I need 3000 Hz Output sampling in MCUs

[PCB.Wiz]

So I have to rewrite the firmware to turn it into a higher frequency EMG after I figure out how the HCTO4, HC4052M is connected to it. Has anyone tried rewriting firmware of existing hardware? Is it easy? 

I have already read the firmware of the chip via 4 pin JTAG and MSP Flasher, even ran it at CSS and IAR and saw the assembly language and even single stepped it. But it's 27k of firmware and very complex.

You have answered your own question there.

But first I need to know if the MSP430F2617 with serial pins can even output 3000 Hz to the FT232 chip.

If you cannot extract that from the datasheet, I'm not sure creating your own firmware is within your abilities ?
As already said, 3000Hz is 'slow' , the serial port speed is not an issue.
The ADC is likely to be more of an issue, I see the ADS124 specs

• 24 BITS—NO MISSING CODES
  19 BITS EFFECTIVE RESOLUTION UP TO 20kHz DATA RATE
  LOW NOISE: 1.8ppm
  FOUR DIFFERENTIAL INPUTS
  INL: 15ppm (max)
  EXTERNAL REFERENCE (0.5V to 5V)
  

That's quite impressive. 
19 bits of useful info can pack into 3 bytes which needs 9000 bytes/sec or 90k Baud.
You have not mentioned channel count, but the image shows 4 opamps, so do you intend to stream 4 channels at 3kHz ?.  That's then > 360kBaud of data flow.

The modern USB-UARTS can manage 12M/N or 24M/N for UART speeds, the common baud rates like 115200 are simply historic suggestions. 100k and 120k and 500k will also be possible. (etc)

So there will be many possible baud rates that both MSP43 and UART can manage.
Check the resonator on your board and use the datasheet to calculate possible baud rates.

The EEG app can only accepts up to 256 Hz.

What limits that ?

I need to find app that can show waveforms up to 3000 Hz. But that's another day.

Do you need real time display, or do you intend to capture then graph ?

google finds items like
https://github.com/xioTechnologies/Serial-Oscilloscope
and that links to
https://www.oscilloscope-lib.com/
and also there is this
https://github.com/ccrause/LazScope


If you use ASCII CSV, that will be slow to convert inside the MSP430, and also wastes link bandwidth, 3 bytes becomes ~ 8.25 bytes, so you need something over 1,11Mbd

It maybe easier to tweak the PC side CSV ascii parser to accept a simpler/faster 21 data bits packed into 3 bytes using MSB tags. That is very fast and simple inside MSP430.

3 byte sets of MSB's of [100] can tag channel Zero frame and [110] can increment the channel, leaving you one spare tag of [101] (for things like timebase/scaling?), this simple MSB tag encode gives channel sync information too.

eg A 4 channel capture frame of 21 data bits, would be streamed as 12 bytes MSB repeating as [100][110][110][110][100][110][110][110]....

[ataradov]

Any MCU can interface this device. You can use SPI, or at this speed you can easily just use regular GPIOs.

I would not worry about the noise, make sure it is functional first. And then transitioning to real PCBs would be necessary for the final debugging anyway.

[uer166]

There is no chance you will get MCU with a 24 bit ADC that actually works well. If you need any sort of analog precision, you have to use external ADC and good analog design.

Eeeh, I agree in general but there are always exceptions, here's something we used successfully with 3 24 bit S-D for revenue grade metering: https://www.ti.com/lit/ds/symlink/msp430f67641a.pdf?ts=1673831875420&ref_url=https%253A%252F%252Fwww.google.com%252F.

[ataradov]

here's something we used successfully with 3 24 bit S-D for revenue grade metering:

With best case SINAD=89 dB, it is barely 15 effective bits. This performance does happen in MCUs. But it is far from what you can get from standalone devices.

[PCB.Wiz]

Ok. What MCU is easiest to interface with the 24 bit ADC ADS1254E?
Also if I put everything in breadboard. What noises would be introduced by using breadbroad only?

Avoid breadboards !

If you want to avoid reverse engineering, you may be smarter to work backwards, and search Digikey etc for Eval Boards of suitable ADCs.  Those already have chips mounted and are tested.
TI and ADi will have plenty, in SPI and i2c

Ones like this, are even already medically focused : https://www.ti.com/tool/ADS1292ECG-FE


Besides UARTS there are now available also i2c and SPI USB Bridge parts, so you could also research if no remote MCU at all is needed !
You might manage this entirely from the PC side.

