Small title bar4

  Home -- article index                                                                    

Complete rewrite 6-25-2012

Greening the HP 331A-334A series of Total Harmonic Distortion Analyzers,
with a focus on the 333A & 334A (and on the Heath IM-5258, too)

Introduction --
The HP 331A through 334A are fine instruments for measuring THD over a very wide frequency range, and they are also competent wideband AC voltmeters. At the time they were designed, their THD measurement floor of around 0.01 - 0.02% was as good as the equipment they were designed to measure. The 331A and 332A do not feature auto-tuning, whereas the 333A and the 334A do. This is important, because for very high performance use, I find auto tuning to be a necessity. The difference between the 331 and the 332, as well as between the 333 and 334, is that the latter two have built-in AM RF detectors -- something I've never had occasion to use.

These instruments are compact, 3U rack mounting or bench-top units, well-built and easy to use. I choose to speak about the 334A -- it is what I have -- but everything I say about it pertains to the others as well, since except for auto-tuning, they are otherwise identical. Note that I include the Heath IM-5258 in this discussion, as it is a nearly identical copy of the HP 333A, with only minor circuit and component differences.

I began the idea of greening the 334 because I felt that only minor circuit changes would be required to significantly improve its resoltuion. Naively, I thought that just using high-performance, low-distortion opamps in place of the discrete-transistor amps would pretty much get the job done.

I was wrong -- significant performance improvements require much more than just using opamps -- they require an extensive redesign and rebuild of the entire instrument. I went far down this road, and ended up dissatisfied and frustrated.

Can -- or should -- the 334A be greened?
The answer depends on the level of performance that can be easily achieved, compared to newer analyzers. See my webpage About THD Analyzers here on the site for a general overview. But getting much higher performance will require big modifications and a lot of work, so it's only worth it if you have the interest and the time, and like this kind of intensive work.

Newer analog analyzers -- Newer analog THD analyzers, like HP 339A, the HP 8903 series, and the Sound Technology units, offer a 20dB improvement in resolution over the 334, and several employ state-variable filters that offer THD floors of 0.001% or better (12-27-2010 -- I got an 8903E and it's floor is around 0.0009% mid-band but it has many gain stages ahead of the notch filter). Some, like Bob Cordell's analyzer, are better than 0.0005% in the audio mid-band, albeit without the HP 334's maximum measurement bandwidth -- but most have a bandwidth of at least 500kHz, meaning reasonable measurements on signals up to 100kHz, if only the low-order harmonics are significant. But commercial analog analyzers are still very expensive, even used, often running well over $600 on auction websites.

Software analyzers -- These are capable of very high levels of performance, like, for example, the units from Audio Precision, which are limited only by the linearity and resolution of their analog input circuitry and by the performance of their analog-to-digital converters (ADCs). What they mostly don't have is wide measurement bandwidth, being limited for the most part to signals under 30kHz for the very best (and expensive) performers, and much less -- a few kilohertz -- for the typical computer system's on-board ADC.

For example, my PC has the on-board "Intel High Definition Audio" chipset, which for input signals has 24-bit amplitude resolution and a 96kHz sampling rate. I've also used a Sound Blaster 24/96 PCI card for input, and I tried an M-Audio Audiophile 192; neither was any better than the on-board system, but they were about as good.

Using the very powerful ARTA spectrum analyzer software, the Intel HDA chipset seems to have a THD floor of about 0.0015% at 1kHz, and is usable to 10kHz. The Sound Blaster has a THD floor of about 0.004%, and is also usable to 10kHz. For mid-band audio work, the Intel on-board chipset is acceptable for all but the most critical work, comparing favorably with an HP 339A in everything except bandwidth. The Sound Blaster just doesn't quite cut it in comparison, though it's still very usable.

Software analyzers are relatively inexpensive, but good ADCs are not, so the total cost can get high fast. But none of the floors of these cards will matter if you use them to look at the output of a notch filter, so that the fundamental is lowered to the point where the residuals of the sound card are down in the noise.

The best combination I've found, and a good reason not to green or mod the HP 334, is the E-MU 0204 24-bit/192kHz USB sound module. Used with an Active Twin-T notch filter and the ARTA software on the PC, this combination allows measurements of harmonics up to about 90kHz, so it works for 20kHz fundamentals if only low-order harmonics are of interest, and offers full performance for 10kHz fundamentals. Once the Active Twin-T filter removes the fundamental by 40dB or more, the E-MU and ARTA allow accurate resolution of harmonics below -130dB for mid-band audio signals -- an astonishing 0.00003% or better.

No hardware THD analyzer can match this performance. So before you entertain the idea of greening -- actually, extensively modding -- an older THD analyzer, consider the low cost and efficiency of building the Active Twin-T filter and buying the E-MU and some audio spectrum analyzer software.

Thinking of going ahead anyway?
334s are regularly available on eBay for under $200 -- often well under. The one I have was $70 delivered, and just needed the tuning dial mechanism lubricated. Thanks to rugged design and great parts, they usually work; the biggest problems typically being lamp burn-out or failure of a CdS photoresistor, both in the auto-tune circuitry. Both problems are easily repaired.

