digital scope question

[Ralph Mowery]

I bought a Hantek 200 MHz digital scope and ahve been learnign to use
the a digital vers an analog.

One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

 

[jeanyves]

On 2016-09-13 16:57:17 +0000, Ralph Mowery said:

I bought a Hantek 200 MHz digital scope and ahve been learnign to use
the a digital vers an analog.

One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

it depends on the frequency of the signal you want to display
but consider that any signal higher than 1/10th of the bandwith (so 20MHz)
won't be displayed properly on a digital scope, as an analog scope will
display ok to the full bandwith.
--

Jean-Yves.

 

[Phil Allison]

jeanyves wrote:

it depends on the frequency of the signal you want to display
but consider that any signal higher than 1/10th of the bandwith (so 20MHz)
won't be displayed properly on a digital scope, as an analog scope will
display ok to the full bandwith.

** Complete nonsense.

The effective sampling rate sets the upper frequency limit for DSOs - which for the OP's scope is 1GSa/s. 200MHz is the analogue bandwidth of the input attenuation & buffer circuitry.

The problem with DSOs is that the sampling rate varies with sweep speed and this affects their ability to show high frequency detail at low speeds and also results in aliasing.

Analogue scopes have fixed bandwidth at all sweep speeds and can be trusted to
show mixed frequency signals correctly.

..... Phil

 

[JW]

On Tue, 13 Sep 2016 21:13:13 +0200 jeanyves <jeanyves@nowhere.com> wrote
in Message id: <57d84fca$0$3348$426a74cc@news.free.fr>:

On 2016-09-13 16:57:17 +0000, Ralph Mowery said:

I bought a Hantek 200 MHz digital scope and ahve been learnign to use
the a digital vers an analog.

One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

it depends on the frequency of the signal you want to display
but consider that any signal higher than 1/10th of the bandwith (so 20MHz)
won't be displayed properly on a digital scope, as an analog scope will
display ok to the full bandwith.

Got a cite for that?

 

[mike]

On 9/13/2016 1:12 PM, JW wrote:

On Tue, 13 Sep 2016 21:13:13 +0200 jeanyves <jeanyves@nowhere.com> wrote
in Message id: <57d84fca$0$3348$426a74cc@news.free.fr>:

On 2016-09-13 16:57:17 +0000, Ralph Mowery said:

I bought a Hantek 200 MHz digital scope and ahve been learnign to use
the a digital vers an analog.

One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

it depends on the frequency of the signal you want to display
but consider that any signal higher than 1/10th of the bandwith (so 20MHz)
won't be displayed properly on a digital scope, as an analog scope will
display ok to the full bandwith.

Got a cite for that?

interesting how people won't believe what they're told, but will believe
someone they reference.

Set the sweep rate to whatever you want to see.
Reduce the source frequency until the display looks like you want.
That's your answer. Nothing anybody says will change the result.
Understanding the math won't change the display.
Digital scopes suck at some things.

 

[Ralph Mowery]

In article <nr9rhs$j98$1@dont-email.me>, ham789@netzero.net says...

On 9/13/2016 1:12 PM, JW wrote:
On Tue, 13 Sep 2016 21:13:13 +0200 jeanyves <jeanyves@nowhere.com> wrote
in Message id: <57d84fca$0$3348$426a74cc@news.free.fr>:



One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

it depends on the frequency of the signal you want to display
but consider that any signal higher than 1/10th of the bandwith (so 20MHz)
won't be displayed properly on a digital scope, as an analog scope will
display ok to the full bandwidth.

Got a cite for that?

interesting how people won't believe what they're told, but will believe
someone they reference.

Set the sweep rate to whatever you want to see.
Reduce the source frequency until the display looks like you want.
That's your answer. Nothing anybody says will change the result.
Understanding the math won't change the display.
Digital scopes suck at some things.