[pcprogrammer]

I would like to add something here in light of the original posters seemingly inexperience with it all, the thread is about a medical device, and there are special rules to take into account.

When powered from mains voltage it needs proper insulation between the circuits and the electrodes attached to the test subject. This is not really needed when it is battery powered and can't be connected to a mains voltage in any way, this also means a connection to a computer via USB.

Before even starting with a design of any electronics I would make sure to know about all the regulations surrounding this type of device.

[planc]

The EEG device has 4000V  H11L1 opto isolator (white chip) to the electrodes. The USB socket is the silver at the bottom.

It's all difficult to trace so I think I'll just get the ADS1292ECG-FE suggested by PCB.Wiz.

But just curious. As a PCB.Wiz. Can you trace all the circuits above?

Also you asked why the EEG pc app limit it to 256 samples per second.





This is because the brain waves are only up to 50 Hz. So the 256 Hz sampling is almost 5 times (satisfying the Nyquist sampling theorem).

Also brain waves are in microvolts.. While the high frequency EMG I need is in millivolts. So the EEG equipment has more amplification than I need.

I think your suggested ADS1292ECG-FE will be great. https://www.ti.com/tool/ADS1292ECG-FE  I downloaded their free app.  It seems to do 2000 Hz?? I did even try inputiing up to 100,000 Hz.  Or did I misread something?

[ataradov]

The low pass filter cut off frequency has nothing to do with sampling rate.

You will have to take this one step at a time. First of all get any MCU board and blink and LED and send some characters over UART. Get familiar with peripherals. Then connect the ADC board and start sending real samples.

There is too much going on to get is all working at once from scratch.

[planc]

I have written so much programs with the 68HC11 hence familiar with MCU.

What is meant by low pass filter then?  Sampling of wave is taking samples of it. I want it to sample at 3000 Hz so I could get signal at 1000 Hz.

[ataradov]

There is no need to attach pictures explaining trivial things.

That tool does further filtering of the samples. As you can see it has adjustable low-pass and high-pass filter plus a notch filter to reject mains (50/60 Hz). None of this tells anything about the sample rate. Generally to implement 2 kHz low pass filter, the sampling frequency must be at least 4 kHz.  But as you said, this field lets you enter any value, so who knows what the actual sample rate is. But I bet it is documented somewhere in their documentation or can be configured in settings.

You say that you have experience with 68HC11, but this is a 40 year old device. And this whole task is trivial for anyone familiar with MCU programming.

[planc]

https://www.ti.com/lit/ug/slau384a/slau384a.pdf?ts=1673847243143&ref_url=https%253A%252F%252Fwww.ti.com%252Ftool%252FADS1192ECG-FE%253FHQS%253Ddis-dk-null-digikeymode-dsf-pf-null-wwe%2526DCM%253Dyes%2526distId%253D10



You are right. The app PC actual sample rate seems to be only 500 samples per second. I need at least 3000 samples per second.

So PCB.Wiz, the eval board you suggested won't meet the requirements.

Hmm.. Ok. Where can I find service (where I'll pay) to design and build the circuit and make the firmware and PC app for 10mV 3000 samples per second signal to be outputted to PC and display filtered waveforms at 600 Hz, 700 Hz, 800 Hz and every hundred up to 1500 Hz.

What company can do that? or anyone of you can do that (for a fee)?

[ataradov]

The idea of a kit is that you get the proven hardware right away, you can then use it to make your own software that does whatever you want.

I'm not aware of any contract services like this, but they all would require a very detailed specification. Something we have not seen here yet. So, your starting point would be come up with this and it might help us understand what you want exactly.

In low noise systems every part matters, it is not just a matter of the ADC board design. It is not an insignificant project, it would be very expensive.

[Siwastaja]

Instead of a consult firm who is tasked to develop "just the code" or "just electronics + code" based on a specification, it might be a better idea to find someone who takes a more holistic approach to the problem, i.e. you will spend several hours just discussing about the problem, brain waves and whatnot in general. As a result, they will gain understanding what you are really trying to achieve and can suggest ways to simplify the problem - and then design electronics / write firmware which does what you need. Without this wide and deep discussion into the physics side you and any subcontractors are possibly going to solve the "wrong problems". Similarly, resources might be better spent if the analog/firmware/algorithm designer learns about this EMG thing directly from primary sources, instead of / in addition to yourself acting as a middle man trying to come up with analog specifications.

