|
I've spent the last couple of weekends working on an open-hardware power rail oscilloscope probe: connect it to a low-impedance source via a 50¦¸ cable/termination and it applies an adjustable offset, allowing you to view ripple on relatively high voltage rails.
Commercial probes are quite expensive (~$4k), so I've taken a first pass at a low-cost version based on an . Currently it's just a schematic, but it's been a few years since I went this deep on a low-noise design (and my maths is terrible) so I'd appreciate any early review people here can offer.
You can find the schematic and analysis at?. Feedback appreciated on-list or via GitHub issue.
Kind regards |
| Patrick Coleman |
|
| Technical Director |
| BLINKENLIGHT Ltd |
| web. |
|
Well, I tried looking at the schematic you supplied, and I can't zoom in on it.? I did look at the original design, and I have at least one question.? The offset voltage is applied to the gain stage, which suggests that any noise on that signal is then amplified.? Would it not be better to apply the offset to the output stage instead?
Just leafing through his design, and ignoring most of it, when he gets to comparing op-amps, the one with the least output swing has the smallest noise figure.? Think you'd want that for the gain stage, and perhaps carefully calculate the effect of a bit more gain in the output (more swing but higher noise) stage?
I'm definitely not a low noise type, but those at least may (I hope) be decent questions.
Harvey
toggle quoted message
Show quoted text
On 7/30/2023 8:03 PM, Patrick Coleman wrote: I've spent the last couple of weekends working on an open-hardware power rail oscilloscope probe: connect it to a low-impedance source via a 50¦¸ cable/termination and it applies an adjustable offset, allowing you to view ripple on relatively high voltage rails.
Commercial probes are quite expensive (~$4k), so I've taken a first pass at a low-cost version based on an existing public design <>. Currently it's just a schematic, but it's been a few years since I went this deep on a low-noise design (and my maths is terrible) so I'd appreciate any early review people here can offer.
You can find the schematic and analysis at . Feedback appreciated on-list or via GitHub issue.
Kind regards
*Patrick Coleman*
Technical Director
BLINKENLIGHT Ltd
web. <>
Sent from Front
|
Hi, this is a pretty neato thing, if you commercialize your design I'd be interested in buying one! tx!
J
Patrick Coleman wrote:
toggle quoted message
Show quoted text
I've spent the last couple of weekends working on an open-hardware power rail oscilloscope probe: connect it to a low-impedance source via a 50¦¸ cable/termination and it applies an adjustable offset, allowing you to view ripple on relatively high voltage rails.
Commercial probes are quite expensive (~$4k), so I've taken a first pass at a low-cost version based on an existing public design <>. Currently it's just a schematic, but it's been a few years since I went this deep on a low-noise design (and my maths is terrible) so I'd appreciate any early review people here can offer.
You can find the schematic and analysis at . Feedback appreciated on-list or via GitHub issue.
Kind regards
*Patrick Coleman*
Technical Director
BLINKENLIGHT Ltd
web. <>
Sent from Front
|
Hi,
Harvey, try pressing CTRL and use the mouse wheel to zoom. The offset pot needs to be in this position to be able to adjust the input to the first op-amp close to 0V.
Patrick,
I'm no expert in low noise, but only issue I see with this design, is that the lower frequency noise is divided down and then amplified again, so any noise in the probe power supply or introduced by the op-amp will be added to the noise you try to measure. The higher frequencies will bypass the op-amps though the capacitor. In this design, there is no amplification of the noise, so you will be limited by what your scope are able to resolve.
I have been thinking about making a probe to measure power supply noise one day.
Not got around to it yet, and it might be some time before I do, but here are a few links of projects I have found, that I would use as basis:
Regards, Askild
On Mon, Jul 31, 2023 at 3:17?AM Harvey White < madyn@...> wrote: Well, I tried looking at the schematic you supplied, and I can't zoom in
on it.? I did look at the original design, and I have at least one
question.? The offset voltage is applied to the gain stage, which
suggests that any noise on that signal is then amplified.? Would it not
be better to apply the offset to the output stage instead?
Just leafing through his design, and ignoring most of it, when he gets
to comparing op-amps, the one with the least output swing has the
smallest noise figure.? Think you'd want that for the gain stage, and
perhaps carefully calculate the effect of a bit more gain in the output
(more swing but higher noise) stage?
I'm definitely not a low noise type, but those at least may (I hope) be
decent questions.
Harvey
On 7/30/2023 8:03 PM, Patrick Coleman wrote:
> I've spent the last couple of weekends working on an open-hardware
> power rail oscilloscope probe: connect it to a low-impedance source
> via a 50¦¸ cable/termination and it applies an adjustable offset,
> allowing you to view ripple on relatively high voltage rails.
