Editor’s note: As you can see, Michael Steffes puts a great deal of ‘blood, sweat, and tears”, with a generous helping of his long and extensive Op Amp knowledge, into these blogs. This series of highly technical and informative blogs are awesome tutorials for all levels of designers. It’s the kind of articles I would cut and paste from print magazines at EDN and EETimes for my future designs when I was a circuit designer for 40 years. I hope you enjoy these bits of wisdom and experience that will surely help you in your work and pique your creative imagination as well as stimulate your brain’s knowledge centers. Enjoy.
Some years back, an excellent blog appeared on EDN stepping through common issues and questions surrounding lower speed Voltage Feedback Amplifiers (VFA) (ref. 1). Those first appeared, with questions and comments, on the TI E2E forum where it seemed a fair number of those queries were on similar issues with higher speed op amps. At that time, those types of questions were referred to the ‘TI high speed E2E forum’. Those may well have been answered there, but a wider discussion of the unique issues that come with higher speed op amps will be the intent here.
While my colleague Bruce Trump did a fabulous job in succinctly stepping through quite a few op amp issues for the slower VFA type of device (ref. 2), in high speed we have the VFA plus Current Feedback Amplifiers (CFA) and Fully Differential Amplifier (FDA) types that bring their own set of issues. Very roughly speaking, most vendors segment off devices with >50MHz “bandwidth” as being in their high-speed product mix. This discussion will lean heavily on Bruce’s original blog for VFA types, but will add any unique or new considerations for CFA’s and FDA’s in the most pertinent topic areas. This first blog will pick up the I/O range issues (ref. 3) and include industrywide parametric tables where useful. To get some overlap, VFA’s down to 20MHz bandwidth will be included in those tables.
It might be fair to ask, “Who the heck is Michael Steffes?”. After 8 years of analog IC design at IBM and StorageTek, I concluded my high-speed analog IC design career contributing (slightly) to the design of one of the first integrated CFA op amps, the CLC400 – which is basically an IC version of the hybrid CLC231. From there, I traversed a 30-year career in applications, marketing, and business management across 5 suppliers, always focused on high speed amplifier development, introduction, and design-in tools. Along the way, I participated in the introduction of over 70 high speed amplifiers while publishing >100 application notes plus contributed articles. My product development contributions were always some combination of device definition, specification line items and limits, characterization plots, applications text, macro-modeling, and design-in support. Similar to Bruce, I also have had the good fortune to work closely with some of the top analog IC designers where basically no amplifier question was off limits. All of the best questions come from customers – where having those IC designers the next cube over for help is almost unfair.
Bruce Trump’s “”The Signal” came out weekly covering only VFA devices. This monthly blog will necessarily be a little longer covering the 3 types of high speed devices appearing in modern amplifier portfolios. Questions are welcome.
Input/Output Range issues for higher speed Voltage Feedback Amplifiers (VFAs)
In IC design, the output swing range is a separate issue from the input swing design requirements. The working assumption here is that overdrive into the swing limits is best avoided. For all speed ranges and topologies, the outputs usually have a symmetric swing range relative to the supplies. If it is Rail-to-Rail Output (RRO) it is Rail-to-Rail towards both supplies. If it is not RRO, it is not RRO towards both supplies. Also, if the outputs are not RRO, then the input range is usually not Rail-to-Rail Input (non-RRI) with a few exceptions. Again, op amps have no ground pin so any combination of voltages on the two supply pins are allowed as long is the total supply voltage (Vs) across the device is in the specified operating range. The available output swing is always a “headroom” to the supplies; how much overhead voltage is needed to allow for “linear” operation. Rail-to-Rail output stages reduce that headroom as far as possible. But there is still a 10mV to 200mV overhead requirement for proper op amp operation. Non-RRO outputs require much higher headroom, but often offer other benefits in terms of output linearity vs quiescent power.
The early op amp “slang” for output swing was an available +/-Vout on some specified supply voltages. That is a bit cumbersome where a more useful “headroom to the supplies” sometimes appears in more recent product releases like the OPA838 specification (ref. 4) shown in Figure 1. This headroom specification remains the same across any combination of supply voltages.
Example I/O headroom specification for the Negative Rail In (NRI) and RRO OPA838.
At the same time, a relatively new, and very useful, output swing range vs output current plot has come to be a standard RRO characterization curve. This “claw” curve has emerged as a typical way to give a little more information on the increased headroom requirement that comes with increasing source/sink output current and is even included in the most recent TINA (ref. 5) macromodels (ref. 6). Figure 2 shows this curve from the RRIO, 38MHz, OPA350 data sheet (ref. 6). The left scale is the available output swing referenced off the +/- supply voltages (V+ & V-). The delta from the top and bottom maximum is the headroom vs. source or sink current.