This is how I usually work, for example, but of course finding someone close to you who also has time for this problem could prove quite difficult.

[PCB.Wiz]

...The app PC actual sample rate seems to be only 500 samples per second. I need at least 3000 samples per second.

That's likely merely a default, so I would look for the source code for the TI Eval board.

Google finds this "Kit source code / ADS1292ECG-FE Labview PC software source code."
and TI link says this

SUPPORT SOFTWARE ADS1X92XEVM-SW — Software for ADS1x92xECG-FE = Requires export approval (1 minute)

... Can you trace all the circuits above?

It just takes time. 
Make yourself a low voltage milliohm sensor (low voltage current source and a opamp and LED ) - that gives quick trace checks with no damage risk.

Next, print out the datasheets and then use your eyes and meter to find the PCB connections, starting with the most likely pins. Slowly, you create a schematic.
 

.... firmware and PC app for 10mV 3000 samples per second signal to be outputted to PC and display filtered waveforms at 600 Hz, 700 Hz, 800 Hz and every hundred up to 1500 Hz.

That's a strange wording.
Do you want a FFT histogram with 100Hz spacing, or do you want what you have written, which is 9 time domain traces, each with a slightly different filter  (100Hz filter increments)
Notice the TI demo is a time domain display.

[eutectique]

You've omitted the most interesting part of page 318:

SUMMARY OF AURIC COLOR FREQUENCIES
Color Approximate Central Frequency
Low blue (muscle) 2 or 200 Hz
Green 3 or 300 Hz
Yellow 4 or 400 Hz
Red 5 or 500 Hz
Orange 6 or 600 Hz
High blue 7 or 700 Hz
Violet 8 or 800 Hz
Cream 10 or 1,000 Hz
White 11 or 1,100 Hz up

I scrolled through all 368 pages and didn't see a single formula. In the book titled "Science of the Human Vibrations
of Consciousness".

Why do I have a feeling that someone is training chatgpt4?

[ataradov]

Yeah, I suspected that some quackery is going on.

[eutectique]

Sorry, could not resist. Page 315.

The small equipment shown beside the person is a miniaturized, modular sensing instrument, which is like a battery-powered telemetry EMG. Actually, it is a small FM broadcasting system with an amplifier and transmitter for each channel. Note the 12 modular plug-in, white boxes comprise six IRIG channels — with an amplifier and transmitter for each channel. The IRIG channels that we used were numbered 11, 12, 13, and 14 to indicate the recording capacity of each channel. The highest channel that we used was IRIG 14, with a sensing capacity up to 20 thousand Hz.

We had discovered that the regular, hard-wire instruments (operated by electricity) used for EMG, ECG and EEG did not pick up the entire millivoltage signal of the energy field. Hard-wire equipment we designed later allowed us to capture up to 200 thousand cycles per second from the energy field. When the two instruments, telemetry and hard-wire simultaneously recorded the energy field, the telemetry information was a perfect harmonic of the hard-wire instrument, yet of lower frequencies.

In Exhibit 1, the bipolar (2) surface sensors of silver, silver chloride are affixed to the solar plexus or "emotional body" areas. The ground, or neutral reference electrode, is on the wrist.Continuous energy field signals were beamed via the FM transmitter to receiving instrument panels. These data were observed from oscilloscopes, and a print-out, and by further amplifying the sound. At the same time, the energy field data were recorded on a magnetic tape.

[PCB.Wiz]

The TI export approval is just to make sure you won't use any subroutines in the software and integrate it in attack drones in some country in Europe. So just click Accept and you will be able to download it fine.

So you actually did that, and got the software and circuits ?

With the EEG now I have with separate 24 bit ADC external to the MCU, it's difficult. Anyway.
Do you know a working MCU (any brand) with external 24bit ADC that has already existing schematic and even firmware (like from a library)?
I'd like to see how the MCU and 24 ADC are connected.

