This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

VGA Video Splitter

Other Parts Discussed in Thread: OPA2674, OPA3693, THS3121, THS7327, LMH6733, LMH6738, LMH6739, VCA820, LMH6703, THS3201, THS3202, LMH6704, OPA695, OPA694, OPA3695, LMH6715, OPA2691, OPA2690, THS7374, OPA615


I have to design a precision RGBHV video splitter, 1x2/1x4/1x8. Could anybody suggest me with closest parts and schematics that does the work. Most of my experience is on digital side and this would be my full-fledged analog project. Reference designs would be really helpful to kick start immediately.



  • Hi Albert,

    But my requirement is for RGBHV, so can I use the same circuit for all the video components, R,G, B individually?

  • I think so, although (mind you) I have not tested this schematic, only found it.

    For the H and V signals you may be able to use another circuit, with simpler opamp, since bandwidth requirements are not as big as those of color components.

  • Honestly, I did find similar circuits but I couldnt proceed for the same reason that they are not accompanied by sufficient literature to support the reliability and moreover my requirement is for customer product. However, I can test it while looking for better designs.

  • There are a few amplifier options for the RGB part of this need including the OPA3693. I was originally thinking the OPA2674 but it is a dual so it may not be as "ideal" but the benefit is it can drive alot of current (i.e. alot of video loads on each output). The OPA3693 was not designed to drive as many loads, but it can definately drive 2 loads on each output, and probably 3 comfortably. Be sure to use +/-5V supplies for this to simplify things. 5V single supply is possible but you will need some more components around it to make sure the signal is biased properly and within the linear input/output range.

    As for an application circuit, see figure 50 (pg 16) in the THS3121 datasheet- very simplified but gets the point across - note +/-5V supplies should also work well. A real application circuit should also include ESD protection on each output (I recommend BAV99 or BAT54S as they work well and are low cost) and a 10pF cap on each connector for EMI concerns. This part may also work well as it can drive alot of current, but the bandwidth may be a little bit of an issue depending on how fast of a RGB signal you want to drive. Definately fine for VGA and SVGA, but if you want to do QXGA then you need a faster part like the OPA3693.

    For the HV distribution, I am not a digital guy but I suspect simply using 5V logic buffers should work fine.

    Hope this is helpful.



  • You may also check THS7327 and similar

  • Thats really educative Randy. Thanks a lot. Do you suggest not to use dual op-amp though it can drive multiple loads. Like I said, i would require 1x2, 1x4, 1x8 distributors, obviously high-current outputs are required. On the other hand, if 3693 can be used in a cascaded  model without compromising the quality, then 2674 can be avoided.

    Coming to the BW, as I am studying this, I understand that resolution of the picture decides that BW of the op-amp to be used, Am I right? Thats what you meant by fast RGB signal?

  • THS7327 as a buffer, is it something what an amplifier does or diff? In digital, buffer is used as something that stores data, I was wondering if 7327 does the same thing?

  • A buffer is a very generic name for many functions. Even in the digital world, you have gates that are inverting and non inverting buffers, and whose functions is to increase the fanout of a digital line. In this case, the buffer name is similar, but in the analog world.

    I think that you should read the datasheet and search for similar products in the family. Although the device has an inverted function from what you want (not a splitter, but a mux), it could be that it can make your desired function. For example, if you shortcut all three inputs of one channel, you would have a one to three splitter. I don't know if such a connection between inputs is OK for the video analog signal.

  • A dual op-amp is fine, it just lends itself to have an extra amp doing nothing in a RGB 3-channel situation. Nothing bad about it though.

    If the OPA3693 solution is used, or any other solution where multiple amps are needed on a signal, then I would recommend to simply connect all the + inputs together (parallel approach) rather than doing a cascade approach (where another amp is fed from the previous amp).

    As for bandwidth, here is a list of common resolutions (not all encompassing but the most common) and required amplifier parameters based significantly (but not entirely) on VESA VSIS specifications:

    VGA - 640 x 480 x 60; Pixel Frequency = 25.175MHz, Analog Bandwidth (-3dB) required = 22.7MHz, Slew Rate = 112V/us

    SVGA - 800 x 600 x 60; Pixel Frequency = 40MHz, Analog Bandwidth (-3dB) required = 36MHz, Slew Rate = 179V/us

    XGA - 1024 x 768 x 60; Pixel Frequency = 65MHz, Analog Bandwidth (-3dB) required = 58.5MHz, Slew Rate = 291V/us

    SXGA - 1280 x 1024 x 60; Pixel Frequency = 108MHz, Analog Bandwidth (-3dB) required = 97.25MHz, Slew Rate = 484V/us

    UXGA - 1600 x 1200 x 60; Pixel Frequency = 162MHz, Analog Bandwidth (-3dB) required = 146MHz, Slew Rate = 726V/us

    WUXGA - 1920 x 1200 x 60; Pixel Frequency = 193.25MHz, Analog Bandwidth (-3dB) required = 174MHz, Slew Rate = 866V/us

    As you can see, even the OPA2674 in a gain of +2V/V bandwidth (needed for this situation) will not be high enough to support all of these resolutions. But the OPA3693 definately can support these. Since we are talking about high speed amplifiers, be sure to use high speed layout techniques for any design.