>
> Commercial probes are quite expensive (~$4k), so I've taken a first
> pass at a low-cost version based on an existing public design
> <>.
> Currently it's just a schematic, but it's been a few years since I
> went this deep on a low-noise design (and my maths is terrible) so I'd
> appreciate any early review people here can offer.
>
> You can find the schematic and analysis at
> . Feedback appreciated
> on-list or via GitHub issue.
>
> Kind regards
> *Patrick Coleman*? ? ?
> Technical Director
> BLINKENLIGHT Ltd
> web. <>
>
> Sent from Front
>
|
Interesting project.?
For this function, I would use a Tek 7000 series mainframe withe
7A13 OR 7A22, although neither would go to 100 MHz.
Seems to me that if the Offset pot is towards either extreme end,
and the input V was high and of the same polarity, that the Vin +/-2V Absolute Maximum rating would be
exceeded. A pair of diodes from the - to the + input would protect
against that.? In addition, a pair of LED's (Red for 1.8V drop) on
the output of U1 could indicate an overload condition and turning
direction to correct.? An op amp with gain to the U1 output could
make this a much more sensitive indicator when the scope is not
visible while probing.
John??? KK6IL
On 7/30/2023 5:03 PM, Patrick Coleman
wrote:
toggle quoted message
Show quoted text
I've spent the last couple of weekends working on an
open-hardware power rail oscilloscope probe: connect it to a
low-impedance source via a 50¦¸ cable/termination and it
applies an adjustable offset, allowing you to view ripple on
relatively high voltage rails.
Commercial probes are quite expensive (~$4k), so I've taken
a first pass at a low-cost version based on an .
Currently it's just a schematic, but it's been a few years
since I went this deep on a low-noise design (and my maths is
terrible) so I'd appreciate any early review people here can
offer.
You can find the schematic and analysis at?.
Feedback appreciated on-list or via GitHub issue.
|
Kind regards
|
| Patrick Coleman |
|
| Technical Director |
| BLINKENLIGHT Ltd |
| web. |
|
On Sun, Jul 30, 2023 at 06:17 PM, Harvey White wrote:
Well, I tried looking at the schematic you supplied,
Are you using Kicanvas? then... CTRL down... and use the wheel (on your mouse) to zoom in out. wheel to scroll up and down SHIFT down... and use the wheel to scroll left and right
|
There is no mouse.? There is only a touchpad.? Yes, you can zoom in, but it's a bit tricky.
I did look.? I'm not sure about the power supply, I'd be tempted to perhaps do it differently.? You can get a 5 volt 200 ma isolated converter (5v-5v) which might be better.? The grounds then don't matter.? If noise is a concern, then run the converter only when the batteries are being charged.
Harvey
toggle quoted message
Show quoted text
On 7/31/2023 11:02 AM, Roy Thistle wrote: On Sun, Jul 30, 2023 at 06:17 PM, Harvey White wrote:
Well, I tried looking at the schematic you supplied,
Are you using Kicanvas?
then... CTRL down... and use the wheel (on your mouse) to zoom in out.
wheel to scroll up and down
SHIFT down... and use the wheel to scroll left and right
|
On Sun, Jul 30, 2023 at 05:57 PM, Patrick Coleman wrote:
You can find the schematic
yup... can't view the schematic with anything but your link (doesn't load anything in Altium 365, or directly on the Kicanvas Webpage.)
|
Sorry about this - I realised this afternoon that KiCanvas is still in alpha and doesn't work everywhere yet (eg, mobile!). I added an old-fashioned PDF schematic to?, see -
Hi, this is a pretty neato thing, if you commercialize your design I'd be interested in buying one! tx!
Thanks Jesse! I'll update here as soon as I've got a prototype ready (I'm hoping to slide it in with another order early next week).
Harvey, John - I really appreciate the comments. I wanted to think about them a bit more (and run the different configurations through my noise calculation spreadsheet), and I'll respond tomorrow. John, my instinct was to avoid diodes in the low-noise section but you may well be right that the input voltage is too high on one end of the potentiometer. I'll take another look.
Kind regards |
| Patrick Coleman |
|
| Technical Director |
| BLINKENLIGHT Ltd |
mob. +44 7709 045 513 |
| web. |
 On July 31, 2023 at 5:30 PM GMT+1 roy.thistle@... wrote:
On Sun, Jul 30, 2023 at 05:57 PM, Patrick Coleman wrote:
You can find the schematic
yup... can't view the schematic with anything but your link (doesn't load anything in Altium 365, or directly on the Kicanvas Webpage.)
|
Interleaved, please: On 7/31/2023 2:42 PM, Patrick Coleman wrote: Sorry about this - I realised this afternoon that KiCanvas is still in alpha and doesn't work everywhere yet (eg, mobile!). I added an old-fashioned PDF schematic to , see -
Oh, I have a copy of Kicad, and just opened the file in that. Since I really haven't used it for anything yet (EAGLE in my configuration works for me), running it on a laptop has some interesting features. Hi, this is a pretty neato thing, if you commercialize your design
I'd be interested in buying one! tx!