Example “claw” curve for the RRIO OPA350.
This claw curve is in fact one way to estimate another very important op amp output specification – linear output current. This is easily one of the more difficult specifications to either deliver, or extract out consistently, across decades of data sheets from different vendors. Lacking any more definitive specification, the parametric tables shown below used a 0.5V headroom limit from this type of claw curve (if available) to estimate a “linear” output current using the minimum of source or sink. This 0.5V headroom with load current is approximately at the limit of where you would call a part RRO (Rail-to-Rail Output) and crossing over into the regime of non-RRO devices. The other approach used, lacking this curve or a clear specification line, is ½ of the short circuit current. A more recent effort at showing an available “linear” output current appears in the unity gain stable OPA837 (ref. 8) data sheet as Figure 3. The “A” at the far right indicates 100% ATE screened where the conditions show those test setups and limits on a +/-2.5V supply.
More recent “linear” output current specification
While very useful, the “claw” curve of Figure 2 is implicitly showing a 2-quadrant output VI capability. Some older high-speed op amps like the OPA690 (ref. 9) attempted to show a more thorough 4-quadrant VI capability as illustrated in Figure 4. While there is a lot of information crammed into this plot (including the claw curve portion), it didn’t generate much interest and has largely been dropped at this point from the TI datasheets – while it does show up in the newer Microchip data sheets (Figures 4-6, ref. 10). Note the higher output currents available from this 5.5mA quiescent current device. That is the main advantage of a non-RR output design in that they can have much higher “power gain” from the required quiescent supply current to the available output current.
Example 4-Quadrant output VI operating range for the non-RRO OPA690.
Input Range Options for High Speed Voltage Feedback op amps
Unlike the two options for output swing range, there are three typical input range options found in wideband voltage feedback op amps. There are many older non-Rail-to-Rail Input (non-RRI), devices that are usually also non-RRO (with some exceptions). These require typically 1 to 2V of input operating voltage headroom to the supply pins. As single supply, RRO VFAs became prevalent, the earliest offered Negative Rail Input (NRI) is where the input can usually swing to the negative supply pin but requires typically 1 to 1.5V headroom to the positive supply. This quickly migrated to numerous CMOS and bipolar devices offering a Rail-to-Rail Input (RRI) with minimal voltage headroom on the input pins to the supplies. As ref. 3b explains, there are numerous tradeoffs to the various techniques to deliver RRI.
As the application bandwidth increases, there are relatively fewer applications that require RRI to the positive supply – where a high side current sense application stands out as one. For single supply, DC coupled, applications, there are many more applications that only require an input range to the negative supply for an input swing from ground to some positive voltage. For high speed complementary bipolar op amps, NRI only requires a PNP input stage where many more of those types are found vs the relatively fewer bipolar RRI devices. For higher speed applications, some of the reasons a positive rail input might “NOT” be required include:
- Normally operating at some gain. For non-inverting operation this reduces the swing requirement on the input vs the output. There are relatively few Rail-to-Rail Input/Output (RRIO) unity gain buffer applications above 50MHz.
- Often operating AC-coupled, so the input DC operating voltage can be set wherever it operates best and (if gain>1) the input swing is safely within the supply rails.
- Where inverting operation can be used, the input pins stay at the bias voltage on the non-inverting input with no voltage swing on the inputs.
Most NRI with RRO op amps, operating on a single supply with ground on the negative supply pin, cannot truly swing completely to ground on the output with good linearity. Most PNP input stages can accept true ground inputs for single supply operation. There are a range of circuit techniques that can be used to move a 0V → +Vin input swing, with DC coupled gain, to operate with some headroom to ground on the I/O pins, improving the signal path linearity. One of those techniques appears in the OPA830 (ref. 11) data sheet as shown in Figure 5. Here, the supply is used to level shift the input (and hence the output) slightly above ground where now the 0V to +Vin input swing operates at the output with a bit more low-end headroom. This approach does require the source signal to sink that level shifting current and that level shift supply now has a signal path to the output voltage. Where PSRR might be a concern, use a reference voltage for this level shift if available.
DC coupled, single supply, 0V input to level shifted output approach.
The more prevalent issue in bipolar NRI with RRO devices is on the output pin. PNP inputs used for NRI can usually swing with good linearity to ground but most RRO devices lose linearity on the output if forced to within 200mV of ground. That 200mV of output headroom can also be easily provided using the fixed -0.23V negative rail generator – the LM7705 (ref. 12) as the negative supply (as suggested in the comments of ref. 3c). Using a switching regulator like this does raise PSRR concerns in op amp applications. As will be seen later, bench testing this approach with the LM7705, using FDAs, shows complete switching noise rejection for those differential output high speed amplifiers. One device (LT6360, ref. 13) includes a built-in negative supply generator to get the linear swing to ground.
