USB digital scope

[Helmut Sennewald]

Hello Davide,

Thanks for this info. I am interested to see the hardware design.
Could you please send me the schematics/photos. I promise not
to use it commercially.

Best regards,
Helmut

----- Original Message -----
From: "Helmut Sennewald" <[email protected]>
Newsgroups: sci.electronics.design
Sent: Monday, December 05, 2005 11:49 PM
Subject: Re: USB digital scope

Hello Davide,

Thanks for this info. I am interested to see the hardware design.
Could you please send me the schematics/photos. I promise not
to use it commercially.

Best regards,
Helmut

Hello Davide,

Thanks for sending your schematic to me. I have taken a look to it.

I have read in another message from you that this circuit
can save 1k-samples per channel.

Is it possible to have a higher number of samples stored
with a pin-compatible FPGA?

Best regards,
Helmut

 

[Davide]

Hi Helmut,
it is possible to fit in the same pcb the ACEX1K50 in place of the
1K30: in this case the memory for each channel is increased to 1.5K
samples.

Davide

 

[Bob Smith]

Dear Bob,
thanks for your offer. I saw the linuxtoys.org website, and this is
very interesting. I'm thinking about the functions to include in the
applicative. I'm not a software specialist and the "XMLHttpRequest()
for your interface..." is not very clear for me....

XMLHttpRequest() is a JavaScript subroutine (so it runs
on the web browser) that monitors the web server for new
data. It does this without polling so it is very fast.
This routine is the core of what is called AJAX and is
what makes sites like Google maps work.

A web interface is just one possibility. We could also
to a TCL interface if you wish. We probably can't do
native X-windows or MS-windows interfaces since you'll
almost certainly want a portable interface.

Bob

 

[Joel Kolstad]

The PCB is intended to mount also two 12 bit 50MSPS A/D converter (I do
not yet tested this A/D converter but I hope this works fine...) in
place of the two 12 bit 20MSPS; The PCB has an "expansion port" for a
daughter board for:

3. To stack multiple board in order to have 4,6 or 8 channel scope.

I've thought about doing something like this myself on and off, but there's a
couple of challenges with the approach:

-- Aligning all the paths from one box to the next pretty much implies that
you need to generate a common clock somewhere and distribute it to all the
individual boxes with some known relation from box to box. If you do restrict
people to physically plugging one box into the other (or connecting them only
with cables of known lengths that you've provided), this is not difficult.
-- You need to distribute a high speed "trigger" signals to all the boxes so
they know when to stop capturing. That's not too hard, but what's harder is
that you have to consider whether or not you want every box to be able to
trigger or whether you just want to have a specialized "trigger" box... the
problem is that, while repetitive random-time sampling is great (with
repetitive signals), your trigger circuitry has to operate at the "equivalent
time" rate, i.e., much faster than the ADCs themselves.
-- With digital (logic analyzer) boxes, it becomes challenging to start
coordinating triggers when the trigger event is something like, "bit pattern
1010xxxx on box #1, then 00xxxx00 on box #3, back to xx0xxxxx on box #1,
etc..." (And this is a very simple trigger event compared to what contemporary
logic analyzers can do.)

At the end of the day I figured it's be easier just to design one big PCB and
populate it differently based on the options a user wants, just like the big
boys do. I think there's a huge market out there for a USB-based
instrument that implements most of what something like a Tektronix TDS3054
does -- 4 analog channels, 500MHz analog response, 5GSsp. With '3054s going
for $10k, if you could profitably sell a USB-based version for, say, $2500,
you'd have more orders than you could fill, I imagine!

The Bitscope guys (www.bitscope.com) out of Australia have been quite
successful, even though their products are nowhere near as fancy as a TDS3054
(they're basically 100MHz/100Msps with ONE channel and something FAR less with
two channels).

---Joel Kolstad

 

[Davide]

Dear Joel,
thanks for your considerations: you are right, the channel
synchronization is VERY difficult...I approached this problem in this
way:

1. I distribute the clock signal (20MHz) from the first scope board
(channel 1 and 2) to my second board (channel 3 and 4). The clock goes
only to fpga (then the fpga distribute them to the A/Dcs), and for this
the capacitive loading on this wire is quite low.
2. I have a bidirectional wire, the trigger, that comes from first
board to the second if the trigger is on channel 1 or 2 or from the
second board to the first if the trigger is on channel 3 or 4 (this is
set by the applicative program). The trigger is a "normal" signal
synchronous with the 20MHz clock signal: this will generate the sweeps
with exactly the same start point and the same end point (referred to
the 20MHz clock cycle).
3. The equivalent time sampling will be rebuilt after this phase,
calculating with interpolation the trigger instant on a sub-cycle basis
(interpolation) and shifting the overall waveform of this fractional
value. For this the waveforms from the different boards needs to be
synchronized at the clock cycle and not at higher frequency.

Finally, this scope has not very high features but from my point of
view is quite cheap (about 100€ for components) and easy to use
(powered by USB only). Another very important point (always from my
point of view) is the "teaching" peculiarity of this instrument: you
can change the DSO firmware (FPGA) directly from the USB with no
programmers or other things, only Quartus and a USB cable.

hi,

Davide

 

[Davide]

You're perfectly right. My goal was a 100$ instrument and not 1000$
instrument!!! The comparison is not from instruments in the same
"segment".

