LLow Frequency AC Repsonse of Digital Multimeters

[David Hess]

I just put together a ported subwoofer and ran across a slight problem with the tuning.  My best multimeter only has a flat AC response down to 20 Hz which is marginal for making the measurements to properly tune the subwoofer enclosure.  So my question is, what modern or available AC voltmeter has the lowest AC frequency response?

A DSO using DC coupling could make the measurements I need, albeit with lower resolution.  Another option I am considering is the DATS V3 which can be had for about $120.

[joeqsmith]

No idea what your requirements are but most of my DMMs have a DC mode.  Some of them have cutoff filters above 20Hz. 

Maybe you could just rectify and filter it. 

[bdunham7]

Wouldn't a TRMS AC+DC meter essentially be flat to DC??  Hold on, I'll try one.

EDIT:  TRMS AC+DC doesn't work so well below 10 Hz because I can't slow down the sample rate and thus the reading fluctuates, as you'd expect.  So, a quick low frequency test of a few meters set to AC using a 2.000V signal:

Fluke 8846A w/ filter set to 3 Hz reads pretty much spot on down to 3.0 Hz, and then drops off starting at 2.9Hz.  I guess when they say 3Hz they really mean it.  This is supposedly a 3 or 4 pole filter, I don't remember which.  Accuracy is specified to 3 Hz.  3.0000 Hz, that is.

Fluke 289 stays accurate down to 10 Hz and then fluctuates due to sample rate.  Accuracy specs go to 20Hz.

Fluke 189 stays accurate to 5 Hz, then fluctuates.  It isn't 'better' than the 289, but has a slower sample rate.  Accuracy specs go to 45Hz.

Simpson 270, reads nicely and probably more than good enough for subwoofer tuning--the error is a LOT less than 1dB, probably more like 0.1dB.  Below 10Hz the needle starts to shiver, but it is accurate and reasonably readable down to 2 Hz.  Suprisingly, the Simpson is specified for AC from 20Hz to 100kHz.  Apparently the needle shivering is the only thing limiting the lower end.

So the winner is the Simpson.   

[Tomorokoshi]

Measurements using an HP 3403C True RMS Voltmeter.

Using HP 3310A Function Generator as a source. Set up sine wave amplitude for 1.000 V RMS at 1000 Hz. Passthrough 50 ohm termination at voltmeter.

Code: [Select]

Frequency  Voltage  Jitter
1 MHz      1.006    +   0.001
100 kHz    1.003    +/- 0
10 kHz     1.002    -   0.001
1000.0     1.000    +/- 0
 100.0     1.000    +/- 0
  10.0     1.000    +/- 0
    3.0    0.999    +/- 0
   1.0     0.996    +/- 0.001
   0.3     0.953 to 1.005
   0.1     0.659 to 1.217 on a 5 second cycle
   0.01    0.033 to 1.408 on a 50 second cycle

So it seems good to around 1 Hz.

[bdunham7]

Well, he said 'modern'.  Mine are all current production.....

I have an old HP 403B.  It's not RMS, but I wonder how it would do.

[floobydust]

I find the peak is very sharp and narrow bandwidth (depends on system Q). I use a Lissajous with scope (sampling V, I) to see the resonance.
The tiniest of air leaks will cause a huge shift at low levels, as well as the need to break in a new driver and loosen up the suspension a little.
I think it would be difficult but possible to find it even with a DMM rolling off. An old VU meter can even be used.

At higher power, I found the room interacts with the loudspeaker and sometimes I've locked onto a room resonance by accident.

[David Hess]

Maybe you could just rectify and filter it.

That would definitely work, or I could extend the time constant of the rectifier in one of my bench meters.  I have several which are fully documented so this would be relatively easy.

Wouldn't a TRMS AC+DC meter essentially be flat to DC??  Hold on, I'll try one.

EDIT:  TRMS AC+DC doesn't work so well below 10 Hz because I can't slow down the sample rate and thus the reading fluctuates, as you'd expect.  So, a quick low frequency test of a few meters set to AC using a 2.000V signal:

That was one of the first things I tried but I got the same result which in retrospect makes sense.  I also got the same results in average and RMS modes.

Measurements using an HP 3403C True RMS Voltmeter.

A thermal RMS meter is an interesting option but I do not have one.

[Kleinstein]

The DMMs that use digital RMS have a good chance to work also down to rather low frequencies. The HP 34410 starts from 3 Hz in the specs , the newer 3446x should be similar.

The ADC on most DSOs is only 8 bits, but they used quite a lot of points. With a reasonable well chosen range the resoltion is not that bad. It is more that the calibration and the rage steps are often not that accurate.

