Another curve tracer option to consider: http://www.fullnet.com/~tomg/gooteect.htm
Tom is actually a member on this forum (gootee).
Tom is actually a member on this forum (gootee).
BWRX said:
That is cool. the more the merrier! thanks for the link.
I had looked at that unit but it was not the right product for me for many reasons. at $600 for a kit and $1200 assembled that sucker is not cheap!
There is also a tracer made by Madel that is is $589 new and it comes assembled and has a screen built in! no need to connect to a scope.
Plus this unit does not have PC connectivity or any sort of way to store a trace or log devices. and that is the main reason i have not bought the Madel unit already.
Lets get the the Circuit-ed people and gootee together and combine forces! lets take the gootee kit and combine it with the PC finctionality of the Circuit-ed system and now you have something!
We are just so close.....
Zc
We are doing our testing and have come up with some solutions and more ideas to increase the pure USB-powered solution. A board hack of the present design can achieve around 50ma comfortably with +/-15V, which can cover a large number of small transistors.
Exclusive USB-powered tests are inherently power limiited. We can probably achieve 100-125ma @ 12-15V. USB is by specification limited to 500ma @5V on a port. Adding a wall-wart much more is possible.
After looking at the scope screens I think it becomes apparent about the advantages of using the PC to plot and evaluate curves. Here are some screen shots of CT100 curves:
NJFET multipass test using same device:
PNP test:
A Resistor using the PNP test...can you calculate the value?:
Exclusive USB-powered tests are inherently power limiited. We can probably achieve 100-125ma @ 12-15V. USB is by specification limited to 500ma @5V on a port. Adding a wall-wart much more is possible.
After looking at the scope screens I think it becomes apparent about the advantages of using the PC to plot and evaluate curves. Here are some screen shots of CT100 curves:
NJFET multipass test using same device:
An externally hosted image should be here but it was not working when we last tested it.
PNP test:
An externally hosted image should be here but it was not working when we last tested it.
A Resistor using the PNP test...can you calculate the value?:
An externally hosted image should be here but it was not working when we last tested it.
BWRX said:
Thanks, very much, for mentioning my curve tracer, here!
Zero Cool said:
I just saw this thread and haven't had time to think this through, so that I could post a comprehensive (or even intelligent, maybe) answer. But, while I'm here right now, I want to at least give some comments:
I know my curve tracer is not cheap, any more. Sorry. You might not believe how expensive some of the parts are, that I use now. I gave up on the idea of trying to do it as cheaply as possible, quite a while back, and decided to make it extremely accurate and very high-quality, instead. The switches and Molex connectors, alone, are well over $100 per unit! (However, since then, I've been toying with the idea of also producing a "cheaper" version, again, but with the much-better circuit topologies and easier-to-build layout that resulted from the latest re-design efforts. A cheaper but low-current version could also be done. I could probably also come up with a board-level "core" module that was relatively cheap.)
The next "big" design evolution, that I have had in mind for my curve tracer product, for "quite a while", is PC connectivity, and all the enhanced functionality, et al, that that might entail. Unfortunately, I got very bogged-down, for about the last year and a half or more, on the re-design of the analog stuff, or else I'd probably already have the digital parts done. (I just now finally got the last of the backordered parts for the very-latest version, and even still have a few long-backordered units to get out the door. So I should soon be able to get back up to speed for attacking "new stuff".)
I'll take a look at the Circuit-Ed information and see if I might be able to use it more-or-less directly. But it should also be fairly easy to "roll my own", otherwise. All suggestions are welcome!
BTW: I also have some relatively-easy options I could implement to increase the test voltages and currents, but don't really know how desirable that might be. Again, all suggestions are welcome.
I'll certainly try to keep up with this thread, from now on. Thanks again!
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
I showed the following screenshot in the "Matching Transistors" thread under Amplifiers/SolidState. It's a comparison between 2- 2N3906 PNP's, the first is Fairchild and the other is Motorola. By the way, it's an open offer for DIYAudio members who order and get free US shipping and discounted shipping outside of the US (do I get in trouble putting that here...
) Please note that what is available right now is the CT100 with a max sweep voltage of 0 to 15, and 22mA max current.
PNP 2N3906 Comparison Test:
) Please note that what is available right now is the CT100 with a max sweep voltage of 0 to 15, and 22mA max current.PNP 2N3906 Comparison Test:
An externally hosted image should be here but it was not working when we last tested it.
mightydub said:
This may be slightly off-topic, but, speaking of LT-Spice:
I hope that everyone here is aware of the fact that
---------------------------------------------
**** LT-SPICE CAN PROCESS WAV FILES ****
---------------------------------------------
as both inputs and outputs!
The possibilities are so intriguing that it's almost mind-boggling.
Details are in the files section (and message archive) of the LT-Spice group, at http://www.yahoogroups.com .
