Is this a major advance in PI sensitivity ??

[moodz]

I had been thinking about a comment by Carl-NC I read in another thread about 12 bit resolution VS 8 bit, combined with the subject of dynamic range mentioned in this quote:

"In theory, given the same voltage range for single-ended and fully-differential inputs, the fully-differential inputs will have double the dynamic range" from this link:

http://www.maxim-ic.com/app-notes/index.mvp/id/1108

Which lead me to the LME49724 easily located on eBay and low in cost. Problem is then I end up with an ADC such as the AD7450 and again am in a situation where so far only evaluation boards seem to be readily available as well as cost being an issue from a mass production perspective.

http://www.analog.com/en/analog-to-d...s/product.html

http://www.analog.com/en/evaluation/EVAL-AD7450/eb.html

Therefore I end up stuck in an endless loop of one difficulty or another until I can think of a better part for the ADC. Unless of course I find a cheap source for the AD7450.

The usual design process that I find is most efficient is to build it first an then make it better. You don't choose a part that offers something like better dynamic range unless you "need" it. There is always a better part available than the one you have now ... You will never build anything if you are always looking at the grass on the other side of the fence.
PS a 24 bit converter with differential inputs is like $5. More dynamic range than you will need.

[Midas]

The usual design process that I find is most efficient is to build it first an then make it better. You don't choose a part that offers something like better dynamic range unless you "need" it. There is always a better part available than the one you have now ... You will never build anything if you are always looking at the grass on the other side of the fence.
PS a 24 bit converter with differential inputs is like $5. More dynamic range than you will need.

Depends on what your designing I think Moodz. If its truly ground breaking and experimental and the concept can be proved with cheap components then sure use whatever. If your designing something based on well proven concepts then its inefficient to choose components that you know are ultimately going to limit performance, sending you back to the drawing board. Its like sitting down to build a better motorbike and starting by making a bicycle. You get to the stage where its time to put the motor in and have to scrap the frame and the wheels you've already made because they aren't strong enough. But you should have already known that because you've seen other motorbikes..

BTW, What's this $5, 24bit convertor of which you speak? I kind of suspect it's not going to be anywhere near fast enough for direct sampling.

[moodz]

Depends on what your designing I think Moodz. If its truly ground breaking and experimental and the concept can be proved with cheap components then sure use whatever. If your designing something based on well proven concepts then its inefficient to choose components that you know are ultimately going to limit performance, sending you back to the drawing board. Its like sitting down to build a better motorbike and starting by making a bicycle. You get to the stage where its time to put the motor in and have to scrap the frame and the wheels you've already made because they aren't strong enough. But you should have already known that because you've seen other motorbikes..

BTW, What's this $5, 24bit convertor of which you speak? I kind of suspect it's not going to be anywhere near fast enough for direct sampling.

Well that is another point of view ... however there is not a single design posted on this forum that actually uses an ADC ( commercial schematic or user submitted ) ... happy to be proved wrong here.
The design process should begin with the GOAL ( or outcome ) ... which I would imagine is to design a detector with performance at least in the same ballpark as commercial detectors.
Secondly the main functional subsystems that make up the detector system should be identified, how they interact and specified.
Thirdly each of these subsystems should be designed, prototyped and tested.
Fourthly the whole system should be integrated and the proof of concept be verified.
Fifth at this point the required performance gap should be identified AND IT IS AT THIS POINT THAT ANY EXOTIC COMPONENTS ARE UTILISED as each of the subsystems is brought up to the required performance level. If you can avoid it you wont use exotic components as they often suffer from supply problems during the product life cycle.

Finally thats why these forums ... ( although very interesting so dont get me wrong here ) ... never result in a finalised design. Everyone is either simulating or looking up chip catalogues for better and better chips when in reality I think there would be much more satisfaction if all the subsystems that actually make up a detector were actually posted up.

So am I being hypocritical here .... I think not.

I have specified and actually made ( as against talking about it ) a PI controller chip.
I have specified and actually made a frontend ( actually several ) but the two main ones are the differential and the more recent mono (with diff amp to reduce EMI).
I am working on the sampling and amplifier section.
I have specified and actually made a fully FPGA based DSP system that is a full direct sampling system but subject to current patent applications.

All of these systems followed the methodology I outlined above and did not use especially exotic components.

As for the ADC chip ... the detector system model is a wide bandwidth / high frequency system that is modulated by a very low frequency signal ( the target / ground ). The real challenge is to extract the signal not the carrier by demodulating / filtering / amplifying / sampling IN THAT ORDER. Thus the low frequency signal being sampled is more than adequately achieved with a commodity type ADC used in sound cards etc ( in fact most of these chips have 2 x differential inputs at 24 bits and 192 Khz sampling rate. )

The WM8788 is available from Farnell for A$ 2.58 in single quantities... you can get two for $5 !! ( cheaper now than when I last bought them .. wow ).