[radiolistener]

Update: posted to wrong topic, Brymen BM867S and UNI-T UT210E frequency response results moved here:
https://www.eevblog.com/forum/testgear/frequency-response-of-dmms/msg3572871/#msg3572871

[bdunham7]

The DMMs that use digital RMS have a good chance to work also down to rather low frequencies. The HP 34410 starts from 3 Hz in the specs , the newer 3446x should be similar.

The 34401A also works to 3 Hz or so if you go into the settings and change the filter.  I don't really see why the method of RMS conversion, whether it be thermal, a TRMS chip or digital, actually matters much--they all should convert down to DC.  The limiting factor seems to be the integration time or measurement period, which if not long enough, will cause a fluctuating reading.  That fluctuating reading isn't 'wrong', it just reflects the actual RMS value during the measurement period.

[Kleinstein]

The analog RMS convertes need the filtering to give the right correction. With a very low frequency leading to large fluctuations the errors go up quite a bit.  As the filter slows down the response quite a bit, it is not so practical to have a very slow filter, especially if only 1 setting is available. The ouput voltage is not only filtered for the direct reading, but also for the RMS calculation, that  is done as  the square of the input voltage divided by the filtered result. So the filtering is not just for a more stable result, but required part of the RMS conversion.

The digital RMS can use a FIR fitler and this gives less problems with settling. Also the fitler can be automatic adjusted to the frequency. So it is possible to use a better filtering with low error even for very low frequencies. Ideally one needs 1 period for a new result.

The thermal RMS converters are by nature relatively slow and if the filterung is insufficient to give a stable result, the result is still correct on average.

[David Hess]

Thanks for helping me think this problem through.

No idea what your requirements are but most of my DMMs have a DC mode.  Some of them have cutoff filters above 20Hz. 

Maybe you could just rectify and filter it.

I think I am going to first try modifying one of my old Tektronix bench meters for lower frequency operation.  They are fully documented and easy to work on so this should not be too difficult.

The AC rectifier is an interesting circuit.  How do you make a precision rectifier with zero offset voltage?  You AC couple the amplifier.

[David Hess]

The operational amplifier is the NE531 which is 100 V/mV, 1 MHz GBP, and 35 V/us when compensated for unity gain, and perhaps an estimated 20 MHz GBP and 700 V/us as shown.  (1) I do not want to order just some low leakage capacitors so if I replaced the operational amplifier, what do you think would be a good choice?

Most video amplifiers have low open loop gain which would compromise accuracy but maybe the AD829 (100 mV/V, 120 MHz, 230 V/us), LT1226 (150 mV/V, 1000 MHz, 400 V/us), or LT1222 (200 V/mV, 500 MHz, 200 V/us) would work well?  The later two are decompensated.  The LT1222 is especially interesting because it supports clamping but the circuit would need to be changed to take advantage of that.

LT1022   400 V/mV   8.5 MHz   23 V/us   $6.71
LT1122   400 V/mV   8.5 MHz   26 V/us   $7.34
LT1468   9000 V/mV   90 MHz   22 V/us   $7.06

OPA604   100 V/mV   20 MHz   25 V/us   $3.50
TLE2071   200 V/mV   10 MHz   45 V/us   $2.30

LT1222   200 V/mV   500 MHz   200 V/us   $10.20

I am limiting myself to 36 volt and higher operational amplifiers in 8 pin DIP packages unless there is something outstanding available only in an SO8 package.

(1) I do not believe that an NE531 gets 700 V/us when uncompensated but that is what the math says and I am too lazy to test one, but it *is* fast.

[bdunham7]

I'm trying to follow along and have a few questions.

How are you using the AC measurement to tune your subwoofer, exactly?  What do you measure?

Are you thinking that the frequency response limitations of the Tek plug-ins are due to the output filtering or the AC coupling of the op-amp (U195)?  IOW, do they bounce around or just roll off without fluctuation?

Is TRMS not needed in this case?  IOW, is the signal you are measuring not subject to any sort of distortion or change other than amplitude?

[David Hess]

How are you using the AC measurement to tune your subwoofer, exactly?  What do you measure?

A rough measurement of the current using a current shunt will reveal the impedance which depends on the loading of the driver.  With a ported enclosure, two peaks in the impedance are observed and their magnitudes can be used to tune the enclosure without making any absolute measurements.

I am doing it this way because nothing extra is required.  An impedance analyzer would make the job much easier, and I may still go that route if I do not get the results that I want, but so far the results have been good.  Previously I did this with woofers but the resonate frequencies were higher so there was no problem making the measurements.

Are you thinking that the frequency response limitations of the Tek plug-ins are due to the output filtering or the AC coupling of the op-amp (U195)?  IOW, do they bounce around or just roll off without fluctuation?

First it rolls off, and then it starts fluctuating.