 

[Davide]

I'm sorry, I'm not sure that the DLL works in this envronment. I
checked the DLL only with delphi and Visual Basic. I think this can
work also with Visual C++.

The A/D is the ADS805 or ADS807 from TI. (the price is the same and the
second has 50 MSPS smaple rate).

hi,

Davide

 

[Rene Tschaggelar]

Check out the cleverscope for a nice USB scope that has 4MB of sample
memory (SDRAM). It trys to compete with the Agilent mixed signal scopes
and actually doesn't do too bad a job:
http://www.cleverscope.com/
The software (Labview) lets you do lots of clever custom programmable
maths directly on the waveforms too.

Labview is the killer. Did you ever spend a month
trying to bring the wires through some silly walls
while knowing with some decent compiler it would
have been a matter of minutes ?

Rene

 

[David L. Jones]

Labview is the killer. Did you ever spend a month
trying to bring the wires through some silly walls
while knowing with some decent compiler it would
have been a matter of minutes ?

Yup!

I love it how supposed "specialists" can't even explain how their
Labview "code" works, nor can they fix the damn thing 12 months later
because they can't understand what they did. Love it how you can't
readily "peer review" it either.
And NI push it like it's the greatest thing since sliced bread and what
wonderful benefits it has for our organisation, all we have to do is
send every one of our programmers on the Labview BASICs course and
we'll all be instant experts :->

Dave

 

[David L. Jones]

You're perfectly right. My goal was a 100$ instrument and not 1000$
instrument!!! The comparison is not from instruments in the same
"segment".

Yes, you are right.
It is a great design for the price, although the lack of memory really
cripples it, regardless of how cheap it is. I suspect more people would
happily pay double or triple the cost for say a 1MB scope, or even 10KB
worth of sample memory.

How much does a few MB of SDRAM cost in order to get a vast improvement
in functionality?
I've seen inside the CleverScope, and it is not much more than 2 SDRAM
chips, an FPGA, and the input circuitry. Pretty pricey, but volumes are
low and it fills a niche.

Dave

 

[David L. Jones]

I've thought about doing something like this myself on and off, but there's a
couple of challenges with the approach:

-- Aligning all the paths from one box to the next pretty much implies that
you need to generate a common clock somewhere and distribute it to all the
individual boxes with some known relation from box to box. If you do restrict
people to physically plugging one box into the other (or connecting them only
with cables of known lengths that you've provided), this is not difficult.
-- You need to distribute a high speed "trigger" signals to all the boxes so
they know when to stop capturing. That's not too hard, but what's harder is
that you have to consider whether or not you want every box to be able to
trigger or whether you just want to have a specialized "trigger" box... the
problem is that, while repetitive random-time sampling is great (with
repetitive signals), your trigger circuitry has to operate at the "equivalent
time" rate, i.e., much faster than the ADCs themselves.
-- With digital (logic analyzer) boxes, it becomes challenging to start
coordinating triggers when the trigger event is something like, "bit pattern
1010xxxx on box #1, then 00xxxx00 on box #3, back to xx0xxxxx on box #1,
etc..." (And this is a very simple trigger event compared to what contemporary
logic analyzers can do.)

At the end of the day I figured it's be easier just to design one big PCB and
populate it differently based on the options a user wants, just like the big
boys do. I think there's a huge market out there for a USB-based
instrument that implements most of what something like a Tektronix TDS3054
does -- 4 analog channels, 500MHz analog response, 5GSsp. With '3054s going
for $10k, if you could profitably sell a USB-based version for, say, $2500,
you'd have more orders than you could fill, I imagine!

I suspect not.
Those after that sort of performance would most likely not want a USB
based scope, they want a "real" scope they can play with on the bench.
500MHz analog bandwidth is incredibly difficult to do, and at 1GS/s+
you need to roll your own ADCs which is what the big scope
manufacturers do. That takes massive R&D time and $$$$$$$
You can't just slap together a bunch of parts from Digikey and get
anywhere near that performance.

The Bitscope guys (www.bitscope.com) out of Australia have been quite
successful, even though their products are nowhere near as fancy as a TDS3054
(they're basically 100MHz/100Msps with ONE channel and something FAR less with
two channels).

They aren't as fast as the real scopes because you can't go much above
that with off-the-shelf parts.

Dave

 

[Joerg]

Hello Riscy,

I think it do need 12 bits, it give better range and resolution to do
some post calculation.

Yes, if you want to do FFT and stuff like that. But I thought this
thread was more about the scope nature of a digital scope. When you want
to use it to track down runt pulses, phase jitter, setup and hold
violations and the other usual stuff the ADC needs to be as fast as can
be. Even if that comes at the expense of bits. I have done quite some
work with 6-bit digitizers because that's all we had in them days when
you needed more than 100MSPS. We actually needed 400MSPS. The ENOB up
there had probably been 4 bits or less but it did the job.

Regards, Joerg