# When GND Isn’t GND, Single-Ended Circuits Become Differential

The system ground return, or GND, symbol is often taken for granted when drawing schematics. GND symbols are placed all over the schematic with the assumption that the different GNDs will all be at the same electrical potential on the printed circuit board (PCB). In reality, current flow through the GND impedance can create voltage differences between the GND connections on the PCB. Single-ended dc circuits are particularly sensitive to these GND voltage differences because the expected single-ended circuit is converted to a differential circuit causing output errors.

Let’s use the standard non-inverting amplifier circuit shown below as an example. When the GND potentials of the input source, VIN, and the input resistor, RI , are equal, the familiar circuit gain of 1+RF /RI applies. Therefore, the 100mV input signal is multiplied by a gain of 10 V/V and the output is 1V.

In the circuit shown in the figure below, a voltage source, VGND2 , has been inserted between the input source GND and the RI GND connection. The result is a modified transfer function with the addition of the VGND2 voltage multiplied by the inverting circuit gain of – RF /RI . A 10mV difference in GND potentials reduces the desired 1V output by 90mV to 0.91V. This corresponds to a 9% relative error compared to the desired 1V output.

In the circuit shown below the circuit transfer function is further affected when the output voltage is referenced to a third GND potential, VGND3. The VGND3 voltage will directly subtract from the previous output transfer function. Therefore, a 20mV VGND3 voltage further reduces the output voltage to 890mV, corresponding to an 11% error compared to the desired 1V output.

The issues shown in the two examples above can be reduced by using proper PCB layout techniques to keep GND potentials of the circuit input source, input resistor, and output voltage at the same potential. The best solution is to place critical GND connections physically close to each other using a common “star” GND method. This will reduce the PCB impedance developed between the GND connections and will therefore reduce any voltage potential differences between them. In the example circuit schematic and layout shown below, the distances between the GND connection of the input source, output voltage, and input resistor are all located close together on the top layer of the PCB. This should prevent the single-ended circuit from becoming a differential one!

In summary, next time you’re having issues with dc circuit performance verify that the voltage potentials of all important GND connections are equal.

## 81 comments on “When GND Isn’t GND, Single-Ended Circuits Become Differential”

1. Davidled
July 5, 2014

Design and quality engineer should understand the cause of signal distortion whether it comes from either misinterpretation of circuit design or ground.  Ground might be related to IC die technology. The integration of digital and analog circuit might produce more noise or ground issue.  Board designer and digital and analog engineer might work together to have the board provide the precise signal in any environment.

2. goafrit2
July 6, 2014

>> The integration of digital and analog circuit might produce more noise or ground issue

My rule #1 is have different GNDs for both digital and analog grounds to avoid cross-talk and noise. But generally, we need to design systems to be hardened from some of these. A differential strategy is the best.

3. Victor Lorenzo
July 6, 2014

@goafrit2, I agree with you on separating analog and digital grounds, it is an effective noise reducing PCB design technique.

Some other useful hints are given by TI applications engineers here: http://www.ti.com/lit/ml/sloa089/sloa089.pdf.

4. etnapowers
July 7, 2014

In case of a multi-site testing, an Automatic Testing Equipment having a dedicated ground for each site is desirable, provided that the layout of each site has to be realized by separating the analog and the power ground.

5. etnapowers
July 7, 2014

That's a good rule, I think, I add that the layouts of the PCB has to be symmetric and the ground planes have to be opportunely insulated one from each other.

6. Netcrawl
July 8, 2014

@Victor I agree with about separating analog and digitall grounds, it still one of the most effective methods of noise suppression, separate grounds does not mean that the grounds are electrically separated in a certain system, they need only to be common at some point, perhaps a single low-impedance point.

7. Netcrawl
July 8, 2014

@etnapowers, in creating a good PCB layout, we need to evaluate first potential problems  in the early design stages. We need to identify the quite and noisy portions of the circuits, when it is known, we can now implement noise reduction techniques more effectively

8. etnapowers
July 8, 2014

@Netcrawl, agreed, that's a good strategy. A good planning of the layout of the PCB has to take into consideration the simulation data and the test results of the most critical points of the board, by considering the noise issue.

