|Elliott Sound Products||Project 191|
Most of the time, most of us actually don't know how much of our amplifier power we're using when listening to music. Some amplifiers have clipping indicators that show when we've gone beyond the limits, but that isn't necessarily particularly useful. Unless one connects an oscilloscope to the speaker outputs of the power amp and watches it like a hawk to see the maximum level, we never really know if the amp is big enough to avoid clipping, or way more than we ever actually use.
Voltage isn't everything though, as some speaker systems can demand far more current than we may have expected, and this may be the reason the amp sometimes 'complains' by creating distortion, even though it's nowhere near clipping. Unless you have a way to monitor both voltage and current, you never really know if the current amplifier is up to the task. This project will normally be used in conjunction with a digital multimeter, which only needs to measure DC voltage.
The peak voltage and current are captured by rectifiers and peak hold detectors, which retain the highest voltage for as long as it takes to make a measurement. Depending on how well the peak hold part of the circuit is insulated, the voltage can be held for 5 minutes or more with only a small loss of the retained voltage. The TL072 has an input impedance that's claimed to be 1TΩ (1012 ohms - yes, you did read that right). Input current for the TL07x series of opamps is around 65pA (typical). The main limitations are the insulation resistance of the capacitor's dielectric and that across the reset switch.
This isn't a 'power meter'. It's designed to show the maximum (peak) voltage and current, and if you multiply the two together the result is not 'power' per sé, but is simply the maximum voltage and current provided by the amplifier. The maxima don't necessarily occur at the same point in time (or even at the same frequency), so you can't use them to measure the actual power delivered. What you can do is ensure that the two peak values are within the design range of the amplifier. Should you measure a peak voltage that's close to the amplifier's supply voltage, your amp is under-powered for the level that you listen to music. This indicates that a more powerful amp is needed, purely to get the required peak voltage without clipping. If you think that you need a wattmeter ('power meter'), then see Project 189 which is a true multiplying wattmeter.
Determining the actual power you need is difficult, because the detector cannot show any voltage greater than the amp can deliver. Turn down the volume to the point where it sounds about half as loud (or use a sound level meter and set the level 10dB lower), and run the test again. The peak voltage displayed is at a level of 10dB less than your 'preferred' level. If you multiply the peak voltage by 3.16, that tells you the peak voltage you need.
For example, let's say that your amp uses ±35V supplies and an 8 ohm load, and the peak detector shows (close to) 3.5V on the 50V range. This indicates that the amplifier is clipping on peaks, since the output voltage is equal to the amplifier's supply voltage. When the level is reduced so it sounds half as loud, you measure 1.5V on the same range (15V peak). Multiply 15V by 3.16 and you get 47.4 volts. That means you need an amplifier that can deliver at least ±47.4V peak - a 150W/ 8 ohms amplifier. Unless you have particularly low impedance speakers (or ones that are known for having an 'unfriendly' impedance curve), the usefulness of the current detector is not so great. It can be omitted, but for the sake of a few cheap parts, it's worth having.
If you have two multimeters, you can connect one to the voltage output and one to the current output, so you'll be able to see the peak value in 'real time'. If you reduce the volume, you must reset the integrators, as they will maintain the voltage stored for quite some time (if done very carefully, that could extend to an hour or more (but it will not maintain an accurate voltage for more than a few minutes at most).
The idea for this was prompted by an article published in Audio Express magazine in 2015. Despite some superficial similarities (as must be the case), the design shown here is different, in that a full wave rectifier is used to ensure that the highest peak voltage or current of either polarity is detected. Half wave is probably alright for a long term test, as the two polarities will eventually have equal peaks, but this isn't something I'm normally willing to leave to chance.