Looks like a digital scope will not display an AM signal like the analog
one will. I finally found an explination on the internet. They just
can not take enough samples fast enough.
I even tried going to 1 MHz carrier and 500 Hz for the modulation tone
and the display was still messed up. I would have thought a scope rated
for 200 MHz would work at 10 MHz even if from China. It does do as good
or beter than the 200 MHz band width while displaying a sine wave.

I think the artical was written around 1998 and would have thought the
scopes would be beter by now, but just can't beat the math.

http://www.rocketroberts.com/techart/sigproc.htm

While the scope I bought was one of the China versions, it had good
reviews. I though it might be the China copy,but doubt any scope would
be beter , atleast in the under $ 50,000 catagory.

 

[Phil Allison]

Ralph Mowery wrote:

If I did not post it, the scope is a Hantek dso5202p which is a 200 MHz
scope and a 1 GB sample rate if only one Chanel is use. It does the sin
(x)/x thing to smooth the curve between points. The scope may blow up
but for the money it seems to be a very nice hobby scope. Bought it new
for only $ 299 from a company in the US. It does have a 3 year warranty
I feel safe with it.

I am retired now and just thought it would be nice to play with a
digital scope. I am now finding out some of the limitations of them.
Glad I still have an old Techtronics 465B analog. Seems that going to
the X-Y display is not very good either if I want to compare 2 signals
that are very close in frequency. Never could get the circle to show up
and rotate slowly.

** Nothing approaches using an analogue scope with Lissajous figures - and you can make very complicated and rather pretty ones too with the help of some phase shift networks.

Back in the 1980s I needed to check the AC supply frequency to see how it varied and not having a period counter I used my scope in X-Y mode. I also had a valve audio generator with very fine frequency adjustment ( due to the use of a 30:1 reduction drive on a large tuning gang ) and a regular frequency counter.

Producing a nice circle was dead easy but only resulted in an +/- 1 Hz precision due to uncertainty with the last digit on the counter. So I set the generator to 500Hz and steadied the pattern, giving me 0.1Hz resolution on the counter. Next I tried 5kHz and that gave me 0.01Hz accuracy - going any higher proved impractical.



..... Phil

 

[Dave Platt]

In article <MPG.32424894fd79e297989749@news.east.earthlink.net>,
Ralph Mowery <rmowery28146@earthlink.net> wrote:

Looks like a digital scope will not display an AM signal like the analog
one will. I finally found an explination on the internet. They just
can not take enough samples fast enough.

It really depends a lot on the specifics of the digital scope.

I even tried going to 1 MHz carrier and 500 Hz for the modulation tone
and the display was still messed up. I would have thought a scope rated
for 200 MHz would work at 10 MHz even if from China. It does do as good
or beter than the 200 MHz band width while displaying a sine wave.

It's important not to conflate the scope's analog bandwidth, with its
sampling rate... or to confuse a "real-time" sampling rate with an
"equivalent time" sampling rate!

Some of the less-expensive digital scopes may quote a high analog
bandwidth, but sample at a fairly poor rate... e.g. a 100 MHz scope
which samples at only 150 or 200 megasamples/second. A scope of this
sort can only let you look at fast waveforms well, using "equivalent
time" sampling, and this requires a "repetitive" waveform that doesn't
vary significantly between one scan-sweep and the next.

On the other hand, a "100 MHz" scope with a 2-gigasample rate can do a
very decent job on fast, nonrepetitive (i.e. modulated) signals.
It'll also cost more, due to the need for faster ADCs and memory.

Other tidbits:

There are a number of things that can influence how well the scope can
show you what you want to see.

If your scope is simply displaying the individual samples, and drawing
straight lines between the points, then you probably need a sampling
rate of at least 20x the carrier frequency (or more) in order to avoid
having the display confused by the "chopping off" of the tops of the
waveforms when the sampling time doesn't happen to fall right on the
peak of the waveform. 50x is even better.

Some digital scopes have the ability to interpolate... they'll do some
sort of polynomial or sin(x)/x curve fitting or filtering, and will
show you a display of what they *think* the actual waveform would have
looked like on a continuous (i.e. analog) display.