9. Davidled
July 8, 2014

What if automatic testing equipment (ATE) is connected to the testing module?  In that case, ground of each will be connected. There will be possible ground potential between ATE and testing module. If ATE has capability to minimize the ground voltage, it provides more precise evaluation result.

10. etnapowers
July 9, 2014

@DaeJ, in this case there are two important factors that have to be guaranteed:

1. The ground voltage, that has to be accurately controlled by a feedback signal of a control system inserted in the ATE hardware.
2. The connections between the ground planes of the test board and the pin grounds of the ATE , that should offer low resistance to avoid the presence of a voltage drop across the path.
11. Davidled
July 9, 2014

For #1, I think that this comment mentioned a kind of PID controller. PID controller could be used for system control consisting of input and output. I am wondering what type feedback signal could control the ground voltage. I guess that the desired voltage coudl be 0 [V].

12. etnapowers
July 9, 2014

The feedback signal is the output of an error amplifier that has a zero reference voltage.

This type of control is very sensitive to the precision of the reference voltage.

13. Davidled
July 9, 2014

In the most application, PID controller will be used for any parameter of input in the system, but, it is hard to say that ground voltage could be one of input signal due to its characteristics. As alternative way, it might be suggested that engineer designs the circuit better for high power distribution with AC.

14. goafrit2
July 9, 2014

>> @goafrit2, I agree with you on separating analog and digital grounds, it is an effective noise reducing PCB design technique.

Besides PCB development, it works well on mixed signal (integrated circuit) design.

15. goafrit2
July 9, 2014

>> That's a good rule, I think, I add that the layouts of the PCB has to be symmetric and the ground planes have to be opportunely insulated one from each other.

And always make sure that your main signal lines (the critical signals) are well protected (guarded). That ensures they are not perturbed by noise.

16. goafrit2
July 9, 2014

>> separate grounds does not mean that the grounds are electrically separated in a certain system, they need only to be common at some point, perhaps a single low-impedance point.

You can pass that signal as a unity follower by wasting one amp so that by the time you are using it the other side, you have a fairly stable clean signals. In biomed design where signals are very small that noise budget becomes an issue, every trick to get noise on ground planes low is used.

17. goafrit2
July 9, 2014

>> A good planning of the layout of the PCB has to take into consideration the simulation data and the test

That is why those auto-routing wizards they package in PCB CAD is a marketing buzz and largely useful in practice.

18. goafrit2
July 9, 2014

>>  that should offer low resistance to avoid the presence of a voltage drop across the path.

When in doubt, use a buffer to get the low impedance for optimal result. Test equipment is part of the system and its noise contribution must be considerred. A stable power source generated and passed through a stabilizer before it makes its way into the DUT will help.

19. fasmicro
July 9, 2014

Some other useful hints are given by TI applications engineers here:http://www.ti.com/lit/ml/sloa089/sloa089.pdf.

That is my best way of learning circuits and systems. If you read Application Notes from some of these great firms, you will get a better practical insight than textbooks.

20. fasmicro
July 9, 2014

In the most application, PID controller will be used for any parameter of input in the system, but, it is hard to say that ground voltage could be one of input signal due to its characteristics.

If you make a rail-to-rail opamps, there is a chance you can accomplish anything you want in the system. It effectively opens the window to use signal in any way because there will not be need to worry if there will be attenuation because of dynamic range.

21. etnapowers
July 10, 2014

“it might be suggested that engineer designs the circuit better for high power distribution with AC.”

@DaeJ: could you specify what do you mean by high power distribution with AC?

22. etnapowers
July 10, 2014

@goafrit2: I agree with you and I add that the signal paths have to be separated in the PCB layout , to avoid cross-talking.

23. etnapowers
July 10, 2014

A simulator of the effect of PCB load for the circuit is highly recommended to optimize the effectiveness of the design process.

24. etnapowers
July 10, 2014

@fasmicro: Really nice link thank you for this. The tutorial of Texas Instruments are a great resource for an engineer who has to design and debug a circuit board, there are good suggestions to improve the effectiveness of the testing process and the reliability of the overall system.