And just for reference the first motorbike was a bicycle with a motor bolted to it

[Aziz]

Indeed,

my first "motorbike" was made with a sound card. Brrrouuumm, brouuuuum, broooouuuuum, well, it's running a hell. (A two cylinder "motorbike", which means a two-channel ADC).


Moodz is correct. Use the most used chips from the commodity (cheap and available), which can't be bought off from the ******** chip trolls (buying all the available stocks to sell them at a much higher price). They do the job as well.

And sometimes it's better to know how to make a discrete solution.

Aziz

[Midas]

Well that is another point of view ... however there is not a single design posted on this forum that actually uses an ADC ( commercial schematic or user submitted ) ... happy to be proved wrong here.
The design process should begin with the GOAL ( or outcome ) ... which I would imagine is to design a detector with performance at least in the same ballpark as commercial detectors.
Secondly the main functional subsystems that make up the detector system should be identified, how they interact and specified.
Thirdly each of these subsystems should be designed, prototyped and tested.
Fourthly the whole system should be integrated and the proof of concept be verified.
Fifth at this point the required performance gap should be identified AND IT IS AT THIS POINT THAT ANY EXOTIC COMPONENTS ARE UTILISED as each of the subsystems is brought up to the required performance level. If you can avoid it you wont use exotic components as they often suffer from supply problems during the product life cycle.

Finally thats why these forums ... ( although very interesting so dont get me wrong here ) ... never result in a finalised design. Everyone is either simulating or looking up chip catalogues for better and better chips when in reality I think there would be much more satisfaction if all the subsystems that actually make up a detector were actually posted up.

So am I being hypocritical here .... I think not.

I have specified and actually made ( as against talking about it ) a PI controller chip.
I have specified and actually made a frontend ( actually several ) but the two main ones are the differential and the more recent mono (with diff amp to reduce EMI).
I am working on the sampling and amplifier section.
I have specified and actually made a fully FPGA based DSP system that is a full direct sampling system but subject to current patent applications.

All of these systems followed the methodology I outlined above and did not use especially exotic components.

As for the ADC chip ... the detector system model is a wide bandwidth / high frequency system that is modulated by a very low frequency signal ( the target / ground ). The real challenge is to extract the signal not the carrier by demodulating / filtering / amplifying / sampling IN THAT ORDER. Thus the low frequency signal being sampled is more than adequately achieved with a commodity type ADC used in sound cards etc ( in fact most of these chips have 2 x differential inputs at 24 bits and 192 Khz sampling rate. )

The WM8788 is available from Farnell for A$ 2.58 in single quantities... you can get two for $5 !! ( cheaper now than when I last bought them .. wow ).

And just for reference the first motorbike was a bicycle with a motor bolted to it

I mostly agree with your design philosophy, but I'm going to insert a new third step:
Research whether the subsystem you need has already been developed and is available for use, perhaps in another field. Make an honest assessment of whether you think you capable of improving it, if not use it as is. In other words, don't waste time reinventing the wheel.

Other than that my only point was that your 5th step is going to frequently send you all the way back to your 3rd step. Time is by far the biggest investment in this sort of development. So if you can avoid some of these cycles by starting with suitable quality components then even if you do spend a little more than strictly necessary on the bits, your still be well ahead. If you achieve the performance your after and start thinking about production then you can start thinking about economizing on components.

THe WM8788 is indeed a surprisingly cheap performer. I'm a little suspicious of how the datasheet is written though. Nothing is specified at its full data rate or pre-low-pass-filter.

Oh and I said a better motorbike...a moped is never going to be better than an R1.

[moodz]

I mostly agree with your design philosophy, but I'm going to insert a new third step:
Research whether the subsystem you need has already been developed and is available for use, perhaps in another field. Make an honest assessment of whether you think you capable of improving it, if not use it as is. In other words, don't waste time reinventing the wheel.

Other than that my only point was that your 5th step is going to frequently send you all the way back to your 3rd step. Time is by far the biggest investment in this sort of development. So if you can avoid some of these cycles by starting with suitable quality components then even if you do spend a little more than strictly necessary on the bits, your still be well ahead. If you achieve the performance your after and start thinking about production then you can start thinking about economizing on components.

THe WM8788 is indeed a surprisingly cheap performer. I'm a little suspicious of how the datasheet is written though. Nothing is specified at its full data rate or pre-low-pass-filter.

Oh and I said a better motorbike...a moped is never going to be better than an R1.

Touch`e I guess I would lump research and not reinventing the wheel under the design step but I guess it is good to be explicit about it.
Recently I came across a power switching device that switches 8 amps, breaks down at 1700 volts and only has 20 pF output capacitance. So I bought 5 of them ( $20 a pop ) however there appears to be only one manufacturer and the application field in high efficiency solar regulators is changing so fast that I can just see these devices not being available in a couple of years.
Will I play with it ... yes .... Would I recommend it for a new design .... no ... for the reasons above.

On motorcycles ... We would expect an R1 to perform better than a moped. The real engineering challenge is to surprise everyone by designing a moped that out performs an R1 not the other way round.