Thanks Jesse! I'll update here as soon as I've got a prototype ready (I'm hoping to slide it in with another order early next week).
Harvey, John - I really appreciate the comments. I wanted to think about them a bit more (and run the different configurations through my noise calculation spreadsheet), and I'll respond tomorrow. John, my instinct was to avoid diodes in the low-noise section but you may well be right that the input voltage is too high on one end of the potentiometer. I'll take another look.
My thought would be to have the gain stage without an input offset adjustment.? As long as the offset is not too bad, it can be compensated for by simply sending the pot voltage to the inverting output driver.? Noise on the pot would not be amplified by the gain stage.? This almost suggests a 20 turn trimpot to adjust a "zero" reference and then a front panel offset (you may want steps and the like). I'd also consider putting in an isolated supply running off the USB inputs.? They're surprisingly inexpensive.? Not sure how good the noise figure is, but if that were a problem, then plugin to charge, but the grounds are now isolated. Harvey Kind regards
*Patrick Coleman*
Technical Director
BLINKENLIGHT Ltd
mob. +44 7709 045 513
web. <>
Sent from Front
On July 31, 2023 at 5:30 PM GMT+1 roy.thistle@... wrote:
On Sun, Jul 30, 2023 at 05:57 PM, Patrick Coleman wrote:
You can find the schematic
yup... can't view the schematic with anything but your link (doesn't load anything in Altium 365, or directly on the Kicanvas Webpage.)
|
The originator was mystified by the fact that the measured noise was much higher than expected. The mistake in his noise estimate was assuming that the gain of the first stage is 1. While that¡¯s true for the signal path, the noise gain of that stage is about 12. So the voltage noise of the op amp is multiplied by 12, not 1.?
At least that¡¯s what I saw after quickly skimming the article. Mighta missed something along the way¡
Gary NA6O
|
LTspice may be helpful for all this ... I'm not volunteering to do it :) but I'm doing my first serious low-noise design in it now. Like any simulator, the garbage in-garbage out phenomenon is always hanging over you, but it could be instructive and help verify analytical analysis. One need not have the design-intended opamps and transistors, just something reasonably in the same realm, and swapping in & out other models can give an idea of how sensitive a component / circuit section is to overall noise levels.
Have fun ...
J
Gary Johnson via groups.io wrote:
toggle quoted message
Show quoted text
The originator was mystified by the fact that the measured noise was much higher than expected. The mistake in his noise estimate was assuming that the gain of the first stage is 1. While that¡¯s true for the signal path, the *noise gain* of that stage is about 12. So the voltage noise of the op amp is multiplied by 12, not 1.
At least that¡¯s what I saw after quickly skimming the article. Mighta missed something along the way¡
Gary NA6O
|
To revive an old thread, I wanted to give an update on the power rail probe project - I built out the prototype in October and spent the last few months characterising it. It works pretty well!
The frequency response (attached) is flat to 500MHz, as hoped, aside from a -3.7dB dip @ ~1.2MHz. That had me stumped for a long time, but after messing around in TI (TINA?) SPICE I believe it to be poor compensation in the inverting opamp U2 (see GitHub below for schematics).?
I have some ideas for improvements, but the probe works really well for day-to-day use, and the frequency response is now well characterised should you wish to take more accurate measurements. I've attached a chart of 20mV p-p noise on a 20V rail from my bench PSU, taken with my cheap 10MHz Picoscope thanks to a 20V offset applied by the power rail probe.
My plan is to release the 8 boards I have here as v1.0 (with fancy machined front and rear panels, no less), and continue iterating on the design as time allows.
There's a detailed writeup on GitHub:??(there's also a link there to purchase an assembled unit, or a blank PCB, both shipping from London, UK)
...and some more colour in this thread on Mastadon:?
I really appreciate the detailed advice and feedback from the group back in August - it influenced and improved the design. If you get in touch directly, I can add a discount to your order (anyone on the group is welcome).