I dislike half wave rectifiers for a variety of reasons, so a simple full wave design has been incorporated. This is followed by a peak detector and hold circuit, which retains the highest peak voltage or current so it can be measured easily. The rectifier used has a particularly useful feature, in that it has variable gain. By using an appropriate input resistance, we can set the gain without having to include attenuators or additional amplifiers (see ESP Application Note 001 for precision rectifier descriptions). Consequently, we can have different ranges simply by using a different input resistance. After you've taken a reading, the 'hold' capacitor (C2) needs to be discharged, and this is done using Sw1 (Reset). It needs to be held for at least 10 seconds to ensure that the cap is fully discharged.
R7 is used to 'bootstrap' diode D5, and ensures that it has no voltage across it. With no voltage, there can be no current, so diode leakage is all but eliminated. The impedance at the junction of C2, D5, U2.5 and Sw1 is extremely high, and any PCB leakage will cause the voltage across C2 to droop with time, and you may miss the full amplitude of a peak reading. Ideally, this connection will be 'sky-hooked', meaning that the parts will be connected in mid-air, with no connection to a PCB or Veroboard track. The smallest leakage resistance will cause the peak-hold detector to lose voltage.
C2 should be a polypropylene (MKP) or a high voltage (250V minimum) film capacitor, because the insulation resistance of 63/100V MKT caps (polyester aka Mylar/ PET film) simply isn't good enough to hold the charge for more than a few 10s of seconds. The insulation resistance of a typical MKT polyester cap is around 5GΩ, vs. 500GΩ for a similarly rated polypropylene cap. For 99.9% of 'normal' audio applications this is of no concern at all, but for this application it's critical. I measured a 1kV WIMA 220nF MKS4 (polyester) cap (well, I tried), and its insulation resistance was well over 100GΩ (the WIMA website says greater than 30GΩ for voltage ratings of 250V and above). A 220nF, 275V X2 (polypropylene) cap measured 'only' 14.7GΩ.
The storage (hold) time is determined by the insulation resistance of C2, and any loading from the opamp (U2B) plus switch and diode leakage. The latter is dealt with fairly well by R7, but with a 220nF capacitor, the stored voltage will fall by about 4.5mV/s with a total leakage of 1GΩ. With a total leakage of 50GΩ, the voltage will fall by around 50µV/s, so maintaining a high impedance is essential. Expecting better than around 20mV/minute is probably unrealistic. Most tests will not take long enough for this to be a problem.
Ultimately, it appears that the limiting factor is actually the opamp - even with a claimed input impedance of 1TΩ for a TL072, its bias current (65-200pA at 25°C) is the overriding factor, and it is not low enough to maintain the voltage. Most other devices are similar or worse, but you could try the CA3240, which is rated for 20pA at 25°C, but the input current climbs alarmingly at higher temperatures. This is unlikely to be a problem in a domestic (or laboratory) setting, but it needs to be considered if high temperatures are anticipated. For example, if you use resistors instead of a current transformer (see below for details), the internal temperature may be far higher than desirable.
For good results at low levels (below 100mV), D1 and D2 should be matched for forward voltage. If this isn't done, low level rectification won't be 'true' full wave, but will show unequal peaks for positive and negative half cycles. It's up to the constructor to determine if this is a problem or not. For what it's worth, I didn't match the diodes on my prototype, and it is fine for 'typical' measurements. Also, beware of light. Under strong lighting, the reverse resistance of glass encapsulated diodes falls, which may cause the stored voltage to collapse a little faster than expected.
The datasheet says that the reverse leakage current for a 1N4148 is 25nA (at 20V), and I've measured the reverse resistance at between 1.13GΩ and 1.17GΩ in darkness, reducing to between 930MΩ and 1GΩ under bright lighting on my workbench (See Appendix). The difference isn't great, but without R7 the diode leakage would be a real problem. Reverse leakage was measured with 10V across the diode, and I obtained figures of around 8.5nA (dark) and 10.6nA (light). Of course, these values will be different with different diodes as verified by testing more than one.