Representative higher speed VFA devices with different I/O range capability
Modern “complementary” bipolar (vertical PNP’s) high-speed monolithic VFA’s began to appear in about 1989 from multiple suppliers. With almost 30 years of product development behind us, there are a vast number of devices to choose from. A small sampling of those, with some key parameters, appear in the following 3 tables. These parameters are from a best effort review of the public data sheets at some point over the last 10 years. While there are hopefully minimal errors in these tables, a close review of the most current device data sheet would be prudent for any designer considering them.
The max operating supply voltage (max Vs) is a clue to the process technology. Higher voltages are usually bipolar (HV CMOS is more common in slower devices) while lower maximum Vs (<7V) are commonly CMOS or complementary SiGe. The input bias current also gives a clue to whether the device is bipolar or CMOS/JFET. Where the input bias current is in the μA region, probably bipolar, while if it is in the pA region, probably CMOS/JFET. There are always exceptions – some bias current cancelled inputs show nA type input bias currents but are in fact bipolar inputs. Some lower noise CMOS inputs show >1nA input bias current along with higher input capacitance.
To trim these tables down, the following screens were used. Also, a swing to rail designation included anything with <0.5V headroom. These non-RRO devices are generally the oldest, as newer (after 2008) designs normally target RRO. Most of these non-RRO will be complementary bipolar as shown by the relatively high max. input bias current. Generally, the original supplier is listed as that links to the part number prefixes used (e.g. LMH means “Linear Monolithic Highspeed”). Three of these non-RRO devices include one rail on the input swing range.
- To get more modern devices (remember, the full list goes back to 1989), screen to 1k price <$2.20, max Icc < 12mA, output headroom <2V, input spot noise < 12nV/√Hz.
- Single channel, unity gain stable, VFA only.
- Unity gain bandwidth in the range of 20MHz
Non-RRO high speed VFA, single channel, unity gain stable.
The RRO VFA devices can be broken into the NRI variants and RRI devices. There are a very few RRO VFAs with neither input rail included to get lower input noise as shown in Table 2. To trim this RRO, with NRI table, down, the same screens as Table 1 were used except now RRO only and all RRI devices are excluded. Clearly, there are more CMOS devices (Ib <1nA) and more numerous devices with Vs (max) <7V. Also note the maximum unity gain SSBW came down quite a bit vs. non-RRO (450MHz vs 900MHz). Rail-to-Rail output stages have more propagation delay limiting the unity gain stable closed loop bandwidth. Three of these RRO devices include neither rail on the input to deliver lower input voltage noise options.
RRO high speed VFA, single channel, unity gain stable.
And lastly, those higher speed VFAs that are considered RRIO. Again, the same screens as Table 2 with RRI only. These rail-to-rail input stages typically include either a crossover network at the input or an internal charge pump to bias the input stage above the positive supply voltage (ref.3b). Sometimes, the part description BW does not match the unity gain SSBW – for reasons that will come up in a later blog.
RRIO high speed VFA, single channel, unity gain stable.
For simplicity, these tables have excluded quite a few VFA variants. Those excluded were decompensated devices, multi-channel devices, variants with or without disable where only the disable version was included here, automotive or high rel, and discontinued devices. These do give a good cross-section to get started with across the different suppliers.
The next installment for this “The Signal Sped Up” blog will extend this I/O range discussion to Current Feedback Amplifiers (CFA) and Fully Differential Amplifiers (FDA).
References to blogs on I/O ranges in high speed amplifiers.
- Full collection of the original article series. “The Signal: A compendium of blog posts on op amp design topics”.
- Specific blogs covering input output range issues for VFA op amps. These include the Q&A:
- TI OPA838 op amp, 1mA, 300MHz Gain Bandwidth, Voltage Feedback Op Amp
- TINA simulator available from DesignSoft for <$350 for the Basic Plus edition. Includes a wide range of vendor op amps and is the standard platform for TI op amp models.
- TI OPA350 op amp, High Speed, Single Supply, Rail to Rail Microamplifier series
- TI OPA837 op amp,Low Power, Precision, 105MHz, Voltage Feedback op amp
- TI OPA690 op amp, Wideband Voltage Feedback Op Amp with Disable
- MicroChip MCP651 op amp, 50MHz, 200μA, Op Amps with mCal
- TI OP830 op amp, Low Power, Single Supply, Op Amp
- TI LM7705 regulator, Low noise negative bias generator
- LTC LTC6360 op amp, Very Low Noise Single Ended SAR ADC Driver with True Zero Output