As usual, what started as a loudspeaker project has turned into a refurbish the multimeters project, but I find that acceptable.  If I have to order parts for one, I might as well take care of the others.  I investigated a broken DM502 tonight and found that one of the bridge rectifiers in the power supply was completely open.  How does that even happen?

Is TRMS not needed in this case?  IOW, is the signal you are measuring not subject to any sort of distortion or change other than amplitude?

Everything is linear so there are no added distortions unless something is broken.  The only thing I am measuring are sine waves.

[Kleinstein]

Changing the caps in the circuit is tricky - at high speed the parasitic capacitance may matter. So the caps at the OP that effect the lower frequency limit should not be physical larger.
The input to the OP is relatively high impedance, so the input capacitance may also be a factor.
I don't think the open loop gain would be so critical for accuracy, as the AC measuremen is not very accurate anyway.

For the measurement in the low audio range, I would consider using a soundcard and build an input amplifier / protection for this instead of modifying the meter with questionable results.

[David Hess]

Changing the caps in the circuit is tricky - at high speed the parasitic capacitance may matter. So the caps at the OP that effect the lower frequency limit should not be physical larger.

The input to the OP is relatively high impedance, so the input capacitance may also be a factor.

Those would definitely be a concern with the higher frequency decompensated amplifiers simply because of the relatively high impedances in the circuit, so a redesign would be required anyway.  I will not be going that route.

Is there a modern replacement for the NE531?  Usually I see old designs like this use the LM318 but its only virtue is availability.  Did the LM318 functionally replace the NE531?

I don't think the open loop gain would be so critical for accuracy, as the AC measuremen is not very accurate anyway.

Most video amplifiers only have an open loop gain of a few thousand which would definitely decrease accuracy.  The ones I listed are exceptions.

For the measurement in the low audio range, I would consider using a soundcard and build an input amplifier / protection for this instead of modifying the meter with questionable results.

I will have to look into that.  Thanks for the suggestion.

[mawyatt]

David,

I get somewhat stable readings with the KS34465A down to ~2Hz using the AC filter set to 3Hz, and Smoothing Filter slow(100 rds). For example, using a HV amp (I'm working on) with 2Hz input the DMM reads 50V AC, 50V Average with SD 0.001V, below 2Hz the readings begin to fluctuate, at 1.5Hz SD rises to ~0.07V.

Best,

[David Hess]

For the measurement in the low audio range, I would consider using a soundcard and build an input amplifier / protection for this instead of modifying the meter with questionable results.

I will have to look into that.  Thanks for the suggestion.

For those who might be looking for software to support impedance measurement using a sound card, REW (Room Acoustics Software) is available.  That leaves finding a suitable sound card interface to sacrifice.

I get somewhat stable readings with the KS34465A down to ~2Hz using the AC filter set to 3Hz, and Smoothing Filter slow(100 rds). For example, using a HV amp (I'm working on) with 2Hz input the DMM reads 50V AC, 50V Average with SD 0.001V, below 2Hz the readings begin to fluctuate, at 1.5Hz SD rises to ~0.07V.

A new high resolution bench meter costs more than I want to pay but it is interesting that they can have provisions to extend their AC response that low.

[David Hess]

I don't really see why the method of RMS conversion, whether it be thermal, a TRMS chip or digital, actually matters much--they all should convert down to DC.

They almost always remove any DC offset with AC coupling, and then have a separate AC+DC measurement mode if desired although this can be done with a second average DC measurement and a tiny bit of math.

The limiting factor seems to be the integration time or measurement period, which if not long enough, will cause a fluctuating reading.  That fluctuating reading isn't 'wrong', it just reflects the actual RMS value during the measurement period.

That comes up when making low frequency noise measurements.  The integration time of the measurement sets a lower limit on the frequencies observed.  So for instance if you want to measure down to 0.1 Hz, then you *must* integrate for 10 seconds.

[mawyatt]

A new high resolution bench meter costs more than I want to pay but it is interesting that they can have provisions to extend their AC response that low.

Think of this as an "excuse" to get a new meter 

Best,

[floobydust]

I went through my notes and for port tuning I found ACV too sloppy, I'd also tried SPL (at the port and woofer) but in the end I had to look at phase-angle using a scope with a 1R sense resistor.

[Caliaxy]

The limiting factor seems to be the integration time or measurement period, which if not long enough, will cause a fluctuating reading.  That fluctuating reading isn't 'wrong', it just reflects the actual RMS value during the measurement period.

That comes up when making low frequency noise measurements.  The integration time of the measurement sets a lower limit on the frequencies observed.  So for instance if you want to measure down to 0.1 Hz, then you *must* integrate for 10 seconds.

