Blog The Signal Sped Up

DC precision considerations for high speed amplifiers, Insight #3

The development of higher speed amplifiers, from a DC precision perspective, has been one of improving that precision while also providing higher slew rates – which ties to available Large Signal BandWidth (LSBW). The highest slew rates intrinsic to the Current Feedback Amplifier (CFA) approach come with relatively poor DC precision. Enhancing the available slew rate in the Voltage Feedback Amplifiers (VFA) has progressed through a number of approaches that will be shown here. The Fully Differential Amplifier (FDA) offers both options where the CFA-based versions provide very high slew rates, with poor DC precision, while VFA based FDAs have progressed using similar techniques to improve precision with higher slew rates as the core VFAs. The FDA brings a number of new DC precision considerations due to the added common mode control loop (ref. 1). This part 3 will progress through the different VFA approaches with Part 4 moving on to CFA and FDA DC precision considerations.

Paths to Increasing Slew Rate in VFA Op Amps

Achievable DC precision (and noise) is largely an input stage design issue. The VFA brings some natural advantages to lower DC errors. Recall, the first level DC errors for the VFA are the three input error terms. Using these to calculate the output error has been covered by many earlier references (ref. 2).

  • Input offset voltage and drift
  • Input bias currents and drift
  • Input offset current and drift

Occasionally, some other error terms are lumped in with the input offset voltage (ref. 3). While it is certainly true the physical input error voltage will have a Vout /Aol term, since that term is output signal level dependent, it is properly captured by a gain error to the output. The input error voltage due to open loop gain (Aol ) is not added to the static DC input offset voltage. It is instead accounted for by the gain compression term described by LG/(LG+1) where LG = Aol /NG, NG = Noise Gain. So, what is the input offset voltage when the VFA op amp is operating at unity gain, split balanced supplies, with a grounded non-inverting input? Since the output voltage will be operating very near ground, internal to the input differential stage, there is an error voltage that should be very nearly 0V. The observed external offset voltage is the required compensating voltage for input stage imbalances to get the actual error voltage internally to zero.

As classic VFA Gain Bandwidth Products (GBP) steadily increased through the early 1980s, their slew rates did not keep pace where a typical unity gain stable, high-speed, Harris op amp (ref. 4) would have a Small Signal BandWidth (SSBW) far exceeding the available LSBW. The two dominant approaches to increased LSBW in VFA devices through about 2004 were to use an input stage that emulates CFA operation or to provide de-compensated amplifiers. Non-unity gain stable, de-compensated, VFA devices reduce the compensation capacitor value and often increase the input stage gm (lower degeneration R in the input stage giving lower input noise as well). Hence, for the same approximate slewing current available at the compensation cap, more slew rate can be made available. Improving DC precision and noise while increasing slew rate is considered in the context of lowering supply current as well. It is always possible to increase slew rate and reduce input voltage noise with more quiescent current. Delivering solutions that improve both at ever lower quiescent current is implicit in this discussion.

First, let’s review unity gain stable, non-slew enhanced, high speed (>20MHz) VFA options. Slew enhanced unity gain stable VFA can be sorted out by ranking a descending estimate of LSBW/Icc . Those will appear in later tables, but for now isolate down to modest slew rate, lower noise, precision, unity gain stable VFA. As described in ref. 5, many of the more recent VFA introductions include Rail-to-Rail Output (RRO) along with Negative Rail Input (NRI) or Rail-to-Rail (RRI). Some of the lower noise devices are non-RRIO designs. The RRI devices include either a x-over network or internal charge pump to bias the input stage beyond the supplies applied (ref. 6). To show better DC offset devices over the total useable input range, RRI devices will be excluded until Table 4. Unity gain stable VFA devices long struggled to deliver low input voltage noise where several recent releases seem to have broken through that limit – initially showing higher supply current or much lower slew rate but with improved solutions included in Table 1. There are also many low offset, higher speed, unity-gain stable CMOS or JFET input devices. Those generally show higher (>4nV/√Hz) input voltage noise and are not as power efficient in their LSBW/Icc ratio. They will have the benefit of removing input bias and offset current errors in the output DC offset calculation.

This Table 1 will show representative, non-slew enhanced, single channel, unity gain stable VFA devices that offer both low offset voltage (<=1mV max) and lower input voltage noise (<4nV/√Hz). The next 3 tables will progress through different approaches to increase the slew rate. These low noise and offset devices are all bipolar. Low output offset will therefore require bias current cancellation in the design (ref.2). Done correctly, that will reduce the output DC error due to input bias currents to Rf *Ios . Note, once again, the highest speed devices require a non-RRO design. The reported gain of 1 SSBW shown here often exceeds the true Gain Bandwidth Product (GBP). For instance, the fastest 800MHz OPA820 actually shows a data sheet GBP of 240MHz. This is again due to phase margin at unity gain crossover for the LG being < 65deg extending the closed loop bandwidth considerably (a topic for a later blog).