  • I was going through a document which is calculating the parameters required to select an op-amp for this application.

    It says BW (-3dB nominal)  = BW (-3db min) x 1.5 and BW (-3db min)  = BW(signal) (-0.1db) x 6.55

    Using the above equs, required BW (-3dB nominal) with 0.1dB attenuation for a XGA and SXGA are 290 and 529MHz respectively, which falls in OPA3693 range but UXGA is out of its range. However, similar values for Slew-Rate are 1820 & 3321V/uS which is out of OPA3693 range.

    It calculate like this. SR(min) = 2 x BW x pi x Vpeak and SR(nominal) = 2 x SR(min). So do we need to identify another part which suits this equation.

    However, with a 0.5db attenuation, BW range will fall well within the 3693 range, however I'm not sure how much will that comprise the output quality with that attenuation.

    Also, say with G=+2, how can we decide what can be the cable length that can be used without performance degradation, assuming cables used are of reasonable quality.

  • Hi Hanumanthu,

    Regarding high slew triple amplifiers:

    These triple amplifiers meet the Slew Rate requirements you've quoted and have closed loop bandwidth > 750MHz:

    LMH6733 (3750 V/us)

    LMH6738 (3300 V/us)

    LMH6739 (3300 V/us)



  • The equations you show are a bit over optimistic about what is needed. In an ideal situation, having an amplifier meet those BW and SR needs is wonderful. But at some point you reach the point where no amplifier could meet those needs - and you are seeing this arise.

    The numbers I show above in this thread come primarilly from the VESA VSIS specification and should be considered the foundation for computer analog RGB signal characteristics. I do not think the OPA3693 should have any issue supporting the resolutions I mentioned above.

    As for cable length, the biggest factor is that the cable should be 75-ohm characteristic impedance. I have seen many cables that do not hit this spec and with very high frequency signals (ex: UXGA) this can cause some image issues. For length, the cable attenuation factor is the important factor. Usually this is specified by the cable manufacturer in terms of x-dB / 100ft or 1000ft or .... There will also be some frequency dependant attenuation too, but at the passband frequencies involved, it should not be a huge issue unless you use very very long cables with very hgih resolutions. One good thing is that most displays have an AGC function in the front-end ADC/decoder. This will compensate for attenuation of the cable, but generally not for frequency roll-off. At the end of the day, if you are <25 feet, then there should be no issue. If you are 100feet, then the cable needs to be very good and there is a chance that very high frequency roll-off may start impacting the sharpness of the display (not as sharp of a step transition). 

  • Hi Randy,

    How far the 0.1dB and 0.5dB attenuation factor effects the signal quality. Are the freqs you have mentioned will change if I'm looking for 0.1dB attenuation?

  • Hanumanthu,

    0.1dB can be a difficult spec to hit. Some stray capacitance in the wrong spot, or PCB material choice, or connector quality, or... could impact this point very easilly. 0.5dB is a little more reasonable to look at.

    In terms of "signal quality", this can be very subjective and is left open for alot of interpretation. Professional video systems strive for the upmost in quality and are willing to spend significant amounts of money to achieve excellent specs - such as 0.1dB flatness. However, consumer is significantly more relaxed and cost is more of an issue.

    One way to think about this is what will be the impact on the voltage? With -0.1dB down, this would be a ~0.692Vpp signal at the display when 0.7Vpp is perfect, or -1.14%. With 0.5db down, the signal is ~0.661Vpp or 5.6% error. For a general person, seeing 5% difference can be difficult. For someone with a "Golden Eye", they may be able to see it.

    I would refer you to the VESA VSIS specification. In this it states that the Max voltage shall be 0.7V +0.07/-0.035V , or +10%/-5%. Video channel-to-channel matching shall be 6%. Voltage at the display shall be better than or equal to -6dB at the pixel frequency, -4dB at 0.85xPixel Frequency, and maximum of +/-2.4dB (voltage) up to 0.8xpixel frequency. As you can see, the standard allows quite a range with respect to frequency response. This is the minimum (or maximum depending on the item) and the system should strive to be better than this. But how much better is up to you.



  • Can you direct me to some PCB design guidelines and other features like taking care of stray capacitance etc that should be taken care off, atleast what TI has identified.

    Also, one document is talking about AC Vs DC coupling for such boards and understood that either have their pros and cons. However, professional players employ DC coupling though it has its own disadvantages. Pls advise. 

  • Also, my requirement is to drive a cable length of upto 160-170 feet, while it can be assumed that OPA3693 can source upto 100ft without compromising the quality. In such case, do I need to use cable drivers?

  • Hi Hooman,

    How many VGA outputs can LMH6738 drive? In general, what is output current required to drive each video signal, R/G/B for 2 VGA ports?

  • Each part identified in this thread has application notes in the back of each datasheet that has several very good points about high speed layout. I would highly recommend you look at those. Some other app notes on high speed layout include:


    Even thogh SCAA0082 is focused on clocks, the information is very relevant to any high speed design consideration. There is probably a lot more app notes out there, but these come to the top of my mind.