Finally - I'd really like to verify the noise measurements I've taken, because I'm not sure I entirely trust my setup(/methodology) here. If you happen to have access to a good spectrum analyser and well-characterised noise source, with the ability & patience to measure noise down to say -100dBm, and would be willing to measure the noise power across a bunch of frequency bands, I'd love to get in touch. I can pay you ... in power rail probes :^)
Cheers
Patrick
|
You can save a lot of money on panels by designing the panels as double sided PCB's solid copper on both sides. The corner mounting holes as through holes with the OD of the throughole being close
to the diameter of the screw head. The just specify black soldermask with white ink. and label as needed. Works like a champ. I would recommend? backing off the copper from the board edge someting like
25 mils or so.
As an edge finish you either can use a matte black paint pen or stack up a bunch of them held together with painters tape and masked accordingly and then paint the edges with Krylon flat black spray paint. As long as they are tightly together and you just dust the edges with enough paint to cover (light coats) no paint will get between them.
Alternately you can try a black Sharpie and see how that works but one of my friends says that fades over time.?
|
Excellent writeup and project! Would buy one immed but had some significant family medical costs in november. Tucked the info away for hopefully the near future. Thank you again, this will be very useful!
J
Patrick Coleman wrote:
toggle quoted message
Show quoted text
To revive an old thread, I wanted to give an update on the power rail probe project - I built out the prototype in October and spent the last few months characterising it. It works pretty well!
The frequency response (attached) is flat to 500MHz, as hoped, aside from a -3.7dB dip @ ~1.2MHz. That had me stumped for a long time, but after messing around in TI (TINA?) SPICE I believe it to be poor compensation in the inverting opamp U2 (see GitHub below for schematics).
I have some ideas for improvements, but the probe works really well for day-to-day use, and the frequency response is now well characterised should you wish to take more accurate measurements. I've attached a chart of 20mV p-p noise on a 20V rail from my bench PSU, taken with my cheap 10MHz Picoscope thanks to a 20V offset applied by the power rail probe.
My plan is to release the 8 boards I have here as v1.0 (with fancy machined front and rear panels, no less), and continue iterating on the design as time allows.
There's a detailed writeup on GitHub: (there's also a link there to purchase an assembled unit, or a blank PCB, both shipping from London, UK)
...and some more colour in this thread on Mastadon:
I really appreciate the detailed advice and feedback from the group back in August - it influenced and improved the design. If you get in touch directly, I can add a discount to your order (anyone on the group is welcome).
Finally - I'd really like to verify the noise measurements I've taken, because I'm not sure I entirely trust my setup(/methodology) here. If you happen to have access to a good spectrum analyser and well-characterised noise source, with the ability & patience to measure noise down to say -100dBm, and would be willing to measure the noise power across a bunch of frequency bands, I'd love to get in touch. I can pay you ... in power rail probes :^)
Cheers
Patrick
|
As another thought, you can drill holes and reverse mount surface mount LEDS (gull wing ones?) on the back of the board.? Put a sheet of transparent plastic over the PC board to protect it and the LEDs.? Will also work well with non touchscreen LCD displays. Any wiring needed for the front? panel controls can be on the back side.
Harvey
toggle quoted message
Show quoted text
On 12/31/2023 2:18 PM, Sam Reaves wrote: You can save a lot of money on panels by designing the panels as double sided PCB's solid copper on both sides. The corner mounting holes as through holes with the OD of the throughole being close
to the diameter of the screw head. The just specify black soldermask with white ink. and label as needed. Works like a champ. I would recommend? backing off the copper from the board edge someting like
25 mils or so.
As an edge finish you either can use a matte black paint pen or stack up a bunch of them held together with painters tape and masked accordingly and then paint the edges with Krylon flat black spray paint. As long as they are tightly together and you just dust the edges with enough paint to cover (light coats) no paint will get between them.
Alternately you can try a black Sharpie and see how that works but one of my friends says that fades over time.
|
That's great advice. I tried something like that for a switch panel in my van, with an aluminium core PCB, but the finish wasn't great - I foolishly specified no silkscreen, hoping for a matte finish, and that was a mistake. I'll try again at some point, because it's a cheap front panel and you can reverse mount SMD LED indicators as Harvey suggested.
For this project, I wanted to test out an aluminium sheet metal manufacturer someone recommended () - the panels machined and anodised to spec came in at $3 each, which was ridiculous (about ?7 each once shipping is included for 2*10 panels). They took a while, and I'd recommend asking for sandblasting to reduce the risk of scratches on the finish, but otherwise great.
The PCBs were another experiment with a new supplier (). They were absolutely fantastic - quite expensive in small quantities, but a big step above the likes of JLCPCB. The test report received with the finished boards ran to 11 pages, and you get a big choice of prepreg, arbitrary stackups, and exotic core materials if required. Their volume pricing wasn't wild, either.
-Patrick
|
John Kolb
Sam Reaves