You can also run into problems where a scope deliberately reduces its
sampling rate, because it's been configured to capture a large amount
of the waveform "before" and "after" the trigger point, and let you
scroll through it. The scope has only so much memory available, and
if you tell it you want to capture a long interval of the signal, it
has to reduce the number of points it captures per (micro or
milli)second in order to fit the data in memory. This can drop the
*actual* sampling rate well below the Nyquist point, and results in
all sorts of confusing-looking aliasing of the signal on the display.
It might be worth turning down your scope's "capture width" to be no
more than can fit on the screen... and checking the display to see if
you can confirm that the *actual* data-capture rate is high enough.

Some digital scopes drop their sampling rate in half if you use both
channels. The popular Rigol DS1052E/1102E can sample at 1
gigasample/second on one channel, but only 500 gigasamples/second if
both inputs are active. If you don't need the second channel, turn it
off!

And, part of the problem is sometimes that the scope's display just
doesn't have enough horizontal resolution. If it only has, say, 640
pixels in each horizontal line (VGA resolution), or worse yet only
half that many, you're not going to see a terribly smooth waveform no
matter how fast you sample it. In addition, most inexpensive digital
scopes are sorta "colored monochrome" - they either light up a pixel
in a color, or don't, but have little or no ability to vary the
*brightness* of a pixel.

You may get better results if you use the scope to sample, then upload
the waveform data to a PC via USB or GPIB and view it on a
higher-resolution display (and/or with a plotting program that can
implement an anti-aliased display with at least 8 bits of intensity).

So, there are often things you can do to make a digital scope more
usable for signals like this... but I agree with you that an analog
scope is often a much more satisfying solution. I have both types,
and haul out the Tek 2232 (or wheel out the 7904) as often as I do any
of the digital scopes.

 

[Ralph Mowery]

In article <4ujmad-f6h.ln1@coop.radagast.org>, dplatt@coop.radagast.org
says...

In article <MPG.32424894fd79e297989749@news.east.earthlink.net>,


It's important not to conflate the scope's analog bandwidth, with its
sampling rate... or to confuse a "real-time" sampling rate with an
"equivalent time" sampling rate!

Some of the less-expensive digital scopes may quote a high analog
bandwidth, but sample at a fairly poor rate... e.g. a 100 MHz scope
which samples at only 150 or 200 megasamples/second. A scope of this
sort can only let you look at fast waveforms well, using "equivalent
time" sampling, and this requires a "repetitive" waveform that doesn't
vary significantly between one scan-sweep and the next.

On the other hand, a "100 MHz" scope with a 2-gigasample rate can do a
very decent job on fast, nonrepetitive (i.e. modulated) signals.
It'll also cost more, due to the need for faster ADCs and memory.

Other tidbits:

There are a number of things that can influence how well the scope can
show you what you want to see.

If your scope is simply displaying the individual samples, and drawing
straight lines between the points, then you probably need a sampling
rate of at least 20x the carrier frequency (or more) in order to avoid
having the display confused by the "chopping off" of the tops of the
waveforms when the sampling time doesn't happen to fall right on the
peak of the waveform. 50x is even better.

Some digital scopes have the ability to interpolate... they'll do some
sort of polynomial or sin(x)/x curve fitting or filtering, and will
show you a display of what they *think* the actual waveform would have
looked like on a continuous (i.e. analog) display.

You can also run into problems where a scope deliberately reduces its
sampling rate, because it's been configured to capture a large amount
of the waveform "before" and "after" the trigger point, and let you
scroll through it. The scope has only so much memory available, and
if you tell it you want to capture a long interval of the signal, it
has to reduce the number of points it captures per (micro or
milli)second in order to fit the data in memory. This can drop the
*actual* sampling rate well below the Nyquist point, and results in
all sorts of confusing-looking aliasing of the signal on the display.
It might be worth turning down your scope's "capture width" to be no
more than can fit on the screen... and checking the display to see if
you can confirm that the *actual* data-capture rate is high enough.