    As for AC or DC coupling, the professional world likes to use DC coupling. Computer outputs are also DC coupled (I don't remember the last time I saw a non-DC coupled computer RGB output). But, AC coupling is also acceptable and the good thing is it is considered to be 100% compatible with any receiver/display. There are occasional instances where DC coupling may have had some issue, but this is usually with very old displays. The only real drawback of AC coupling is cost for the large capacitors (I recommend at least 470uF) and the droop/tilt associated with AC coupling (hence the large value caps needed). If you do go the AC coupled route, I recommend to place a 0.1uF cap (possibly 0.01uF) in parallel with the 470uF cap for much better high frequency performance.


  • The OPA3693 or LMH67xx parts can both drive long cables without issue as long as proper double termination (and cable impedance) is utilized. In this usage, these amplifiers are cable drivers. The biggest concern in your situation is the cable characteristics and not the amplifier. IF you know that the cable will always be say 100feet, then you can do some pre-equalization in your amplifier design. There are a few notes about this including:

    a) Figure 82 of the VCA820 datasheet





  • Hi,

    In terms of output current capability to drive a video cable (usually with series 75ohm and back terminated in 75ohm and with 1V video at the load end), the current is 13.3mA (= 2V / 150ohm). The LMH6738 linear output current is 90mA, so it should handle multiple loads, from this point of view.



  • Randy,

    Would you recommend THS3101/3102 for achieving 0.1db flatness (upto certain extent), though the other parameters are an overkill? 

  • Do you mean THS3201/THS3202? If so, then these will also work similiar to the OPA3693/LMH673x parts already discussed. However, you need to make sure you compare G=+2V/V specs and then you will see they are all similiar in terms of bandwidths. There are other single amps that could be used including OPA694, OPA695, LMH6703, and LMH6704.

    Antoher option is OPA3695. But the integrated resistors on the OPA3693 and LMH6739 simplify the layout and reduces layout parasitic risks.

    Again, the layout, selection of RF/RG resistor values, and cable will be more critical than selecting any one of these amplifiers. One advantage of using the OPA369x/LMH673x parts is they are pin compatible so it would be easy to swap parts in testing.



  • Speaking of S-Video distribution amplifier, I have identified OPA2691 and LMH6715 as ideal opamps for 1:3 mode. Would you suggest any other op-amp. However, I couldnt find a reference circuit for this. Can you help me with one?

  • S-Video Y channel needs to support up to say 6MHz flat and the C channel only needs to support from say 1-5MHz (not exact numbers but to get in the right ballpark). The parts you mention will work fine to support these bandwidths.

    Realistically, any dual amplifier that has a -3dB BW of > 90MHz while in a gain of 2V/V should work very well with excellent performance in the passband. You selected Current Feedback (CFB) amplifiers, but there are similiar Voltage feedback (VFB) parts, such as the OPA2690, that has similiar performance. So you do have options.

  • I have attached a file for 1x4 video amplifier. Would you suggest any changes to it.

  • I am attaching a Tina File showing the recommended configuration. When you AC couple, you must always establish a DC reference point. So a resistor is being used to create this. The High Pass corner frequency should be <= 2.5Hz when AC coupling, so hence the RC product values shown. The resistance needs to be kept fairly low due to the input bias current of the non-inverting nodes of the each amplifier. This bias current creates a DC offset (Ibias x R) which then gets multiplied by the gain (2V/V). Note that this assumes +/-5V is being utilized.

    The other thing AC coupling does is to average the input signal around the DC bias point (GND in this example). So the signal's back porch/Hsync voltage level will vary based on signal content. If this is not desireable then DC coupling is the simple answer, if possible. Otherwise a DC restoration circuit (see OPA615) can be used with AC coupling or if precision is not critical, then a part like THS7374 may be an option to combine a simple DC restore + Video amplifier + filter into one package with single supply operation.



  • I got it now. Actually I forgot to add a 910E along with 470uF in the circuit, though.

    Varying of the HSync with AC coupling, will this effect the final quality? If DC coupling doesnt effect the HSync, is that advisable by all means for a better quality output, considering precision an important parameter?

    By the way, why should the High pass corner frequency be <= 2.5Hz?

  • Most systems today are AC coupled on their front-ends. If true, then the AC coupoling of your system is irrelevant. However, there is a slight chance that some systems may expect DC coupled (sync below GND) signals. In this scenario, AC coupling would not work very well due to the averaging effects with an active video signal moving the DC voltage of the backporch/sync level around.

    The general rule of thumb is to have the High pass corner frequency be at least 10X lower than the lowest pass frequency - which is 60Hz or 50Hz. This keeps the phase variance at the 50/60HZ region pretty flat. Then there is the issue of line tilt. Using too high of a frequency would cause too much tilt in the signal. Usually a 100% white signal for the line is used to examine the tilt. Tilt is the DC voltage delta between the start of the line and the end of the active video line. I have heard it is preferred to be less than 1 IRE (~7mV) for high quality systems, but each designer has to make their own choices. I know some designs have used a HP corner of 1Hz or less to minimize this issue.