Is there a modern equivalent to old AC Voltmeters that measure 2MHZ+?

[bdunham7]

Old style high frequency AC voltmeters used either thermal or sampling methods to achieve high bandwidth.  I do not know of any modern instruments suitable to replace them,

A number of the current higher-end DMMs use sampling methods for AC, including the not-so-new 3458A.  The best of them can measure 10VAC to 1MHz with less than 1% error and I've always assumed that the BW limitations were due to input scaling and protection not the sampling system itself. 

For higher BW measurements, I think the DSO can probably step in in most cases and is mostly limited by the same issue--flatness of the input circuitry.  After all, the DSO is just a really fast sampling voltmeter and if it is taking 2 billion samples per second and the samples are 200ps wide, I don't see how it matters that the input voltage is varying at 10MHz instead of 10Hz.  In my own experimentation doing this the primary sources of significant error seem to be gating issues and probes where they are used.  The same applies in reverse to an AWG--unless it is designed oddly or has a defect, flatness is sort of designed in as long as you don't approach the BW limit of the output circuitry.

[David Hess]

Old style high frequency AC voltmeters used either thermal or sampling methods to achieve high bandwidth.  I do not know of any modern instruments suitable to replace them,

A number of the current higher-end DMMs use sampling methods for AC, including the not-so-new 3458A.  The best of them can measure 10VAC to 1MHz with less than 1% error and I've always assumed that the BW limitations were due to input scaling and protection not the sampling system itself. 

For higher BW measurements, I think the DSO can probably step in in most cases and is mostly limited by the same issue--flatness of the input circuitry.  After all, the DSO is just a really fast sampling voltmeter and if it is taking 2 billion samples per second and the samples are 200ps wide, I don't see how it matters that the input voltage is varying at 10MHz instead of 10Hz.  In my own experimentation doing this the primary sources of significant error seem to be gating issues and probes where they are used.  The same applies in reverse to an AWG--unless it is designed oddly or has a defect, flatness is sort of designed in as long as you don't approach the BW limit of the output circuitry.

"Sampling" in this case refers to a specific method of operation which modern oscilloscopes and the 3458A do not qualify for.  It includes "obsolete" instruments like the HP 3406A and Racal Dana 9301 which use a sampling front end where bandwidth is only determined by the sampling strobe width and not the RC time constant of the input circuits.  This also applies to true sampling oscilloscopes.  These instruments have a very predictable non-linear frequency response in contrast to the linear single or multiple pole response of a normal instrument.

So for instance a Racal Dana 9301 has a flat bandwidth to 500 MHz, where "flat" means that no amplitude correction is required because it is within the 1% measurement error of the instrument.  The  HP 3406A seems to be a little less than this, but the manual is difficult to read on this point.  A digital storage oscilloscope would require a bandwidth of roughly 5 GHz to achieve this with single pole response and nobody has proposed using such an instrument.

This is what makes the performance of instruments like the Racal Dana 9301 and HP 3406A so difficult to duplicate.  Another point is that these instruments do *not* require a leveled source for high frequency calibration.  Calibration can be done with an unleveled source by measuring the first null in their non-linear frequency response.

[TimFox]

In the original post, the calibration procedure from -hp- could use an old-fashioned 400E voltmeter, which is an average-responding unit with the meter movement within a rectifier in the feedback loop.
Done carefully, this works up to a few MHz, which is required to calibrate the 4204A generator.
See page 1-1 of the 400E manual for specifications:  https://pearl-hifi.com/06_Lit_Archive/15_Mfrs_Publications/20_HP_Agilent/HP_400E_EL/HP_400E_&_EL_AC_Voltmeter.pdf

[joeqsmith]

+/-3dB on a scope would be bad.  Mine are all 8-bits so there are other problems.   It is possible to characterize the scopes ADC and correct for non-linearity errors and over sample for these lower frequencies and get some pretty good results.  Some of these new high resolution scopes are really impressive though.   

Hi,
I would like to ask you if I can measure high frequency voltage with simple or ordinary multimetre, .
 maybe a picture or vedio to palnation?, let's say 100mv and 75Mhz.
Thank you

https://www.eevblog.com/forum/rf-microwave/measuring-frequency/

[pdenisowski]

If there are 3 dB variations within working bandwidth, then how is working bandwidth defined? Is there another definition of working bandwidth or is this just talking about what's written on the device as working bandwidth?