Some digital scopes drop their sampling rate in half if you use both
channels. The popular Rigol DS1052E/1102E can sample at 1
gigasample/second on one channel, but only 500 gigasamples/second if
both inputs are active. If you don't need the second channel, turn it
off!

And, part of the problem is sometimes that the scope's display just
doesn't have enough horizontal resolution. If it only has, say, 640
pixels in each horizontal line (VGA resolution), or worse yet only
half that many, you're not going to see a terribly smooth waveform no
matter how fast you sample it. In addition, most inexpensive digital
scopes are sorta "colored monochrome" - they either light up a pixel
in a color, or don't, but have little or no ability to vary the
*brightness* of a pixel.

You may get better results if you use the scope to sample, then upload
the waveform data to a PC via USB or GPIB and view it on a
higher-resolution display (and/or with a plotting program that can
implement an anti-aliased display with at least 8 bits of intensity).

So, there are often things you can do to make a digital scope more
usable for signals like this... but I agree with you that an analog
scope is often a much more satisfying solution. I have both types,
and haul out the Tek 2232 (or wheel out the 7904) as often as I do any
of the digital scopes.

If I did not post it, the scope is a Hantek dso5202p which is a 200 MHz
scope and a 1 GB sample rate if only one Chanel is use. It does the sin
(x)/x thing to smooth the curve between points. The scope may blow up
but for the money it seems to be a very nice hobby scope. Bought it new
for only $ 299 from a company in the US. It does have a 3 year warranty
I feel safe with it.
I am retired now and just thought it would be nice to play with a
digital scope. I am now finding out some of the limitations of them.
Glad I still have an old Techtronics 465B analog. Seems that going to
the X-Y display is not very good either if I want to compare 2 signals
that are very close in frequency. Never could get the circle to show up
and rotate slowly. It will show up if I hook both probes up to the same
signal and an elips due the the slight delay between the signal.

 

>"Producing a nice circle was dead easy but only resulted in an +/- 1 >Hz precision due to uncertainty with the last digit on the counter. >So I set the generator to 500Hz and steadied the pattern, giving me >0.1Hz resolution on the counter. Next I tried 5kHz and that gave me >0.01Hz accuracy - going any higher proved impractical. "

Interesting usage. Let me guess, this was because back then most clocks either had a synchronous motor and even most digital ones used the power line frequency as a clock.

That is the only reason I can think of, if it was something else do enlighten me. I know there is alot of weird shit out there. We had TVs build with an SCR regulator which of course also rectified, and they weren't picky so much about frequecy, even triac light dimmers aren't all that picky. Of course they might be a little different for 50Hz but then it is probably a different set of components anyway for the 240 volt mains in such countries. I do know that in the old days power companies had to adjust because under heavy load the generators would slow a bit, so they sped them up a bit to compensate during low load periods, pretty much so the clocks kept time.

I was in an interesting position at my shop there was a ground fault in the heating system but I was getting 3,500 square feet for like $350 a month which is dirt cheap even then, and that included heat. My shop had CEI which is on the TVA grid and the factory where the boiler was was on CPP which is on the PASNY grid and they were not synched. Sometimes I would pick up like 330 volts. I know it was dangerous but we were nuts back then and just put the benches where you wouldn't get shocked. You would have to barefoot and spill something. The reason was obvious, because the load patterns were different. The TVA is south and PASMY is north. This is one of the few cities where people in some areas actually have a choice. Generally the require that choice be made on a buildingwide basis and I guess I see why, but somehow they had both in that building. (in the back they had a plastic injection molding plant making parts for Mr. Coffee, of course that kind of industry is gone now)