25. Davidled
July 10, 2014

Let us say that board might have high voltage transformer which converts AC to DC and there is a few signal conditional circuit less than 5[v]. In this case, incorrect design causes the noise as well as a highly potential ground voltage. Plus, if this kind of board is integrated into high speed communication module, it becomes more complex board and capturing the precise signal would be a challenge, unless engineer follows the correct process.

26. vasanjk
July 12, 2014

PrecisionHub

I think ground filling is done by designers without much thought. One of my friends argues that by filling, the ground differentials could be done away with. Can this be applicable to all situations?

27. Davidled
July 13, 2014

What kind of filtering would be applicable for ground issue? There are three filtering methods: high pass; low pass; and band pass. Plus, I think that ground signal would be a random signal.

28. Victor Lorenzo
July 15, 2014

@DaeJ, FYI, band-reject and all-pass filters are two more filter types which are used for removing a unwanted frequency band and for compensating for phase shifts, respectively.

29. Victor Lorenzo
July 15, 2014

@DaeJ, ground noise can come from many sources and several of them may overlap. Fortunately it is not always completely random and gives us a chance to deal with it.

In many cases, high frequency and high current signals return paths contribute as ground noise which interferes to low level signal paths.

30. Heinzelmann
July 21, 2014

I think the board layout could be improved.

Especially the currend of C1 must circumvent the input terminal to a large amount.

The ground pads of C1 and C2 should be as close as possible to the gnd  of the output terminal.

31. fasmicro
August 3, 2014

>> A simulator of the effect of PCB load for the circuit is highly recommended to optimize the effectiveness of the design process.

How do you do that? Any CAD for that?I design with Altium Designer (old Protel)

32. fasmicro
August 3, 2014

>> In many cases, high frequency and high current signals return paths contribute as ground noise which interferes to low level signal paths.

In my designs, I try to identify my critical paths and also try to protect them with guard rings or special ways to ensure they are not perturbed. When you have HF in a circuit, you have a real problem if you allow those emissions to affect the circuit

33. fasmicro
August 3, 2014

>> What kind of filtering would be applicable for ground issue? There are three filtering methods: high pass; low pass; and band pass. Plus,

That is not a good strategy. It is very expensive and cost your current. A filter needs opamp and that is money. The best is following best design rules and best practices and avoild throwing transistors to the problems.

34. fasmicro
August 3, 2014

>> I think ground filling is done by designers without much thought.

What is really “ground filling”. Any indept explanation on how this is done will be appreciated. Is it having many grounds as many places as possible?

35. fasmicro
August 3, 2014

>> @DaeJ: could you specify what do you mean by high power distribution with AC?

The comment did not make a lot of sense. But I guess he was trying to say “resillence to AC surge”.  You need to protect the power system to avoid failure when the AC fluctuates or the power noisy.

36. goafrit2
August 3, 2014

I agree with you and I add that the signal paths have to be separated in the PCB layout , to avoid cross-talking.

Besides PCB, integrated circuit design follows the same path. Most times, what works in PCB also work in chip design. Differential strategy with some of the ideas in the article are tested and make sense.

37. Davidled
August 3, 2014

IC Chip is conceptually very similar path with PCB. I am wondering if datasheet of IC indicates noise immunization and cross-talking level in detail,when engineer select the chip for PCB board among many chip in the market. It could be possible that one chip might have zero-tolerance to cross-talking compared with other.

38. etnapowers
August 4, 2014

@fasmicro: you can do it by CAD , many times I have asked to the supplier a model of the PCB load effect for a particular package of an IC, whom I was working on to put it in production. Altium Designer is a good and cheap tool which works well.

39. etnapowers
August 4, 2014

The buffer and the stabilizer are good options, but remember that the test board will be utilized for the production of million of parts, and the added components have to be reliable on the long term.

40. etnapowers
August 4, 2014

“Plus, if this kind of board is integrated into high speed communication module, it becomes more complex board and capturing the precise signal would be a challenge, unless engineer follows the correct process.”