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
$50 USB Data Acquisition Modules
Here are a couple of places that have USB Data Acquisition modules starting at about $50, and include free software, etc:
http://www.dataq.com/usb-data-acquisition/usb-data-acquisition.html
http://www.hytekautomation.com/Products/IUSBDAQ.html
(Naturally, there are a some caveats, in each case.)
I'll need something that samples and transfers MUCH faster than those, to be able to use it with the full frequency range of my curve tracer, which can excite the DUT with up to 20 kHz sawtooth, triangle, or sine, at up to 30Vpp (and with up to 1.5A sensed across 1 Ohm), and with a synchronized staircase waveform for base/gate drive that can go to + or - 15V, with up to 150 mA that gets sensed across 100 Ohms. And the staircase waveform's steps have purposely-slowed-to-the-minimum edge times of about 2.5 us.
To preserve the extremely-good linearity of the ramps and the very low THD of the sine, etc etc, I might need to use something closer to 2 MS/s with at least 12 bits per sample, although I haven't actually calculated the requirements, yet.
Unless I'm missing something, it looks like I might want to come up with my own ADC and USB combo, maybe using an FPGA. The software on the PC may be the more-important question. But I think that if the USB part is done "properly" ("HID" protocol, IIRC), then no special drivers are needed, just something like "oscilloscope software".
It should be pretty basic, since the actual "curve tracer" functionalities are all already done, in my hardware! A "bare minimum" PC program would just need to be able to do an X-Y display of the curve tracer's X and Y outputs' data streams. (There might even be some free software that's already available that would do nicely. But, if not, I did C programming about half the time, for about 15 years. So I could probably come up with something pretty good. ;-) )
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Here are a couple of places that have USB Data Acquisition modules starting at about $50, and include free software, etc:
http://www.dataq.com/usb-data-acquisition/usb-data-acquisition.html
http://www.hytekautomation.com/Products/IUSBDAQ.html
(Naturally, there are a some caveats, in each case.)
I'll need something that samples and transfers MUCH faster than those, to be able to use it with the full frequency range of my curve tracer, which can excite the DUT with up to 20 kHz sawtooth, triangle, or sine, at up to 30Vpp (and with up to 1.5A sensed across 1 Ohm), and with a synchronized staircase waveform for base/gate drive that can go to + or - 15V, with up to 150 mA that gets sensed across 100 Ohms. And the staircase waveform's steps have purposely-slowed-to-the-minimum edge times of about 2.5 us.
To preserve the extremely-good linearity of the ramps and the very low THD of the sine, etc etc, I might need to use something closer to 2 MS/s with at least 12 bits per sample, although I haven't actually calculated the requirements, yet.
Unless I'm missing something, it looks like I might want to come up with my own ADC and USB combo, maybe using an FPGA. The software on the PC may be the more-important question. But I think that if the USB part is done "properly" ("HID" protocol, IIRC), then no special drivers are needed, just something like "oscilloscope software".
It should be pretty basic, since the actual "curve tracer" functionalities are all already done, in my hardware! A "bare minimum" PC program would just need to be able to do an X-Y display of the curve tracer's X and Y outputs' data streams. (There might even be some free software that's already available that would do nicely. But, if not, I did C programming about half the time, for about 15 years. So I could probably come up with something pretty good. ;-) )
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
Re: $50 USB Data Acquisition Modules
Cache the data and burst it out -- or you can use a much more expensive data acquisition card from national instruments or measurement computing (the latter is a subsidiary of NATI).
here's the economics:
50ksps $150 (USB Standalone)
350ksps $450- 550 (PCMIA Card)
1.2msps $1,200
if you contemplate doing USB on your own I suggest Jan Axelson's "USB Complete"
for DIY you can't beat Marty Hebel's "StampPlot Pro" -- you can use it with a PIC with ADC's if you don't want to spend $49 or a Stamp --
gootee said:
Cache the data and burst it out -- or you can use a much more expensive data acquisition card from national instruments or measurement computing (the latter is a subsidiary of NATI).
here's the economics:
50ksps $150 (USB Standalone)
350ksps $450- 550 (PCMIA Card)
1.2msps $1,200
if you contemplate doing USB on your own I suggest Jan Axelson's "USB Complete"
for DIY you can't beat Marty Hebel's "StampPlot Pro" -- you can use it with a PIC with ADC's if you don't want to spend $49 or a Stamp --
and if you want to invest in a Measurement Computing USB card you can do this -- the AOUpdate modules drive a pair of DACs and Voltage and Current are read on the AIRead Modules -- this shows a XY Graph being driven by the data -- with Softwire you can write additional tests within a module to test for boundary conditions (i.e. violating the Safe Operating Area).
To get the data into Excel or a text file each measurement point is passed to an array.
To get the data into Excel or a text file each measurement point is passed to an array.