Figure 1 - Full-Wave Rectifier & Peak Detector With Hold (Two Required)
The input of the rectifier circuit is simply fed via a resistance suitable for the range needed. Because some amplifiers are capable of very high output power (and hence voltage and current), two voltage and current ranges are provided. The ideal ranges would be 100V, 30V and 10V, being at roughly 10dB increments (10dB represents 10 times the power). However, unless a suitably calibrated analogue meter movement is used, these ranges are not user-friendly if a digital multimeter is used to measure the outputs. Consequently, two ranges are provided, namely 500V and 50V. The rectifier has very good performance down to the millivolt level, so a 5V range was not considered necessary. It can be added easily if necessary, simply by including a 10k input resistor (with switching) at the input of the rectifier stage.
Ideal current ranges are 30A, 10A and 3A, which would match the ideal voltage ranges when a 4 ohm load is used. However, the same caveats apply as for voltage, so the ranges provided are 50A (10mV/A) and 5A (100mV/A). With a shunt resistor of 0.1 ohm, 500mV is developed for a peak current of 5A, and 5V for 50A. The rectifier (etc.) will accommodate up to 50A (peak). 50A is a great deal of current, and in reality the maximum will be around 25A (100V with a 4 ohm load, 2.5kW !). This still results in a peak instantaneous 62.5W resistor dissipation and a loss of 2.5V (peak) of signal level, but it's not expected that this will be a common occurrence with 'normal' programme material. If you decide to use a resistive shunt, I suggest two 0.22 ohm 5W resistors and a 1 ohm, 1W resistor, all in parallel. The total resistance is 0.099Ω, provided you select the 0.22 ohm resistors for close to the exact value (±1%). An alternative that causes less power loss and dissipation would be to use a 0.01Ω current shunt resistor for the high current measurement, but they aren't easy to get and would require too much gain from the rectifier to get a usable signal (especially at lower current).
A current transformer is a better proposition, as there is no power dissipated and no signal loss. Common (small) current transformers have a ratio of 1,000:1, so with 5A (RMS) the output current will be 5mA (i.e. 1mA/A). When the standard 100 ohm 'burden' resistor is used, the output will be 707mV peak. I've tested a number of current transformers for other projects, and while it may seem unrealistic to expect full output at 20kHz, in practice this is usually achieved easily. The benefit is that no high power resistors are needed, and normal listening conditions are (almost) completely unaffected. Almost? There may be a loss of a small fraction of one dB due only to the additional wiring, but that can be ignored. If you wish to know more about current transformers and how they work, see Transformers, Part 2 (Section 17).
The main disadvantage of the current transformer is that it's comparatively large, but it's likely to be no larger than the resistors, and doesn't require any ventilation. If you use resistors, the case must be ventilated because they will get hot with high current loads. Personally, I'd go with the current transformer every time, because they are an elegant solution for AC measurements.
A typical 5A current transformer (such as the one recommended below) will normally be able to handle peaks of 50A (and usually more) without saturating, but if used with a subwoofer with frequencies below 30Hz it may cause some distortion. While this may cause slightly erroneous readings, the peaks (the only thing of interest) are largely unaffected. Fortunately, getting the gain required from the rectifier is not difficult. This is the approach I've taken, using a common (and cheap) current transformer available almost anywhere. You can get them from China (on ebay) or from 'real' distributors for less than AU$4.00 each, and there really isn't a need for anything better. I've tested one to well over 70A RMS (as much as you'll normally ever get) down to 50Hz easily. There was no evidence of distortion (caused by the core saturating), and the results are pretty good overall. I doubt that it will cause any problems in use.
The schematic for the unit is shown next, with the two rectifier blocks each using the Figure 1 circuit. To keep it simple, range switches are just mini-toggle types, and the reset buttons can be any momentary push buttons you choose. Because TL072 opamps aren't specified for operation with ±4.5V (as obtained with a single 9V battery), two batteries are needed, so the circuit runs from ±9V. This improves the ranges significantly, because the 50V range actually extends from just a few volts and although the 500V range is rather pointless, it's necessary to ensure that the meter reading corresponds to the actual voltage, without having to use maths (other than ×10 or ×100) to work out voltages and currents. There's not enough supply voltage to get to 10V (representing 100V), so the voltage output is limited to 0-5V for both rectifiers. This limitation could be lifted by using ±15V, but that would require a 'proper' power supply.