Did you drill into the schematics of any of the adc eval boards ?
Connection of a SPI ADC is relatively simple - you find the MCUs SPI capable pins, and connect them to the ADC SPI pins. That's a few minutes work.
Then, you need code that activates the SPI peripheral on the MCU to meet the timing requirements of the ADC, testing as you go.
That's what EVAL boards help you with, as they have already working examples.

About what I want to see. It's like the following where one can see the intensity at every 100 Hertz from 100 to 1500.

That's labeled as a frequency-time plot, and I have seen those called a waterfall plot.  (see post below)
These days the axes are usually swapped, but there is existing work you can tap into for waterfall plots.

Examples are here
https://en.wikipedia.org/wiki/Waterfall_plot

What is the program in the second screenshot above? It's made in the 1980s and never duplicated again.

Possible Microsoft's QuickBASIC (or similar)   "Microsoft released the first version of QuickBASIC on August 18, 1985 on a single 5.25-inch 360 KB floppy disk."
It is doing time domain voltage-time plots, and inset boxes called cross plots are XY graphs of two of the channels. Most oscilloscopes can do that plot.

Do you know how to program PC? Help make one for me because I don't do pc programming.

Then you are out of your depth, in even specifying what you need someone else to do.

Maybe you should shift focus to Oscilloscopes, which can capture and record waveforms, and better ones have processing software for waterfall plots.  (google hints Rigol scopes can do waterfall)
If you ultimately plot pixels on a screen, you do not need 24b ADC, just a higher end scope is fine.

eg try this
https://www.zeitnitz.eu/scope_en

and this
DHO1000 HIGH RESOLUTION DIGITAL OSCILLOSCOPES   https://www.rigolna.com › products › dho1000
Set your frame number to capture and visualize using a density, waterfall, perspective, or mosaic display. Utilize UltraAcquire to capture small, ...

[PCB.Wiz]

About what I want to see. It's like the following where one can see the intensity at every 100 Hertz from 100 to 1500.

That's labeled as a frequency-time plot, but it's a little different when I look closely. There are too many related lines, which suggests multiple capture channels rather than a single spectrum spread.

Is there more info on the actual experiment setup ?

What they may have done for this, is construct ~ 10 sharp audio band filters, and then plotted the envelope amplitude output from each of those filters.

However, the plot also varies in density :
FWIR 'way back' there were spark/arc plotters way back, that left a mark and more time in that area, meant a thinker / darker line, or the higher the drive current, the larger the spots.

If we presume they used that tech, and the line thickness is then amplitude, to get some small frequency deviation info, (within each audio band) you could place a monostable after an audio bandpass (that's how tachometers work)

It's rough, but useful, and the plots will be multi 'line' XYZ* in nature, and may cross-over into each others territory, depending on the band-pass and mechanical plotter offsets - which is what your image does. 
*Z is the line thickness.

Guys. I have some questions. She claimed to do it in the 1970s. At that time, what kind op-amps or device or technology can even measure frequency of 20,000 Hz? That would need sampling to be at least 40,000 Hz to satisfy Nyquist theorem. At that time, what is the best sampling possible?  (She claimed a NASA engineer helped her build telemetry to measure that). Can you list the years and what maximum samples rate can be achieved?  She wrote:

You are over thinking this, and in 2023 terms.

Certainly 20,000Hz is just audio band, and Audio amplifiers have been around as long as loudspeakers, which is over 100 years !
LCR band filters are certainly easy 1970's technology, and impulse tachometers date from at least the 60's, and ferrite cores date from the 30's 

Doing this in the 70's is not going to fit in your pocket, but it's bench-sized and repetitive (~10 channels) audio level tech, so certainly within any decent lab ability.

[gf]

Audio CD was invented in the 1970's too.
So yes, it was indeed possible.

Btw, since it is just audio frequency, what about using a PC with a good soundcard to capture the signal? Possibly with an additional pre-amp. 1...10mV is basically microphone level.

[planc]

The TI export approval is just to make sure you won't use any subroutines in the software and integrate it in attack drones in some country in Europe. So just click Accept and you will be able to download it fine.

So you actually did that, and got the software and circuits ?