[moodz]

...and the WINNING ADC chip is .....

MCP3903 ( or MCP3901 if you are a real tight ***** )

This chip was made for metal detecting ( or should have been )

6 separate ( not muxed ) 24 bit ADCs with a programmable gain amp in front of each channel. Also a handy dandy phase shifter built in ... ( hmmmm )
Low power and LOW price ( mine were A$ 3.90 each )
... made by Microchip so they bolt straight onto the dspic chips. ( eat my shorts AtMega )

see full spec pdf here ...http://ww1.microchip.com/downloads/e...Doc/25048B.pdf

My initial testing shows that the spec is very conservative and the noise performance is much better than the quoted figures.

The UNIPI chip code is under test now.

Features
• Six Synchronous Sampling 16/24-bit Resolution
Delta-Sigma A/D Converters with Proprietary
Multi-Bit Architecture
• 91 dB SINAD, -100 dBc Total Harmonic Distortion
(THD) (up to 35th harmonic), 102 dB Spurious-free
Dynamic Range (SFDR) for Each Channel
• Programmable Data Rate up to 64 ksps
• Ultra Low-Power Shutdown Mode with <2 μA
• -115 dB Crosstalk Between any Two Channels
• Low Drift Internal Voltage Reference: 5 ppm/°C
• Differential Voltage Reference Input Pins
• High Gain PGA on Each Channel (up to 32 V/V)
• Phase Delay Compensation Between Each Pair
of Channels with 1 μs Time Resolution
• High-Speed Addressable 10 MHz SPI Interface
with Mode 0,0 and 1,1 Compatibility
• Independent Analog and Digital Power Supplies
4.5V - 5.5V AVDD, 2.7V - 3.6V DVDD
• Available in Small 28-lead SSOP Package
• Extended Temperature Range: -40°C to +125°C
Applications
• Energy Metering and Power Measurement
• Portable Instrumentation
• Medical and Power Monitoring
Description
The MCP3903 is a six-channel Analog Front End (AFE)
containing three pairs made out of two synchronous
sampling Delta-Sigma Analog-to-Digital Converters
(ADC) with PGA, a phase delay compensation block,
internal voltage reference, and high-speed 10 MHz SPI
compatible serial interface. The converters contain a
proprietary dithering algorithm for reduced idle tones
and improved THD.
The internal register map contains 24-bit wide ADC
data words, a modulator output register as well as six
24-bit writable control registers to program gain,
over-sampling ratio, phase, resolution, dithering,
shut-down, reset and several communication features.
The communication is largely simplified with various
Continuous Read modes that can be accessed by the
Direct Memory Access (DMA) of an MCU and with
separate Data Ready pins that can directly be
connected to the Interrupt Request (IRQ) input of an
MCU. The MCP3903 is capable of interfacing to a large
variety of voltage and current sensors including shunts,
current transformers, Rogowski coils, and Hall-effect
sensors.

[6666]

Can be ordered through RS components Sydney.
6-8 days wait , free postage in Oz

MCP3903-E/SS
Qty Unit Price
1 $4.12
5 $3.76

[Altra]

Moodz, Nice find on the MCP3903. The pga, phase delay and 16bits make for an interesting mix. Ordered some today.

Thanks

[nick_f]

Driving the capacitors C6 +C7 with the TC4421, slows down the switch OFF speed considerably. This has a lot of influence on the Flyback and damping time.

Tinkerer

Hi Tinkerer,
I also didn't like the idea to use the same signal driving the coil mosfet and the converters for the power supply. It would be better to use two separate TC4421 or maybe to separate the outputs of the one used, one output for mosfet and one for DC-DC converters.
I am not sure if there will be any improvement in the real circuit, since I have not tried any of this.


PS - I should have read more, Moodz already explained the minimal effect of using the caps C6 and C7 on the same output of TC4421...

Regards,
Nicolae

[simonbaker]

Moodz, Nice find on the MCP3903. The pga, phase delay and 16bits make for an interesting mix. Ordered some today.

Thanks

I'll second that. Bookmarking. It should be added to our permanent "parts bucket" web page that doesn't exist.

However, I'd also like to find a faster ADC, with at least 160 Ksps and 24 bit.

Cheers,

-SB

[moodz]

I'll second that. Bookmarking. It should be added to our permanent "parts bucket" web page that doesn't exist.

However, I'd also like to find a faster ADC, with at least 160 Ksps and 24 bit.

Cheers,

-SB

ADS1675 :-)

[simonbaker]

ADS1675 :-)

Thanks! Is this the best bang for the buck, or is there something cheaper that works in the 160 ksps range? Also important... not surface mount!

Regards,

-SB

[6666]

ADS1675

this has recently dropped to $53.00 here in oz

[moodz]

ADS1675

this has recently dropped to $53.00 here in oz

Yeh ... kinda expensive ...especially if you make a soldering mistake

However it is good bang per buck on a samples per dollar basis for a 24 bit converter. 40000 / dollar which is comparable to your lower sample rate converters.

moodz.