This question should be addressed to Siglent and other oscilloscope manufacturers.

As someone who works for an oscilloscope manufacturer (and produced a video explaining oscilloscope bandwidth):  scope bandwidth is generally defined as the frequency at which the measured amplitude of a sinusoidal signal is 3 dB down from the actual signal amplitude.  (see video below at 1:14)

While it is true that these variations are frequently more pronounced near the oscilloscope's bandwidth limit, this cannot be taken as a general rule. Many oscilloscopes also exhibit ±3 dB deviations in the middle of their operational bandwidth. And even more often at beginning of operational bandwidth (at low frequencies below 500 Hz).

Wow, you need some better oscilloscopes because I'd call that behavior "broken".  There's no logical reason that an oscilloscope would vary that much from DC to low frequencies.  I don't recall seeing a scope with 3db variations mid-band, but I have seen probes that are wonky like that, which is why I recommended a direct connection.

All scopes have some amplitude variation or ripple in their "passband" (see video at 8:52), but this should be on the order of a fraction of a dB.  3dB variation in the passband would be completely unacceptable (even to me as a hobbyist) at any price point.

[tatel]

Restoring an old HP 4204A Oscillator and I am tearing my hair out with the "calibration"  (more like troubleshooting).  The main 2 measurements required to both troubleshoot & calibrate are the AC voltage (in which the frequency can hit 1mhz & around 10v) and obviously the frequency itself  The manual recommends a 3400a or 400E AC Voltmeter.  There are plenty of them out there for $50-$100 but it just seems silly to buy something like that for this limited use.  There are some heathkits but they all seem in various states of disrepair & aren't much cheaper.  Google hasn't turned up much in this regard.  Thank you!

Maybe a Fluke 8920-21-22?

However you would be in the same case that with the devices suggested to you in 4024A manual

[David Hess]

While it is true that these variations are frequently more pronounced near the oscilloscope's bandwidth limit, this cannot be taken as a general rule. Many oscilloscopes also exhibit ±3 dB deviations in the middle of their operational bandwidth. And even more often at beginning of operational bandwidth (at low frequencies below 500 Hz).

Wow, you need some better oscilloscopes because I'd call that behavior "broken".  There's no logical reason that an oscilloscope would vary that much from DC to low frequencies.  I don't recall seeing a scope with 3db variations mid-band, but I have seen probes that are wonky like that, which is why I recommended a direct connection.

There is a reason that has to do with modern DSO design.

Starting in the 1980s, analog oscilloscopes and DSOs changed to using some variation of a two path high impedance buffer, doing away with the temperature compensated dual JFET design.  One consequence of this is that the cutoff frequencies and gain of the two paths must be matched to prevent a mid-band discontinuity in the frequency response.  Early implementation had a calibration step to adjust this, but modern DSOs usually have a fixed calibration which cannot be adjusted.  In practice the discontinuity is always present to some degree, but I have seen examples of modern inexpensive DSOs where it is particularly bad.

The second attachment shows the early two path design from a Tektronix 2232 where the amplitude and frequency calibration adjustments are present.

[bte]

If there are 3 dB variations within working bandwidth, then how is working bandwidth defined? Is there another definition of working bandwidth or is this just talking about what's written on the device as working bandwidth?

This question should be addressed to Siglent and other oscilloscope manufacturers.

As someone who works for an oscilloscope manufacturer (and produced a video explaining oscilloscope bandwidth):  scope bandwidth is generally defined as the frequency at which the measured amplitude of a sinusoidal signal is 3 dB down from the actual signal amplitude.  (see video below at 1:14)

I know that already. My question was rhetorical to get a clear explanation for the claim of "Many oscilloscopes also exhibit ±3 dB deviations in the middle of their operational bandwidth" or maybe a few examples of such oscilloscopes. With the usual definition of scope bandwidth ("the frequency at which the measured amplitude of a sinusoidal signal is 3 dB down from the actual signal amplitude"), it is not possible to have 3 dB variations in the operational bandwidth to start with. The first 3 dB deviation would define the bandwidth. Maybe instead of the usual definition, a revised definition could be something along the lines of "the highest frequency at which the measured amplitude of a sinusoidal signal is 3 dB down from the actual signal amplitude". That sort of definition would make 3 dB deviations within the operational bandwidth possible.

