the layout so as to make it suitable to a probe shape and size. For those interested I've attached an image of the layout - really quite simple I think (here it is optimized for a 0.5 mm thickness PTFE PCB from Neltec).
Amplification varies somewhat with frequency: According to the datasheet it is about 30 times (31 dBs) at 100 MHz falling to about 24 dBs at 1 GHz. I cannot measure the bandwidth with my oscilloscope (200 MHz bandwidth) but again - according to the datasheet - it should verge on 1 GHz. Using a 3.3 nF coupling input capacitor its -3 dB high pass point is 1 MHz. This, however, may be altered by changing the capacitor to a larger value (0805 size).
I've chosen to use a quite large BNC connector on the probe which besides connecting to a "normal" oscilloscope cable also works as a suitably large handle, thus making ergonomics - steering the probe on a PCB - rather straightforward.
I've attached some screendumps of what the RAM-8A+'s noise looks like on the scope (shorted input), the FFT noise (shorted input) and a couple of measurements with the scope's normal probe (x10 setting) and the RAM-8A+ probe on a digital circuit PSU - just for comparison. Please notice that the scope vertical resolution is 50 mVs using the normal oscilloscope probe whereas it is 100 mVs using the RAM-8A+ probe. In this particular case the noise is visible both on the scope and with the RAM-8A+ probe, however, this in my experience often is not the case.
One "curiosity" here is that the oscilloscope's normal probes essentially measure the same amplitudes both in x1 and x10 settings (??) so I have not verified the exact amplification level. In any case this also is less important to me as the intention of this probe is not absolute precision but the ability to better see/get an indication of low level noise that would not be visible with normal oscilloscope probes.
Should anyone be interested in building a similar probe I may be able to post Gerbers of the layout (not done it before) - or a pdf of the PCB layout.
Cheers,
Jesper
Hi All,
Here's an update from me regarding the HF probe that I've been working on on & off for some time now.
Since my last post I've made another version of the LMH6624 based probe with "on-probe" PSU and (IMHO) a very fine PCB layout (looks good ;-)). Amplification is about 140 times and the BW is a bit above 100 MHz. All in all a quite satisfying solution but also complex and time-consuming to assemble (and occasionally with a bit of oscillation it seems).
Realising that I probably wouldn't be able to easily progress further in this direction I reverted to a suggestion 1audio came up with in the beginning of the thread - a probe based on the RAM-8A+ from minicircuits. Link is here:
https://www.minicircuits.com/WebStore/dashboard.html?model=RAM-8A+
To this end I took the rare liberty of ordering some samples from minicircuits and have now built a probe that appears to work very well and also to be very simple to both make and use.
It's based on minicircuits' suggested schematic & PCB layout yet with minor modifications to the layout so as to make it suitable to a probe shape and size. For those interested I've attached an image of the layout - really quite simple I think (here it is optimized for a 0.5 mm thickness PTFE PCB from Neltec).
Amplification varies somewhat with frequency: According to the datasheet it is about 30 times (31 dBs) at 100 MHz falling to about 24 dBs at 1 GHz. I cannot measure the bandwidth with my oscilloscope (200 MHz bandwidth) but again - according to the datasheet - it should verge on 1 GHz. Using a 3.3 nF coupling input capacitor its -3 dB high pass point is 1 MHz. This, however, may be altered by changing the capacitor to a larger value (0805 size).
I've chosen to use a quite large BNC connector on the probe which besides connecting to a "normal" oscilloscope cable also works as a suitably large handle, thus making ergonomics - steering the probe on a PCB - rather straightforward.
I've attached some screendumps of what the RAM-8A+'s noise looks like on the scope (shorted input), the FFT noise (shorted input) and a couple of measurements with the scope's normal probe (x10 setting) and the RAM-8A+ probe on a digital circuit PSU - just for comparison. Please notice that the scope vertical resolution is 50 mVs using the normal oscilloscope probe whereas it is 100 mVs using the RAM-8A+ probe. In this particular case the noise is visible both on the scope and with the RAM-8A+ probe, however, this in my experience often is not the case.
One "curiosity" here is that the oscilloscope's normal probes essentially measure the same amplitudes both in x1 and x10 settings (??) so I have not verified the exact amplification level. In any case this also is less important to me as the intention of this probe is not absolute precision but the ability to better see/get an indication of low level noise that would not be visible with normal oscilloscope probes.
Should anyone be interested in building a similar probe I may be able to post Gerbers of the layout (not done it before) - or a pdf of the PCB layout.
Cheers,
Jesper
Attachments
- Status
- Not open for further replies.