@DaeJ, that's absolutely correct, in that case the design process of the board must be accurate to avoid the presence of undesiderable errors in the data trasmission due to capacitive coupling at high frequencies.
41. fasmicro
August 5, 2014

I am wondering if datasheet of IC indicates noise immunization and cross-talking level in detail,when engineer select the chip for PCB board

I am not sure if the data is published specifically for PCB, but some companies do report their EMIs in their datasheets. It is left for you to see how that will work with PCB. But do not forget that PCBs have their own issues that must be managed as well.

42. fasmicro
August 5, 2014

Altium Designer is a good and cheap tool which works well.

I am very curious, where are you buying the cheap Altium Designer. I think it is the most expensive PCB CAD in the market now. There are freebies but Altium Designer is nowhere cheap. It is indeed expensive when compared to alternatives.

43. etnapowers
August 6, 2014

@fasmicro: There are many free versions of Altium Designer that work effectively and moreover are considered suitable for a good design at industrial level.

44. goafrit2
August 9, 2014

>>  I am wondering if datasheet of IC indicates noise immunization and cross-talking level in detail,when engineer select the chip for PCB board among many chip in the market

In the medical fields, datasheets for products used therein are required to provide that data

45. goafrit2
August 9, 2014

>> you can do it by CAD , many times I have asked to the supplier a model of the PCB load effect for a particular package of an IC, whom I was working on to put it in production

There are many models available in many major CADs that can be ported during designs. Only big companies can actually request for suppliers to provide their model. You need to have big relationships to be in positions to get that. The easiest is using Altium Designer and other popular CADs.

46. goafrit2
August 9, 2014

>> and the added components have to be reliable on the long term.

The weakest link is having a great board and then using poor quality components to populate them. Components matter and sourcing the best in the industry is important for any design.

47. goafrit2
August 9, 2014

>> There are many free versions of Altium Designer that work effectively and moreover are considered suitable for a good design at industrial level.

48. Davidled
August 9, 2014

I am wondering if there is any example for power supply or any audio amplifier and memory associated with noise immunization in the medical field. If any is given, that could be applied to other industry field. I image that any component related to the capture of body signal should indicate these parameters including the passive component in the monitoring circuit.

49. RedDerek
August 20, 2014

That is why those auto-routing wizards they package in PCB CAD is a marketing buzz and largely useful in practice

Autorouting is good to use once the key signals are laid out – power, ground, etc. Then let the autorouter go. But even then I end up retoucing the routes several times before I am happy with a layout.

50. etnapowers
September 1, 2014

I know that it normally depends on the quality of relationship between customer and supplier, but I think that this should be a common practise by the software houses that produce the CAD softwares.

51. etnapowers
September 1, 2014

I fully agree with you, this is a very common situation, that might be avoided by a proper selection process of the suppliers of the components that populate the board.

52. etnapowers
September 1, 2014

The medical products have to meet some stringent requirements because it is important to be sure that the functionality of the IC will not be affected by noise generated by sources of electromagnetic fields. This compliance should be extended to other fields, different from the medical one, that require the immunity to the noise and cross-talk to ensure a good overall reliability.

53. Victor Lorenzo
September 1, 2014

In medium-to-low complexity PCBs I've found autorouting useful for making BGA fanout and escape routing. Autorouting the rest of the tracks requires making a very large effort in rules definition and good results are not warrantied.

Controlled impedance semi auto-routing for differential pairs is really useful too.

54. fasmicro
September 1, 2014

>> Autorouting is good to use once the key signals are laid out – power, ground, etc.

It saves valuable time. I do use it a lot. Just focus on the most critical signals. Get them done. Then later, use auto-router to complete the design. Also, there are very small circuits that auto-routers work.

55. fasmicro
September 1, 2014

>>  know that it normally depends on the quality of relationship between customer and supplier, but I think that this should be a common practise by the software houses that produce the CAD softwares.

It is very rare for anyone to control what the CAD companies do. Most times, you buy these CADs and there is nothing you can do about them. I will argue the best thing is to spend money on more quality CADs over cheap or free ones. In the ASIC sector, Cadence gives you everything you need for serious work. You can get some CADs to do some works but when complexity sets in, you will have issues.