And yes, I have seen the post from David Hess. It supports such a revised definition of working bandwidth and nicely provides the reasoning for it. BTW, if anyone is interested: the book quote in his post is from "The Art and Science of Analog Circuit Design", edited by Jim Williams.

[bdunham7]

There is a reason that has to do with modern DSO design.

Starting in the 1980s, analog oscilloscopes and DSOs changed to using some variation of a two path high impedance buffer, doing away with the temperature compensated dual JFET design.  One consequence of this is that the cutoff frequencies and gain of the two paths must be matched to prevent a mid-band discontinuity in the frequency response.  Early implementation had a calibration step to adjust this, but modern DSOs usually have a fixed calibration which cannot be adjusted.  In practice the discontinuity is always present to some degree, but I have seen examples of modern inexpensive DSOs where it is particularly bad.

I probably shouldn't have said DC, but even there the issue with the DC/LF amp section on modern scopes may not lead to as serious a problem measuring RMS voltage as it might first appear because the step response distortion can be caused by phase shift just as well as amplititude variation.  I gave it a try using an AWG set to 5.000V and my 8506A in parallel with two Siglent DSOs, an SDS1104X-E (with the problematic LF distortion) and an SDS2354X+.  I tested at 10, 100, 1k, 10k, 100k, 200k, 500k an 1M Hz.   Results:

FREQ     10         100         1000         10k         100k         200k         500k         1M

8506A   4.9939   4.9958    4.9955     4.9982      4.9957       4.9866        4.9599       4.9089

1104X-E   5.02           4.98           5.03           5.02           5.04           5.03           5.03           4.97   

2354X+   5.015   5.008   5.027   5.041   5.050   5.051   5.035   4.967

I think the input capacitance of the 8506A was loading the AWG down at 1MHz, removing it brought the values on the scopes back up.  As you can see, either would be adequate for the OP's purpose--and more generally, they exceed any other method I have of measuring AC voltage once you get over 500kHz.  Unfortunately I can't push much further because my AWG is not quite as good when it comes to flatness.  1% is 0.09dB and that's a lot to ask for from sig-gens or spectrum analyzers.  I also did sweep tests on both scopes 0.1 to 100Hz .  The "problem" scope has just-noticeable uneven amplitude on the low frequency sweep and the other scope is as flat as you could possibly hope for.  Pictures attached.

So for instance a Racal Dana 9301 has a flat bandwidth to 500 MHz, where "flat" means that no amplitude correction is required because it is within the 1% measurement error of the instrument.  The  HP 3406A seems to be a little less than this, but the manual is difficult to read on this point.  A digital storage oscilloscope would require a bandwidth of roughly 5 GHz to achieve this with single pole response and nobody has proposed using such an instrument.

I had a look at the manuals for those instruments because I haven't seen them before.  Yes, the "sampling front end" is a specific subset of sampling systems and it does have the advantages you state, but my larger point was that the sampling system itself in a DSO or the 3458A has the same inherent linearity for the same reasons.  It's the input and scaling circuitry that limits BW.  The thermal converter on my 8506A, for example, would have nearly infinite BW and could be used like an HP thermal RF power meter if there were a way to accurately get a scaled signal to heat up the AC resistor heater in the sensor.  The sampling front end avoids any input circuit issues because it puts the sampling up front.  Of course the trade-off is the limitations on input voltage and input impedance.  Still, even with all that the HP has a best-case accuracy of 3% and the Racal Dana is 1.5% of reading plus 1% FS, so 2.5% at best--0.125dB for the HP and 0.09dB for the Racal Dana.  Now given the 500MHz+(but not LF) that those specs apply to, I have to concede I don't know anything that can do that to 0.125dB.  IDK how flat a 5GHz scope would be to 500MHz, but I'm suspecting it won't be a under 0.5dB.  But for 0 to 1MHz, I think the cheap DSO has them beat.