Basic design of the 334A
The 334A (and sibs) utilizes a Wien Bridge filter in the rejection amplifier system, with switched resistors for range selection and four ganged air-variable capacitors, arranged in pairs, for in-range tuning (the Heath IM-5258 uses a two-gang air variable and resistors twice as large). This design makes it easy to use and gives good performance. The rejection/summing amplifier is a low-noise P-channel JFET, which gets the Wien Bridge band-pass filtered signal to its gate (inverting input) and the wideband unfiltered signal to its source (non-inverting input).

The amplifiers of the 334 are all discrete-transistor designs, and generally have extremely low noise. The tuning range of the 334 extends from 5Hz to 600kHz, with a measurement bandwidth of at least 1.2MHz on the lowest 300uV/0.1% voltmeter range, and about 4MHz on the higher ranges.

The standard noise filter is usually a 400Hz high-pass (hum) filter for use with fundamentals of 1kHz and up, but some instruments instead have a 30kHz low-pass filter. But no variant has both -- too bad. Clearly, a significant aspect of greening will be to add low-pass noise filtering. This is feasible because for circuits with very low distortion and a Gaussian response, the only significant products will be low-order, generally 2nd and 3rd harmonics, sometimes fourth or fifth. This means that a noise bandwidth about 10 times the fundamental frequency will capture the important harmonics while significantly reducing noise.

The voltmeter and distortion ranges are in 10dB steps, which is very convenient for reading output on the large meter. The lowest range of voltage/distortion measurement is 300uV and 0.1% THD. This is not super sensitive, but would let 0.002-0.005% be easily readable if the noise and distortion floor performance allow. Having an additional 10dB or 20dB of meter range would be very useful, giving a full-scale THD of 0.03% or 0.01% and being very readable to below 0.001%, again noise and THD floor permitting.

What limits the stock 334 performance?
Input buffer -- Firstly, the unity-gain input buffer amplifier has significant distortion that will limit the performance of the modded 334 to a little under 0.01% -- this is a lot, enough to make the buffer's replacement essential. An OPA134 or a OPA1641 has sufficiently high input Z and bandwidth to replace the discrete-transistor circuit, but the power supply voltages will have to be reduced -- that's a complication here, and is one that affects other modified circuits as well.

Rejection amp/notch filter -- I took a look at the A3 rejection amplifier, using my PC with the Intel chipset and the ARTA software. The performance floor of the rejection amp is limited by a strong 2nd harmonic that is at about 3X (10dB) above the noise floor. The fundamental essentially disappeared and the auto-tuning seems capable of holding a much better than 90dB notch depth, maybe even better than100dB.

So, where is the 2nd harmonic coming from? One possibility is that there are channel modulation effects in the JFET, because the signals levels at gate and source are actually fairly high, running above 1VRMS at null. A second possibility is that the input stage of the rejection amp system just has a bit too much distortion. The post-filter stage that follows the JFET diff amp doesn't bear on this problem particularly, since at null, the fundamental has disappeared and the levels of the harmonics are extremely low. Noise is another matter.

Options for modding the notch filter -- what doesn't work
Wien bridge -- Keeping the existing Wien bridge tuning and modding the A3 rejection amp is a non-starter. Using high-performance opamps, like the NE5534, OPA134, or OPA604, I tried every amplifier configuration that I could think of, from instrumentation amps to convoluted feedback systems. They either didn't work at all or had serious limitations -- high noise, limited bandwidth, or severe frequency response aberrations. All were variations of summing amp configurations, and all were compromised by relatively high 2nd harmonic levels.

Bridged-T -- Using the HP 339A's Bridged-T circuit also didn't work exceptionally well -- in the 334, it too, is plagued by high 2nd harmonic levels, just as it is in the 339A itself -- in the 339, the 2nd harmonic sets a resolution floor of around 0.001%.

Fair to say, getting performance the equal of the 339A is not a bad thing, but the mods required are large, complicated, expensive, and painful to execute. It means removing the tuning circuit elements and rewiring or replacing all of the notch filter tuning switches and parts, and building a completely new A3 rejection amplifier board. In short, it amounts to a complete rebuild of the 334, not a "greening."

Any hope?
A possible mod that could yield exceptionally high performance would be to use a Twin-T filter similar to the one I built, and that was used by HP in the 4333A. I believe the 334's auto-tuning circuits will work with the Twin-T filter, so that's a plus. However, the Twin-T circuitry, like the Bridged-T or a State-Variable filter, will require a significant rebuild/redesign of the 334, and is not a "greening."

That is not to say that it wouldn't be a successful undertaking, but I would want to make many more mods to the 334 than just rebuilding the filter section, such as adding a true RMS meter circuit and variable low-pass filter networks for noise and bandwidth control. All of that requires more panel space than is easily found on the 334.

For those interested in pursuing a mod like the Twin-T notch filter, I suggest looking at the circuitry of the HP 4333A. Much of it lends itself well to implementation with opamps. In fact, I think that using the 4333's circuit topology in the 339A, albeit with opamps, would be a very worthwhile way to improve it's performance, although that, too, is a major rebuild.

I started out with high hopes for this project. If you have read the earlier version of this page and all of the updates, then you know just how complicated it all got. I may yet change the 334 over to the Twin-T notch filter, just to get the chassis off of the floor and doing something useful. But no simple greening of the 334 will occur.

back to top

to Home -- article index

© 2010-2012 Dick Moore