Figure 2 - Inputs, Range Selection & Battery Monitor (Current Transformer)
Figure 2 shows the wiring if a current transformer (CT) is used. The signal lead simply passes through the hole in the centre of the transformer - a complete turn is not required. U3A is essential to ensure that the 100 ohm burden resistor is not affected by the load imposed by the rectifier input resistors. There is provision for calibration (optional), but the main circuit relies on the accuracy of the current transformer and 100 ohm burden. This isn't really a limitation, because the current measurement is 'incidental' - it's interesting to know, but it does not need to be particularly accurate. Use the optional circuit shown to allow calibration if you think you need it.
To see details of a suitable current transformer, have a look at Project 139A. The transformer used is an AC-1005, a 1,000:1 CT rated for 5A. I've run tests on this particular transformer with current up to 70A RMS (See Note Below), with no sign of core saturation. As the frequency is reduced so too is the maximum current, but it's unlikely that the CT will be found wanting in any way. If you do want to calibrate the current range, use a variable resistor in place of the 100 ohm burden resistor. A 150 ohm resistor in parallel with a 1k trimpot will provide more adjustment than you'll ever need. The datasheet for the AC-1005 is available here if you want the details.
Note: You may wonder how I could test the current transformer to 70A, as that really is a great deal of current. The answer is quite easy, and involves nothing more than winding ten turns through the CT, and supplying 7A. The way CTs work is not intuitive, but adding turns provides greater sensitivity, and the transformer itself (and more importantly, its core) doesn't know the difference between 10 turns with 7A and 1 turn with 70A. This rather simplifies the test procedure, as developing a 'real' 70A test current isn't for the faint hearted. It would require a special transformer that few will have available.
The battery monitor is designed so that the LED (used to indicate that power is on) will extinguish if the total battery voltage falls below about 14V (7V per battery). You may need to experiment a little with R10 which feeds the 7.5V zener. The current as shown is only around 1mA, so the zener voltage is not very well defined. While a lower resistance could be used, that would draw more current from the batteries. The LED needs to be a high brightness type, because it has very limited current (around 1.3mA).
Figure 3 - Inputs, Range Selection & Battery Monitor (Shunt Resistor)
If you don't want to use the current transformer for any reason, Figure 3 shows the arrangement used to monitor current with shunt resistors. The two 0.22 ohm resistors in parallel have a total rating of 10W, which can be exceeded easily with a large power amplifier. However, with 'normal' programme material they can be expected to run at no more than around 5W even with heavily compressed music. This is based on material with a 10dB peak to average ratio, which is fairly normal with many modern recordings. It's probably worth noting that the resistors will take up as much (or perhaps more) space than a current transformer, and the case must be ventilated.
Again, accuracy isn't particularly wonderful, and the same comments apply here as were made for the current transformer. It's actually a bit worse because of the additional resistance, which means that a 100V peak (for example) will actually create a peak current of 24.3A into an 'ideal' 4 ohm load because of the added resistance. The total combined resistance is a little lower than 0.1 ohm (0.099 ohm), but the error is small and can be ignored.
The speaker leads are unplugged or disconnected from the speaker and connected to the inputs of the tester. The tester outputs then connect to the speaker. The inputs and outputs are interchangeable - the tester will work normally if they are swapped. Run the amp at your normal listening level, ideally with some varying programme material to get a representative signal level to the speaker. Because the hold circuit will retain the voltage for some time, you can play a number of tracks, and ideally the loudest will be played last.
When you are done playing music (or other material), the peak voltage and current are read from the tester using a multimeter. If you measure (say) 3V DC on the 50V range, that means that the voltage peaked at 30V (2V on the 500V range means 200V peak, assuming your amplifier can deliver that much). On the 500V range, you will measure a peak voltage of 300mV under the same conditions, indicating a peak voltage of 30V (you should use the 50V range, which will show a voltage of 3V). The 500V range is only needed if the output exceeds 5V DC.