To reduce this table of single channel VFA, the following were screened out –

  1. Slew Rate >400V/μsec
  2. Max Vio >1mV
  3. Input noise voltage > 4nV/√
  4. Rail-Rail Input (RRI) devices (those will appear in Table 4)
  5. 1k MSRP > $3.00 to get more active devices for new design.
  6. Obsolete Devices
Table 1

Click here for larger image 
Unity gain stable, non-slew enhanced, single, VFA devices in descending SSBW.

Unity gain stable, non-slew enhanced, single, VFA devices in descending SSBW.

The most common approach to delivering a VFA solution that offers higher slew rate is to decompensate the op amp. This limits their applicability to higher gain (or transimpedance) applications, but can deliver very good DC precision, normally lower input voltage noise, and a higher LSBW in a given supply current range. While there are many JFET or CMOS input decompensated options, most of those either have higher input noise or offset voltage. To show a few very low input bias current decompensated options, expand the screens to remove –

  1. Input voltage noise > 3.0nV/√Hz
  2. Max. input offset voltage > 2.5mV
  3. Price >$4.00
  4. Obsolete devices

The slew rates are generally higher using the de-compensated approach while most devices also show a lower input voltage noise and offset voltage. There are relatively few RRO or NRI devices where one of the most recent (OPA838, ref.7) does offer that combination with a slew enhanced input stage, very low power, low noise, device. Generally, these devices offer a very low input offset current allowing input bias current cancellation techniques to be applied (ref. 2). The descending SSBW*Gmin sort usually far exceeds the actual GBP for each device as the minimum operating closed loop gain has a low phase margin extending the closed loop bandwidth at that minimum recommended closed loop gain. Gmin is the minimum recommended gain. Usually set to hit some safe minimum phase margin – typically in the 30deg to 45deg region. That Gmin will usually be peaking the non-inverting SSBW, but not unstable. So, the “Minimum Stable Gain” is kind of a misnomer, where “Minimum Operating Gain” is more accurate. Inverting operation for a decompensated VFA can operate at any gain (including attenuation) using the simple compensation technique of ref. 8.

Table 2.

De-compensated high speed, single, VFA devices, in descending Gmin *SSBW (at Gmin )

Table 2

Click here for larger image 
Slew Enhanced, Unity Gain Stable, Voltage Feedback Amplifiers.

Slew Enhanced, Unity Gain Stable, Voltage Feedback Amplifiers.

Slew Enhanced, Unity Gain Stable, Voltage Feedback Amplifiers

The earliest approach to closing the gap between SSBW and LSBW in unity gain stable VFA op amps applied matched input buffer stages to a transconductance element. This shows an external VFA feedback characteristic but then emulates CFA slew rates as an increasing error voltage increases the available slewing current in that resistor sitting across the input buffers. An early example is shown in Figure 1, (page 10, ref.9) where the input stage and transconductance element are highlighted. At the onset of slew limiting, the input error voltage increases – increasing the available charging current through the current mirrors to the compensation capacitor, “C” in Figure 1.

Figure 1

VFA input stage emulating CFA slew enhancement

VFA input stage emulating CFA slew enhancement

This approach can normally be identified by the relatively high input voltage noise that comes with it. Also, since the input buffers are not that matched, they usually show higher input offset voltage and currents. No offset current drift information is also a clue to this topology. Similar to the CFA amplifiers (that use similar input buffers), this approach to slew enhancement comes with generally poor DC precision and non-RRIO swing capability. It is sometimes difficult to identify this topology (although LTC provides Figure 1 in their datasheets for this type of part). Table 3 is a best effort extraction of a range of single channel, unity gain stable, very high slew rate devices similar to the design shown in Figure 1. Even though many of these devices have higher input noise and offset voltage, extreme values were cutoff in table 3 where parts excluded included (no screen on input offset voltage here) –

  1. En > 12nV/√Hz
  2. 1K MSRP > $3.00
  3. Non-unity gain stable devices
  4. Obsolete devices
Table 3

Click here for larger image 
Unity gain stable, very high slew rate, single channel VFA in descending Gmin=1 SSBW.

Unity gain stable, very high slew rate, single channel VFA in descending Gmin =1 SSBW.

While these devices essentially emulated the LSBW capability of CFA types, there is a lot of room for improvement in DC accuracy and noise. Also, similar to all CFA solutions, none of these can support swing to either rail on the inputs. The two sub 1mV offset LTC devices are silent on Ios drift in their datasheets.

Starting around 2004, higher speed VFA developments found newer ways to provide slew rate on demand with better offset and noise. Their aim has been to improve the DC precision and noise while adding combinations of swing to rail on the I/O pins. It is not clear how similar these are internally, but great progress has been made to fill in this gap in the designer’s toolkit. Again, these devices can first be identified by sorting a LSBW/Icc metric in descending order and looking for some combination of swing to the I/O rails (and removing the devices of Table 3). The simple metric used in Table 4 is a slew rate based 2Vpp output LSBW divided by the 25C max. quiescent supply current. Keep in mind, however, that any bench LSBW test of a slew enhanced op amp will show a rapidly increasing supply current as the test frequency increases and the slew boosting circuitry operates. This is true of all CFAs as well as the slew boosted VFA devices described here. That is, however, a darn hard plot to find in any of these slew-boosted device datasheets.