If you will read the following Texas Instruments export approval before downloading the software. It doesn't prohibit you from downloading it (you are not in any of the sanctioned country).  In the second screenshot is the BIS description of control product for export. I spent 30 mins researching about it as I nearly ordered the hardware. And it is not prohibited to get it out too. But the reason I didn't order the product was because of Ataradov warning in time that the actual sample rate may not be the one listed in the low pass window (which you can even type as 100,000 Hz).
Please try to download the software now and see if you could adjust the 500 samples per second seeming limit to 3000 samples per second. For now. I only need one channel, not 6 or 12. One channel is enough just to probe the 3000 Hz sector space at 1 to 10mV just to see if there is something there. If not, then ok. I'd be as happy and forget about this all.

https://www.ti.com/tool/ADS1292ECG-FE






MCU products are controlled under 3E002.c if it matches the description:

"c. Designed to perform more than eight 16-bit 'fixed-point' multiply-accumulate results per cycle (e.g. digital manipulation of analog information that has been previously converted into digital form, also known as digital "signal processing".

The hardware doesn't match that description, soyou can adjust the software for 3000 Hz, then anyone outside can order it too without the United States arresting you or putting you in Federal Prison or initiate Extradition proceeding for acquiring an MCU.

[planc]

With the EEG now I have with separate 24 bit ADC external to the MCU, it's difficult. Anyway.
Do you know a working MCU (any brand) with external 24bit ADC that has already existing schematic and even firmware (like from a library)?
I'd like to see how the MCU and 24 ADC are connected.

Did you drill into the schematics of any of the adc eval boards ?
Connection of a SPI ADC is relatively simple - you find the MCUs SPI capable pins, and connect them to the ADC SPI pins. That's a few minutes work.
Then, you need code that activates the SPI peripheral on the MCU to meet the timing requirements of the ADC, testing as you go.
That's what EVAL boards help you with, as they have already working examples.

This is the last time I'll post images of these. You won't find these anywhere because the ordinary public can't buy them but only those with certain license.



You said the SPI ADC connection to the MCU is very simple. You only connect them directly. In the above ADS1254E ADC (at right side) and the MCU at left. There are many ICs detours in between them. INA 122U, HC4052M, HCTO4, MV358I. Which of them do you think is usually added?  (note the white thing are just H11L1 4500 volts opto-isolator).

The following is zoom of the MSP430, and the pinout designations.  Notice that the MSP430 pin P3.6/uca1TXD/UCA1SIMO (top 2nd pin from right) is connected to pin5 (A3 input) to the hex inverter HCTO4. Have you seen the ADS1254E need to pass through the HCTO4 before getting to the MSP430? Why do you have to hex invert the external ADC to the MCU?









Lastly. SPI connections are like this:



Where are the SPI corresponding pins in the ADS1254E?



I have googled for hours yesteday but couldn't find the answer. Many thanks for your help PCB.Wiz!

[ataradov]

ADS1254E is not really an SPI device. You can just use SPI to get the data, but you need to do extra work as well.

it uses SCLK for the clock and DOUT as the data. But DOUT is also used as a ready signal.  So, you need to connect SCLK to the SPI clock and DOUT to the MISO. But then you need to keep MISO pin configured as a GPIO and read its value. When the value goes low, it indicates that the data is ready. You then need to switch to SPI mode and clock in 3 bytes. Then switch back to GPIO and repeat.

[errorprone]

HCT04 could be used to level shift between a 3V MCU and 5V ADC since it has TTL inputs instead of CMOS ones.

[PCB.Wiz]

You said the SPI ADC connection to the MCU is very simple. You only connect them directly. In the above ADS1254E ADC (at right side) and the MCU at left. There are many ICs detours in between them. INA 122U, HC4052M, HCTO4, MV358I. Which of them do you think is usually added?  (note the white thing are just H11L1 4500 volts opto-isolator).

The following is zoom of the MSP430, and the pinout designations.  Notice that the MSP430 pin P3.6/uca1TXD/UCA1SIMO (top 2nd pin from right) is connected to pin5 (A3 input) to the hex inverter HCTO4. Have you seen the ADS1254E need to pass through the HCTO4 before getting to the MSP430? Why do you have to hex invert the external ADC to the MCU?

The H11L1 are schmitt optocouplers, so they look to have isolation between the analog and digital portions here.
The designer may have put some buffers between the opto's and MCU, use your low voltage continuity checker to trace the CLK and DATA signals.