Current is checked the same way. If you measure 2V at the current output on the 50A range, that means the peak current was 20A. Again, you'll likely find that the actual current is somewhat less. The ranges shown on the schematics assume a maximum of 5V output from the hold circuit, and with a pair of 9V batteries you should be able to measure up to (around) 6.5V output. The table below is based on a maximum output of 5V, which will be available even with low battery voltage.
|Range||Maximum Recommended||Conversion||Max Reading|
|50 V||50 V peak||100 mV/ V||5 V|
|500 V||200 V peak||10 mV/ V||2 V|
|5 A||5 A peak||1 V/ A||5 V|
|50 A||50 A peak||100 mV/ A||5 V|
The most usable ranges for most systems will be 50V and 50A. This will work for amplifiers with supply voltages up to ±50V and load impedances down to 4 ohms. The 500V range won't be needed unless your amplifier can deliver more than 50V peak, indicating much higher than 'normal' supply voltages for the amplifier. It's not expected that any amp will be able to reach a peak output voltage of 500V (30kW into 4 ohms !), but intermediate ranges would mean that the output voltage would have to be calculated, rather than determined using some basic mental arithmetic.
The usefulness of the 5A range is possibly dubious, and you can leave it out if you don't think it will be needed. Even on the 50A range, you'll still be able to measure down to 1A or so (an output voltage of 100mV DC), so you might decide that it's not worth the extra switching. Even the opamp buffer can be omitted if the 5A range is omitted, as the error introduced by the 10k rectifier input load is negligible.
Figure 4 - Waveform Of AC Capture And Hold
For the above, the circuit was configured for unity gain (10k input resistor), and the first drop to zero is when the reset switch was pressed. The voltage was then increased in 1V RMS steps from 1V to 5V. The hold stability is demonstrated by the fact that the peak amplitude remains steady for the complete trace, a period of 30 seconds. There is very little droop even after one minute, but the voltage does fall, as explained above. There is no doubt that the circuit works as described, based on the scope trace. The loss of voltage is more noticeable with a digital multimeter, because you can see the voltage falling (albeit slowly).
you can see that with 1V (RMS) applied, the scope shows 1.4V, and 2.8V for 2V (and so on). The highest voltage measured is 7.07V, but that's not easy to see properly on the scope trace. The trace is shown as a demonstration only, and I didn't calibrate the system before the waveform was taken.
This is a project that lets you determine just how much of the available voltage swing (not actual power) you are using from your amplifier(s). You may well discover that the voltage you can achieve is insufficient, and this is shown by the voltage monitor providing a reading that's close to the amplifier supply voltage. Equally, you may find that you never use the available power, and that you can use a smaller amplifier if you wish. Without this information, you never really know if your system approaches (or reaches) clipping, as most systems don't include a clipping indicator.
The ability to monitor the peak current is also handy, although this can be left out if you don't think you need it. Most speaker systems are reasonably well behaved, and it's unusual for them to demand more current than your amp can provide comfortably. However, if your speakers are 'home-brew' or ones known to be a 'difficult' load, you might be surprised at the peak current that's demanded during operation. Speakers aren't a simple resistive load, and they have considerable reactance (both capacitive and inductive). This can lead to much higher (or perhaps lower) current than you anticipated.
Your speakers may have a quoted impedance of (say) 8 ohms, but this is a nominal figure. The actual impedance can be as low as 5 ohms or as high as 50 ohms (the latter at the woofer's resonant frequency, but it varies with the speaker). There are some speakers that fall to much lower impedances at certain frequencies, often determined by the design of the crossover network. By monitoring the current, you can see just how much peak current your amp needs to provide. This also makes the idea of amplifiers that can provide ten times as much (or more) current than the speaker can ever draw look a bit silly.