An externally hosted image should be here but it was not working when we last tested it.
and get a screen which is going to look better than this:
An externally hosted image should be here but it was not working when we last tested it.
Re: Re: $50 USB Data Acquisition Modules
Hi, jackinnj. Thanks for the reply! And thanks fror the pointers to Axelson and Hebel.
This guy (Thomas Grocutt) did some interesting work: http://www.dsoworld.co.uk/ , and can apparently do from 1.6MS/s to 50MS/s for two 8-bit channels simultaneously (or 100 MS/s for one channel), through a standard PC parallel port. He's got a complete description and fully-developed plans and software, at that link, for a digital storage oscilloscope with a 3Mbit sample depth.
He has much more complexity and speed than I would need. So it seems like I might be able to do something simpler, and at the low end of his speed range, fairly cheaply.
It doesn't appear that 8 bit samples would be good-enough, though. That's only 256 different signal levels. (Neglecting any possible issues associated with compressing the range to suit the ADC input,) For 30Vpp, 8 bits would give a "granularity" of about 117 mV (0.4% of 30Vpp full-scale (FS)). 10 bits would give about 29mV (.097% of FS). 12 bits would give about 7mV (.023% of FS). And 16 bits would give about 460uV (.0015% of FS).
For most aspects of my curve tracer's performance, I shot for better than 0.1% temperature-stable accuracy. It would seem like a shame to degrade that by using only 8 bits. But how many bits is "good enough"? 10 or 12? The more the better?
Regarding using a PIC in the USB converter: I've been doing a little web research, over the past day or so. And so far, it "looks like" most PICS would be too slow, feeding the USB. See http://www.semifluid.com/?p=24 . But I haven't checked out the latest PICS, yet.
For the actual low-level USB interface: If I use USB, I probably wouldn't do that part from scratch. DLP Design, for example, has some USB adapter boards (listed at mouser.com), with a FIFO interface, in a 24-pin DIP footprint, that they claim can do 8 megabits per second through the USB cable to the PC, for $25 qty 1. I'm assuming it would need a small CPU or a FPGA to drive it. A PIC could work, too, if it could be fast-enough.
This link, http://scopeonpc.tripod.com/index.htm , has the hardware and software (DOS and Windows) for a VERY low-cost parallel-port-interfaced 100KS/s 8-bit oscilloscope. Total component cost is less than $5!
Surely, with a better ADC chip (and a good PCB layout, buffer amp, etc), I should be able to get 1 to 2 MS/s, with a few more bits ?? And if the DLP USB adapters work as they say, maybe it wouldn't cost much more to use USB instead of parallel (depending on what MCU/PIC/FPGA might be needed).
Any thoughts?
Thanks again.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
jackinnj said:
Hi, jackinnj. Thanks for the reply! And thanks fror the pointers to Axelson and Hebel.
This guy (Thomas Grocutt) did some interesting work: http://www.dsoworld.co.uk/ , and can apparently do from 1.6MS/s to 50MS/s for two 8-bit channels simultaneously (or 100 MS/s for one channel), through a standard PC parallel port. He's got a complete description and fully-developed plans and software, at that link, for a digital storage oscilloscope with a 3Mbit sample depth.
He has much more complexity and speed than I would need. So it seems like I might be able to do something simpler, and at the low end of his speed range, fairly cheaply.
It doesn't appear that 8 bit samples would be good-enough, though. That's only 256 different signal levels. (Neglecting any possible issues associated with compressing the range to suit the ADC input,) For 30Vpp, 8 bits would give a "granularity" of about 117 mV (0.4% of 30Vpp full-scale (FS)). 10 bits would give about 29mV (.097% of FS). 12 bits would give about 7mV (.023% of FS). And 16 bits would give about 460uV (.0015% of FS).
For most aspects of my curve tracer's performance, I shot for better than 0.1% temperature-stable accuracy. It would seem like a shame to degrade that by using only 8 bits. But how many bits is "good enough"? 10 or 12? The more the better?
Regarding using a PIC in the USB converter: I've been doing a little web research, over the past day or so. And so far, it "looks like" most PICS would be too slow, feeding the USB. See http://www.semifluid.com/?p=24 . But I haven't checked out the latest PICS, yet.
For the actual low-level USB interface: If I use USB, I probably wouldn't do that part from scratch. DLP Design, for example, has some USB adapter boards (listed at mouser.com), with a FIFO interface, in a 24-pin DIP footprint, that they claim can do 8 megabits per second through the USB cable to the PC, for $25 qty 1. I'm assuming it would need a small CPU or a FPGA to drive it. A PIC could work, too, if it could be fast-enough.
This link, http://scopeonpc.tripod.com/index.htm , has the hardware and software (DOS and Windows) for a VERY low-cost parallel-port-interfaced 100KS/s 8-bit oscilloscope. Total component cost is less than $5!