A way around that (for low frequency sine waves, when you can't control the acquisition rate and/or the integration time of your instrument) would be to do exactly the opposite: use the fast "peak" (or min-max) mode, then check the min/max values. Of course, you'll get the peak-to-peak value of your signal, not the RMS (if you wait long enough...); if that's a sine wave, you can calculate the RMS (if you really want to). I tried it on a few of my meters and it seem to work quite well on some of them (e.g. on the cheap-ish UnitT UT61E). The downside is that once it finds a peak, you won't get any update anymore so you have to reset the measurement by getting out of the peak mode and back in, which on some meters requires a at least a looong press of a button (among others). It works great on Dave's 121GW, which has a "1ms Peak" button. You can go way under 1 Hz this way. Of course, it doesn't work well if the signal is noisy. A DSO would be by far a better alternative.

Just a thought, for when you have to use what you have at hand...

[David Hess]

A new high resolution bench meter costs more than I want to pay but it is interesting that they can have provisions to extend their AC response that low.

Think of this as an "excuse" to get a new meter 

That was what I was hoping for since my old Beckman RMS225 needs to be replaced, but I am going to work my way up the cost ladder which starts at modifying one of my existing meters for a couple dollars in parts, and I need to order parts to fix one of my DM502s anyway.

One of my DM502s mysteriously failed last year.  I assumed it was the electrolytic capacitors but investigation shows that the bridge rectifier for the +/-12 volt supply went open on the negative side.  Removing it caused it to open on the positive side also.  I have to hope that the RC4194 dual polarity tracking regulator was not damaged because they are in short supply.

I went through my notes and for port tuning I found ACV too sloppy, I'd also tried SPL (at the port and woofer) but in the end I had to look at phase-angle using a scope with a 1R sense resistor.

I admit it is not optimal but it has worked out well enough in the past, and I did get results this time.

For those who might be looking for software to support impedance measurement using a sound card, REW (Room Acoustics Software) is available.  That leaves finding a suitable sound card interface to sacrifice.

For those who might consider this solution, suitable sound interfaces are rare because a stereo input is required.  I am leery of connecting anything directly to an expensive computer through the line-out and line-in connections.  The BEHRINGER U-Phoria UMC202HD appears to be the best of the inexpensive options but I am not sure about its low frequency capabilities.

The limiting factor seems to be the integration time or measurement period, which if not long enough, will cause a fluctuating reading.  That fluctuating reading isn't 'wrong', it just reflects the actual RMS value during the measurement period.

That comes up when making low frequency noise measurements.  The integration time of the measurement sets a lower limit on the frequencies observed.  So for instance if you want to measure down to 0.1 Hz, then you *must* integrate for 10 seconds.

A way around that (for low frequency sine waves, when you can't control the acquisition rate and/or the integration time of your instrument) would be to do exactly the opposite: use the fast "peak" (or min-max) mode, then check the min/max values. Of course, you'll get the peak-to-peak value of your signal, not the RMS (if you wait long enough...); if that's a sine wave, you can calculate the RMS (if you really want to). I tried it on a few of my meters and it seem to work quite well on some of them (e.g. on the cheap-ish UnitT UT61E). The downside is that once it finds a peak, you won't get any update anymore so you have to reset the measurement by getting out of the peak mode and back in, which on some meters requires a at least a looong press of a button (among others). It works great on Dave's 121GW, which has a "1ms Peak" button. You can go way under 1 Hz this way. Of course, it doesn't work well if the signal is noisy. A DSO would be by far a better alternative.

Just a thought, for when you have to use what you have at hand...

I ran a test and that works great up to 120 Hz, although it would of course be slow to use and it comes at the expense of resolution.

[David Hess]

How are you using the AC measurement to tune your subwoofer, exactly?  What do you measure?

I went through my notes and for port tuning I found ACV too sloppy, I'd also tried SPL (at the port and woofer) but in the end I had to look at phase-angle using a scope with a 1R sense resistor.

I ended up getting good results by measuring the voltage across the speaker when driven directly by the 50 ohm source impedance of my function generator.  Tuning of the port length was done by minimizing the voltage, and hence cone movement, at resonance which is between the two peaks in impedance from the speaker and enclosure.  Resonance was found by finding the lowest impedance point (lowest voltage) which shifted slightly as the port length was adjusted.  Conveniently resonance was at about 41 Hz so within the frequency range of my voltmeter so I did not need to modify one of my voltmeters at all.  A phase measurement would be more sensitive but having more resolution on a voltmeter makes up for that.

Besides yielding good passband flatness and transient response when tuned, minimizing the cone movement reduces distortion from large signal non-linearity, which I consider to be the most important.  The non-linearity in the driver produces both harmonic and intermodulation distortion, and the later is particularly objectionable and a problem in low frequency drivers.

The dowel shown below allowed adjustment of the port length while testing.