One thing that this project is ideal for is to characterise a speaker system, so you know just how much power is needed for the required SPL. Naturally, that must also consider the power ratings of the installed loudspeaker drivers. While most loudspeakers can tolerate a little more power than their continuous rating (often referred to as 'programme power'), if it's overdone, the driver(s) will have a short and miserable life. This project deliberately shows the peak voltage and/ or current, since that's what determines the amp rating without clipping. Despite what you might imagine, an RMS voltage (or current) measurement of amplifier output with programme material is not at all useful, as that fails to show if transient clipping occurs. The idea of this project is to catch peaks (transients), not the average or RMS level.
There's no real reason that you need two detectors, because voltage and current peaks can be measured separately. Peak current isn't something that you really need to measure, but of course it's interesting to see just how much current your speakers draw. Remember that all values shown are peak. Dividing by 1.414 to get RMS will only work if the input is a pure sinewave.
If you wish to measure the actual power being delivered to your speakers, then use the Project 189 'true' power meter, which uses an analogue multiplier to compute the power delivered, based on the instantaneous peak voltage and current delivered. It also has an output for an oscilloscope so you can also read the instantaneous power. While it's not especially cheap to build, it does work very well.
You may well wonder how it's possible to measure a resistance of 1GΩ or more, as I did for the 1N4148 diodes and selected capacitors. Obviously, no multimeter can measure that much resistance, but with some trickery it can ! The meter is used on its voltage range, and connected in series with the reverse biased diode. Then a known voltage is applied (say 10V), and the meter will show a reading of perhaps 100mV. Almost all digital multimeters have an input impedance of around 10MΩ (two of mine measure 11MΩ) on the DC voltage range, so a voltage of 100mV across 11MΩ means the current is 9.09nA. The remainder of the voltage is across the diode, which must also be passing 9.09nA. If the applied voltage is 10V, that works out to a resistance of 1GΩ (10V / 9.09nA = 1.1GΩ).
Extreme precision is not necessary (one could subtract the 100mV for example), but the end result is 'good enough' for most measurements. This is particularly true since the insulation resistance of a PCB between adjacent tracks may not be much greater than the reverse leakage of the diode, and even the smallest amount of moisture can affect the reading dramatically. I measured between tracks of a 50mm length of Veroboard, and when dry I obtained 6.2mV (almost 18GΩ), but just breathing on it dropped the resistance to well below 1GΩ (albeit briefly).
This is a very useful technique if you ever need to measure particularly high resistances, and it doesn't appear to be widely known. There are (of course) specialised meters for measuring extraordinarily high resistances, but the humble digital multimeter does a perfectly acceptable job with some care. Quite obviously, the DUT (device under test) must be suspended away from anything that may be ever-so-slightly conductive, and the meter leads also have to be very well insulated. the smallest amount of leakage can create a very large error.
This technique is discussed in more detail in the ESP Application Note AN-016.
Now you also know why I recommend that the junction of C2, U2 pin 5, D5 and Sw1 (Reset) should be suspended in mid air ('skyhooked'), and not connected to the Veroboard tracks. Even a fibreglass PCB may be suspect unless the appropriate points are protected by a guard track or similar (which is not possible with Veroboard). If you've never heard of a 'guard track', see Designing With Opamps, High Impedance Amplifiers. The guard track (or ring) effectively 'bootstraps' the enclosed circuit in the same way that R7 prevents leakage in D5 (Figure 1).
1 Audio Express, Build a Voltage and Current Peak Detector - George Ntanavaras
2 P. Horowitz and W. Hill, The Art of Electronics, 2nd edition, Cambridge University Press, 1989.
3 Applications of Operational Amplifiers, Third Generation Techniques - Jerald Graeme, Burr-Brown, 1973, pp. 123-124
4 Microelectronics: Digital and Analog Circuits and Systems (International Student Edition), Author: Jacob Millman, Publisher: McGraw Hill, 1979 (Chapter 16.8, Fig. 16-27)
5 ESP AN-001 - Precision Rectifiers
6 ESP AN-014 - Peak Detection Circuits
|Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is © 2019. Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws. The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference. Commercial use is prohibited without express written authorisation from Rod Elliott.|