56. fasmicro
September 1, 2014

>> The medical products have to meet some stringent requirements because it is important to be sure that the functionality of the IC will not be affected by noise generated by sources of electromagnetic fields.

Besides the electromagnetic field interfference and other spurrious factors, the key is that in medical, human lives are at risk and that is begging trial lawyers to get to work when you make mistakes. It is different compared to when you have a \$30 camera which you can throw away.

57. goafrit2
September 6, 2014

It is never a good strategy to try to do what they do in the medical ASIC design in other areas. Anyone that does that may be out of business. Medical ASIC has different standards and requirements that make them very expensive. Deploying such in cheap consumer ASIC may not make sense.

58. etnapowers
September 8, 2014

“It is very rare for anyone to control what the CAD companies do”

@fasmicro: that's correct, the only way to have a control is to require a dedicated software by mean of a CAD engineer working on site at a company design department. this could be expensive but it can work.

59. etnapowers
September 8, 2014

@fasmicro: this is the reason for why the effectiveness and the reliability of an IC for  medical device has to be very good, because it has to work well for a long time in an hard environment.

60. fasmicro
September 9, 2014

>> that's correct, the only way to have a control is to require a dedicated software by mean of a CAD engineer working on site at a company design department.

Unfortunately, that will be very expensive. When we think we need resident engineers on site, it means the CAD companies do not make good products. Software should evolve to the level where we do not need people to be working on site.

61. fasmicro
September 9, 2014

>> this is the reason for why the effectiveness and the reliability of an IC for  medical device has to be very good, because it has to work well for a long time in an hard environment.

That is the business – it is not about the product, it is what it will be used for. Gyros for toys can go for \$0.60 while medical device requires gyros that cost excess of \$290. That is the difference and that drives the design strategy since one can cost life while the other is just a product that can tossed if it breaks down.

62. etnapowers
September 11, 2014

“When we think we need resident engineers on site, it means the CAD companies do not make good products”

@fasmicro: the resident engineer may be needed also to develop a special software for a challenging project. I agree with you, it could be very expensive many times.

63. goafrit2
October 1, 2014

>> I image that any component related to the capture of body signal should indicate these parameters including the passive component in the monitoring circuit.

Check some of the works done by Tony Delburich in Zurich. They have done many designs in this area.

64. goafrit2
October 1, 2014

>> that's correct, the only way to have a control is to require a dedicated software by mean of a CAD engineer working on site at a company design department

The big problem with that is that only the biggest account holders can support that cost model. I do not expect Cadence or Synopsis to send an engineer to be on site just to support their product. That will be very bad.

65. goafrit2
October 1, 2014

>> this is the reason for why the effectiveness and the reliability of an IC for  medical device has to be very good,

In medical, a failure could mean death and that is a huge liability issue. That is why when you order a component, most vendors will explain their datasheets that their parts are not for medical applications. Should you want medical applications components, contact for a different type of product.

66. goafrit2
October 1, 2014

>> the resident engineer may be needed also to develop a special software for a challenging project

Always remember: the best CAD or software does not have a customer service line. Think Google and have many times you have needed to call them. We need to make great products like that for the industry.

67. etnapowers
October 13, 2014

@goafrit2, I agree with you, it might be very expensive, only big companies might be able to do that or a middle size company which bets on a high profit project that could justify the costs of this solution.

68. etnapowers
October 13, 2014

@goafrit2: Google is a good example of a great product that not requires support from the provider, but in the industrial environment this is a situation very desirable but at the same time very difficult to achieve. Let's consider for example the case a tool for automatic test equipment that is compatible with Windows XP but not with Windows 7, a support engineer team is necessary for the conversion, even if the tool works very well.

69. etnapowers
October 13, 2014

@goafrit2: that's absolutely true, medical devices must be effective and reliable on the long term basis. the companies that produce devices for medical applications must satisfy very stringent requirements and the importance of this compliance is very high because a human life can be directly impacted by a malfunctioning of the device. It is desirable, when possible, that the designers of the device take into account the possibility of a failure and insert a emergency subroutine which can guarantee a temporary functioning until the failing device is replaced.