Anyway, I never bothered to put a scope on it but I would guess that for 60 Hz it never varied more than +/- 0.5 Hz. The only things that would require that accuracy would be clocks. Well maybe high fidelity turntables. Back in them days even my favorites, Duals used a synchronous motor and a tapered motor shaft for speeed control. However, the speed indicator strobe light ran off the power line so if it was off, your speed was off. (my ?Uncle used to use a stopwatch and put a piece of masking tape on the platter and sit there and count the revolutions, some people are picky about such things ad also made his olady strain the gravy)

 

>"Glad I still have an old Techtronics 465B analog. "

Despite whatever gripes you might have about that scope, it is one of the prime old Teks to have. We have one and played with it and it has the ability to display both A and B sweep simultaneously. Very few scopes have that ability as far as I know. Now this one stopped triggering B sweep, I suspect Q4565 but have had other things on my mind. Plus the way they activate it by driving the emitter I don't want to throw just any transistor in there, except maybe for test. The pulse is there on the emitter, the bias is on the base, but the collector is a flatline, at a higher voltage of course. I suspect I am going to need a really fast transistor for that power range which isn't that high, but it has to ahndle a pretty high base to emitter current, so just a higher hfe, or HFE isn't going to cut it. they generally don't want high base currents, that is the whole idea of higher gain ! I would consider a high speed power switching transistor but it is socketed and I do not want to maul it. And I really don't want to modify it.

I'd like to keep it as stock as possible, it is in decent cosmetic condition except for those leggy thing at the back so it stands up on the floor, apparently they saw some use. But all the knobs and pots and switches are fine and the trace it sharp with no screen burn.

So it is on my bench all apart, but it is not holding up any other jobs, there is an old Pioneer receiver I need to order a bunch of transistors for but there is not rush on that. Other than that I just go to work, fukit. they are slowing down so maybe I'll get back on it. I can probably scare up a transistor for it around here once I do some math. Maybe. I just hate ordering things that only cost like thirty cents.

 

" I have both types,
and haul out the Tek 2232 (or wheel out the 7904) as often as I do >any of the digital scopes. "

Yeah, "wheel out", you got that right. We got a 7834 and I think they weigh about the same. If you don't have a scopemobile or something similar, get rid of it. Or hire some kid to move stuff around.

Actually I got an old 561A in the garage on top of a stand for a little buzzbox fluxcore welder. I need to fix that. It works but has a HV arc and if not plugged into a grounded outlet shocks you. Cool little toy for the Man cave out there. Yup, we got a propane heater for the winter and a stereo with a CD player. Actualy and a mixing board and microphone as well. But none of us have gotten drunk enough to really enjoy it lately...

We also have a bunch of 7000 series plugins, some quite coveted, which might go on the block soon if anyone is interested. We don't have any spectrum analyser or that type of stuff, but there is a 1 GHz plugin that exceeds the limits of the 7834. If you want it make offer, and we do barter. That one is strictly 50 ohm input. There is also a differential amp that can take some serious offset and has an astronomical CNRR. Got a few of the good horizontal plugins as well but I think a couple of them don't work right. But there are like 13 of them in the collection and the scope only takes four. So (LOL) I figure we need no more than eight.

Would definitely trade a bunch of them and other stuff for the spectrum analyser. IF you, or anyone is interested just post and I'll make a list of what they are and their condition. I think there are a couple that work fine and have a broken knob or something, but a knob can usually be taken off a cheaper one.

But you may consider me a Tek CRO afficianado. I will not say fan because that word is based on the word phanatic, I am not a fan of anything. It's just I have never seen scopes that work as well as those old Teks.

 

>"If your scope is simply displaying the individual samples, and >drawing straight lines between the points, then you probably need a >sampling rate of at least 20x the carrier frequency (or more) in >order to avoid having the display confused by the "chopping off" of >the tops of the waveforms when the sampling time doesn't happen to >fall right on the peak of the waveform. 50x is even better. "

Sure, but where is the button for that ?

 

On Tuesday, September 13, 2016 at 12:57:10 PM UTC-4, Ralph Mowery wrote:

I bought a Hantek 200 MHz digital scope and ahve been learnign to use
the a digital vers an analog.