Some of the earliest efforts to increase slew rate in very low power devices were left with very high noise and offset voltage (ref. 10). More recent introductions have lowered both the input noise and offset voltage at reduced supply current. Here, we see many more RRO options and mainly NRI. Though not shown in Table 4, the input offset current and drift for these “precision” VFA devices are typically very low.

There are a few RRI choices where it is not clear if the slew enhancement functions when the input common mode is in the range of operation where the inputs have switched over to the upper stage (ref.6). The LTC RRI devices (ref. 11) specify slew rate in an inverting configuration to hold the input operating voltage fixed near midscale. The ADI RRI device (ref. 12) operates gain of +1 for the slew rate specification, but does not encroach into the crossover region 1.3V below the positive supply.

To trim Table 4 down, the following were screened out –

  1. En > 6nV/√Hz
  2. Vio > 0.5mV
  3. 1k MSRP > $3.00
  4. Obsolete devices
Table 4

Click here for larger image 
Precision, Slew Enhanced, Unity Gain Stable, Single Channel VFA in descending LSBW/Icc

Precision, Slew Enhanced, Unity Gain Stable, Single Channel VFA in descending LSBW/Icc

Higher speed VFAs have striven hard to improve their full power bandwidth while adding I/O range and DC precision features similar to the much larger range of <20MHz precision op amps. Where RRIO options are not required, and only modest slew rates, there are some very power efficient, low noise and offset, choices in Table 1. If your application can benefit from a decompensated solution, those bring intrinsically higher slew rates vs supply currents. Several recent releases in Table 2 deliver very low noise and offset at very low quiescent current (ref. 7). Where CFA type slew rates are required in a VFA architecture, those are readily available but none offer any swing to rail capability. As shown in Table 3, they also suffer the same poor DC precision as CFA’s and generally higher input voltage noise. The latest introductions in Table 4 improve the slew rates vs quiescent power while offering RRO and different swing to rail input options. They are not generally as good on input noise as those in Table1 and no where near the slew rates available in Table 3, but do expand the LSBW envelope at lower quiescent current with some impressive choices. Part 4 will continue this DC precision discussion for the CFA and FDA high speed amplifiers.

References for DC errors for high speed VFA

  1. Input and Output Voltage Range Issues for High Speed CFAs and FDAs, Insight #2
  2. Basic DC error calculations ADI app. Note MT-039 Op Amp Total Output Offset Calculation
  3. Bruce Trump -The Signal, Offset Voltage and Open Loop Gain – their cousins
  4. Harris CA3100, 38MHz, Operational Amplifier
  5. Input and Output Voltage Range Issues for High Speed Amplifiers, Insight #1
  6. Bruce Trump -The Signal, Rail-to-Rail Inputs – what you should know!
  7. TI, OPA838, 1mA, 300MHz Gain Bandwidth, Voltage Feedback Op Amp
  8. Michael Steffes, Aug. 1997, Unique compensation technique tames high-bandwidth, voltage-feedback op amps
  9. Linear Technology, LT6274, 90MHz, 2200V/μsec 30V Low Power Amplifiers
  10. Early slew boosted VFA with high noise and/or offset voltage
    1. TI, OPA830, Low Power, Single Supply, Wideband Operational Amplifier
    2. Intersil, EL8100, 200MHz, Rail-to-Rail Amplifiers
  11. Linear Technology RRI slew enhanced VFA’s
    1. LTC6246, 180MHz, 1mA Power Efficient Rail-to-Rail I/O Op Amps
    2. LTC6252, 720MHz, 3.5mA Power Efficient Rail-to-Rail I/O Op Amps
  12. ADI, RRI Slew Enhanced VFA, ADA4807

2 comments on “DC precision considerations for high speed amplifiers, Insight #3

  1. soufiane_ben
    November 30, 2018

    Hi Michael:

    I think some designers should start by selecting the “right” process. There are compimentary processeses which lend themselves better for optimized speed/power ratios. On the other hand, most “modern” engineers seem to fear the idea of a decompenated op amp, think OP37 which is still one of the best decomp op amps ever made in my opinioon. 

    I think that soon we will see increased needs for high speed and DC precision, this is particularly true for T&M and automotive but I can also see this need in medical instrumentation for instance. 

  2. Tucson_Mike
    December 5, 2018

    Hey Soufiane, 

    It is not just “modern” designers that approach decomp parts with some trepidition. It has always been uphill from the first decomp part I was involved in (the CLC425) to these most recent blazing fast decomp VFA like the OPA855 (8GHz GBP). If you are tasked with a transimpedance design, most designers recognize the need for decomp and deal with it. Otherwise, it is a tough sell even with clear input noise and LSBW advantages. There are certainly added stability risks – but good models can mitigate those. Looking industrywide, I would say LTC has made a commitment to decomp (even more than I have) while Maxim recently spun out a series of decomp CMOS devices. Nice knob to use in product development if you understand the tradeoffs. 

Leave a Reply

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