70. goafrit2
November 3, 2014

>> Let's consider for example the case a tool for automatic test equipment that is compatible with Windows XP but not with Windows 7, a support engineer team is necessary for the conversion, even if the tool works very well.

The major problem is that before you can get that support engineer, you need to be booking a lot of order and that must be a big part of the company revenue. If Apple is your client and books 80% of your revenue, nothing stops you relocating to their office to make them happy!

71. goafrit2
November 3, 2014

>> It is desirable, when possible, that the designers of the device take into account the possibility of a failure and insert a emergency subroutine which can guarantee a temporary functioning until the failing device is replaced.

I have desiged a gyroscope for a medical application. You are not just designing for your client, you are doing same for what FDA will allow to go through. That is why the cost factor would be a factor of 50x between a part for medical and consumer sector.

72. uchiha
November 4, 2014

@goafrit: Yes indeed its something that we cannot do anything about other than wait and watch. Anyway what I feel is that if it's a good company like Apple, why bother so much. They will take care of the rest in a much more better way.

73. fasmicro
November 5, 2014

>> Anyway what I feel is that if it's a good company like Apple, why bother so much.

The question is why is the big company having all the fun in the medical industry? Is there a way to ensure compliance and strong regulation while giving the small guys room to compete.

74. etnapowers
November 13, 2014

@goafrit2: you're right, the volume of business involved is a key parameter to unlock some situations and to purchase the services of a support engineering team. another important aspect is the effectiveness of the support. If the results in terms of quality are good, the incoming business will imply a profit that justifies all the expenses to reach the goal, included the costs related to the support team services.

75. etnapowers
November 13, 2014

@goafrit2: the ICs for medical applications must satisfy very stringent requirements in terms of reliability. This is due to the importance of the application and this justifies a high cost, as you correctly said. The gyroscope for medical applications that you designed for example, has to be reliable on a long time basis, and, if an unexpected malfunctioning appears, there should be the chance to replace the part or to fine tune it, to ensure that the overall system continues to work well even in the case of a problem of an important component like the gyroscope is.

76. fasmicro
December 2, 2014

>>  you're right, the volume of business involved is a key parameter to unlock some situations and to purchase the services of a support engineering team

I have been a fan of companies like Google which does not have a lot of support because they simply make great products that do not need a lot of support. I hope companies need to pivot towards that paradigm instead of the old model, you buy product, you hire bunch of guys around the corridors for years

77. fasmicro
December 2, 2014

>>  This is due to the importance of the application and this justifies a high cost, as you correctly said.

The experience in the medical electronics industry tells me the importance and value of having lawyers that can sue companies. I see everyone trying to make reliable products to avoid trial lawyers calling in coming years over failure.

78. etnapowers
December 15, 2014

“I hope companies need to pivot towards that paradigm instead of the old model, you buy product, you hire bunch of guys around the corridors for years”

@fasmicro: the trend of modern electronic companies is in the direction of Google model, because I guess that it is an effective mode to work: the effort in Research and Development is very important to ensure the success of a company over time and to avoid the presence of “old-style companies”.

79. etnapowers
December 15, 2014

“The experience in the medical electronics industry tells me the importance and value of having lawyers that can sue companies. I see everyone trying to make reliable products to avoid trial lawyers calling in coming years over failure.”

I think that lawyers play an important role to preserve patients from a failure of the electronics medical devices like pacemakers whose reliability must be excellent over the time. For sure it is not simple to guarantee such a reliability from a technical point of view, but I think that the entire control loop works well in case of the presence of lawyers who might sue companies in case of a malfunctioning due to mistakes of the developers of medical electronic devices.

80. Arismart1
February 6, 2018

Thank God for lawyers, for one time I see a benefit in what they do!

81. MelBrandle
July 29, 2018

In setting up circuits, a minor change could lead to a huge difference. One wrong point could mean a total failure or the ability to increase the overall voltage. This is the importance that we need to take note of before we end up with the wrong circuit setup. This is why it is always critical to read of the requirements of the circuit before proceeding to setup anything at all.

This site uses Akismet to reduce spam. Learn how your comment data is processed.