One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

Try a single sweep on the DSO that may look better. (or not.)

George H.

 

[JW]

On Tue, 13 Sep 2016 14:35:54 -0700 mike <ham789@netzero.net> wrote in
Message id: <nr9rhs$j98$1@dont-email.me>:

On 9/13/2016 1:12 PM, JW wrote:
On Tue, 13 Sep 2016 21:13:13 +0200 jeanyves <jeanyves@nowhere.com> wrote
in Message id: <57d84fca$0$3348$426a74cc@news.free.fr>:

On 2016-09-13 16:57:17 +0000, Ralph Mowery said:

I bought a Hantek 200 MHz digital scope and ahve been learnign to use
the a digital vers an analog.

One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

it depends on the frequency of the signal you want to display
but consider that any signal higher than 1/10th of the bandwith (so 20MHz)
won't be displayed properly on a digital scope, as an analog scope will
display ok to the full bandwith.

Got a cite for that?

interesting how people won't believe what they're told, but will believe
someone they reference.

Sure, because that was a ridiculous statement. I'd like to see cite from a
reputable source that would back that up.

Can't find one, can you? Maybe jeanyves can?

Bet not.

Set the sweep rate to whatever you want to see.
Reduce the source frequency until the display looks like you want.
That's your answer. Nothing anybody says will change the result.
Understanding the math won't change the display.
Digital scopes suck at some things.

I wouldn't use a Chinese scope for anything. They just can't be trusted.

 

[root]

JW <none@dev.null> wrote:

[crickets]

Completely off Topic:

I laughed at your response, then I started to think about whether
[crickets] means the same thing to you that it does to me.

I take it as a reference to a Bugs Bunny/Daffy Duck cartoon
about a contest to see which was funnier. Is that your
reference?

 

[JW]

On Wed, 14 Sep 2016 07:11:26 -0400 JW <none@dev.null> wrote in Message id:
<1sbitbh4o4knjiima9j84s9qc8n3et4iq2@4ax.com>:

On Tue, 13 Sep 2016 14:35:54 -0700 mike <ham789@netzero.net> wrote in
Message id: <nr9rhs$j98$1@dont-email.me>:

On 9/13/2016 1:12 PM, JW wrote:
On Tue, 13 Sep 2016 21:13:13 +0200 jeanyves <jeanyves@nowhere.com> wrote
in Message id: <57d84fca$0$3348$426a74cc@news.free.fr>:

On 2016-09-13 16:57:17 +0000, Ralph Mowery said:

I bought a Hantek 200 MHz digital scope and ahve been learnign to use
the a digital vers an analog.

One problem I seem to have is trying to display an AM signal and getting
it to look like my Analog scope when setting it for 100 % modulation.
Out of a signal generator into my analog it looks like it should, but I
havenot been able to get the digital to show anywhere close to what it
should.

Does anyone have any ideas on what I may be missing or are the digital
scopes not set up to do that.

it depends on the frequency of the signal you want to display
but consider that any signal higher than 1/10th of the bandwith (so 20MHz)
won't be displayed properly on a digital scope, as an analog scope will
display ok to the full bandwith.

Got a cite for that?

interesting how people won't believe what they're told, but will believe
someone they reference.

Sure, because that was a ridiculous statement. I'd like to see cite from a
reputable source that would back that up.

Can't find one, can you? Maybe jeanyves can?

Bet not.

[crickets]

 

[JW]

On Thu, 15 Sep 2016 16:03:36 -0000 (UTC) root <NoEMail@home.org> wrote in
Message id: <nregoo$35j$1@dont-email.me>:

JW <none@dev.null> wrote:

[crickets]

Completely off Topic:

I laughed at your response, then I started to think about whether
[crickets] means the same thing to you that it does to me.

I take it as a reference to a Bugs Bunny/Daffy Duck cartoon
about a contest to see which was funnier. Is that your
reference?

https://www.youtube.com/watch?v=d0mNkix4Xnk