Surely, with a better ADC chip (and a good PCB layout, buffer amp, etc), I should be able to get 1 to 2 MS/s, with a few more bits ?? And if the DLP USB adapters work as they say, maybe it wouldn't cost much more to use USB instead of parallel (depending on what MCU/PIC/FPGA might be needed).
Any thoughts?
Thanks again.
- Tom Gootee
http://www.fullnet.com/~tomg/index.html
-
The 18F2550 and 18F4550 PIC's have an ADC capable of 70-80K samples per second. The dsPIC's are much higher, but they don't have a USB module.
HID is about 64KBytes per second in bulk transfer, and CDC is 420KBytes per second. This can be acheived with one of the PIC's above.
I have used several USB chips from FTDI and DLP Designs. They work quite well and they also have some IO capabilites. They may still produce their DAQ board.
TI has a very interesting USB chip designed for DAQ and DAC in the audio spectrum, with sampling at 48KHz: PCM2702. I considered it also but got stuck loving my USB PIC's. This might have some possibilites to simplify the USB connections and audio spectrum DAQ.
HID is about 64KBytes per second in bulk transfer, and CDC is 420KBytes per second. This can be acheived with one of the PIC's above.
I have used several USB chips from FTDI and DLP Designs. They work quite well and they also have some IO capabilites. They may still produce their DAQ board.
TI has a very interesting USB chip designed for DAQ and DAC in the audio spectrum, with sampling at 48KHz: PCM2702. I considered it also but got stuck loving my USB PIC's. This might have some possibilites to simplify the USB connections and audio spectrum DAQ.
wds said:
wds,
Thanks for all of the pertinent information and guidance! I'll do some digging for more details about it all, now that you've given me some much-needed clues.
One thing that I should clarify: I'm don't think that "audio spectrum" defines the needed bandwidth. The beautiful sawtooth-with-VERY-linear-ramp's _repetition_ frequency is 22 kHz. To try to preserve most of its waveform characteristics will take a much higher frequency than 22 kHz. I don't have a good spectrum analyzer, currently. And my math is a little rusty. BUT, I simulated the circuit (with LT-Spice) that produces the ramps, with most parasitics modeled, etc, and simulator's output waveform shape very-closely matching scope display of actual circuit's output. Then I did an FFT of the simulated sawtooth output waveform. IIRC, the harmonics didn't get into the -80 to -90 dB range until between 1.5 and 2 MHz. However I must confess that I don't have a very good feel for how losing everything above "Fc" MHz, or everything below some -dB level, would affect the waveforms. But, now that I think about it that way, I guess it will be relatively easy to use the existing simulation, with an added "brick wall" filter, to find out!
Thanks again!
- Tom Gootee
http://www.fullnet.com/u/tomg/index.html
-
I'm with Jack on how to utilize the USB, bulk transfer only...that's how we're doing it. The eye doesn't perceive changes faster than a few milliseconds, and that's light years of time when it comes to microprocessors to do their work. You can gather and process a lot of raw data and ship it over to the PC en masse very quickly with USB. The PC is much faster and more capable for the math than a microcontroller so it's your best friend for that level of processing.
The USB PIC's have an internal 1KByte buffer which helps to alleviate the pressure on the PIC's processing overhead. When you are sending or receiving with USB the PIC can continue data acquisition and other processing functions.
The USB PIC's have an internal 1KByte buffer which helps to alleviate the pressure on the PIC's processing overhead. When you are sending or receiving with USB the PIC can continue data acquisition and other processing functions.
since it is a curve tracer and there is some predictability to the data you can use a variety of techniques to increase the "granularity" when you need it --
the most useful aspect of the tube tracer I built is comparing pairs of tubes, comparing the two halves of 12ax7's, 6sn7's -- or just plain "dud" parts.
with respect to speed -- you also have to remember that the devices have to settle -- and you should also bookmark the temperature and timestamp the results.
the most useful aspect of the tube tracer I built is comparing pairs of tubes, comparing the two halves of 12ax7's, 6sn7's -- or just plain "dud" parts.
with respect to speed -- you also have to remember that the devices have to settle -- and you should also bookmark the temperature and timestamp the results.
Re: Re: Re: $50 USB Data Acquisition Modules
Ahh here is where the price point options kick in. I dont think 8 bits is enough, 10 maybe for a low priced option. 12 i "think" would be enough for General use. is there any reason for 16 bit accuracy? that would be a mega buck option i would bet. maybe for matching front end transistors with some advanced software.
gootee said:
Ahh here is where the price point options kick in. I dont think 8 bits is enough, 10 maybe for a low priced option. 12 i "think" would be enough for General use. is there any reason for 16 bit accuracy? that would be a mega buck option i would bet. maybe for matching front end transistors with some advanced software.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.