HP 339A Info

MBS = NBS

Why they put them letters right next to each other ?

ng like this desirable. It was bought surplus on the cheap. It is not dead, but I am pretty sure it is not working correctly. I've yet to completely l earn to work the thing actually, and am taking it slow.

ut the fundamental frequency and measures what remains. Not quite rocket sc ience. Unfamiliar territory for me indeed, but I should be able to understa nd this.

nd so forth. I have only gotten so far in the manual but I am on it. The wa y I see it, these things you take a sine wave, put it through whatever and calibrate the distortion meter somehow. Like set it to full scale, flip the switch and you get the distortion reading.

e actually confirmed that it does detect distortion. Instead of the onboard oscillator I used my Wavetek 111. Not the greatest but good enough for wha t I wanted to do. I got the thing to give a useful meter indication and swi tched waveforms. I switched to triangle from sine and the needle didn't go up. OK, now realy I am not sure if the peak/RMS value of a triangle wave is less or more than a sine wave. And I fully expected the square wave to peg the meter, which it did.

eedle rise could I be sure this thing is anywhere near being able to really measure distortion. I don't care what, that half sawtooth with the 50 % de ad time will deflect any measuring device less whether it measures average, RMS, peak or s***ma. PLUS the thing is only positive going, so the peak va lue is half. And yet the needle moved up.

you want to really see the waveform. And it is that waveform like, exactly .

needle pegs and takes a few seconds to settle back down. Now, with HP I am not impressed. some of their designs, certain other things and their ideas of the human interface, I think suck. but the fact is they built this, not me and no matter what I think about some of their dowside, I do not bleieve they designed the thing so it would peg the meter on things like range cha nges. Mode changes. Like from input to distortion.

ks, bulging caps or anyhing of the sort. Of course this is going to be done again but we stuck the lid back on to see how it works.

some come along, as long as the thing works.

basic test. Like if I switch from sine to square, is there a number that it should read like 50 % or something ? I mean like at a standard say 1 KHz a nd a specified rise/fall time. Can we get close somehow ? Like if you buy a used ohmmeter off someone that is supposed to work you would take a resist or and check it to see the thing actually works.

er else might help this thing stay out the dumpster. Calibration attempts l ater.

** The HP339A is a high performance, auto-nulling THD analyser, capable of resolving 0.001% distortion.

To test it you *must* use the internal low distortion oscillator and set t he unit to auto null at say 1kHz. After a few seconds the reading should fa ll to around 0.001 or 0.002%.

Pegging the meter is not a fault, until the auto nulling system has finishe d its job large amounts of fundamental frequency get through the notch filt er/s and the meter is simply showing you that fact.

It is very instructive when using a THD analyser to view the residual signa l on a scope - so you know what it consists of.

Might be mostly AC hum, random noise, second harmonic or crossover distorti on spikes - or a mix of some or all the above.

.... Phil

One simple method is to wire up a comparator to turn the internal oscillator signal into a square wave.

The fundamental component of a 1V amplitude square wave (2V p-p) is 4/pi volts, and the THD is everything else. The THD is sort of a voltage ratio (it's the square root of the power ratio), but whatever it is, the impedance divides out, so for simplicity let's figure based on 1 ohm impedance.

Total power = (1 V)**2 / 1 ohm = 1 W

Fundamental power = (4/pi)**2 / 2 = 0.81 W

Thus

Harmonic power = 0.19 W

and

THD = sqrt(0.19/0.81) = 48.4% (*)

You can take that and add it to the original sine wave via voltage dividers (a decade resistance box would be useful), and calibrate the meter down as low as you like. 90% sine + 10% square = 4.84% THD, etc.

Watch out for the comparator pushing crap back out its input, which can be a problem at very low levels.

Note that because they use a notch filter, these meters measure THD+noise, not just THD.

Cheers

Phil Hobbs

(*) Another definition of THD is sqrt(harmonic power + noise power)/(total power), which for a square wave would be sqrt(0.19/1) =

43.5%. Doesn't make much difference at tolerable distortion levels.

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

I assume that is what is at the "monitor" terminals. Is that correct ? Iff thsat is what it is I would like to have a scope connected there, or later when I get one, a spectrum analyzer maybe ?

"

Very interesting. Once I got the thing to the point where I think it works I will have to look into that. An actual ay to calibrate it, though not rea lly accurate it does prove it works. Maybe with a little bit of trouble it can be confirmeed down to a certain resolution.

The Wavetek has a fixed square wave (in fact all of them) and a variable ou tput that is selectable. Couple of resistors should do it.

So 90/10 should read 48.4 %. That means 99/1 should read 4.84 % ?

Question - am I going by peak value or RMS/average or what here ?

On a voltage basis, 100% square is 48.4, 10% square is 4.84, 1% square is 0.484, 0.1% square is 0.0484, and so on. (On one definition of THD it's slightly inaccurate at 10% because there's only 98% of the sine power at that point, but it's the lower numbers you care about anyway.)

From a signal processing POV, the THD meter reading is the square root of SINAD, i.e. the ratio of (everything except the sine wave)/(the sine wave) on an amplitude basis, not a power basis like a SNR or SINAD. I gave the definition in my last post.

The intention is to have a number you can read off on a scope, so it's basically voltage. You can't usefully add up the amplitudes of different frequency components, so the audio THD definition uses the square root of the power rather than the voltage per se. You can think of it as RMS distortion divided by RMS signal.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

** One needs a test that shows the instrument is doing the things a THD analyser is meant to.

1. Thoroughly remove the fundamental frequency.
2. Pass other frequencies up to 100kHz to the metering circuit.

Frequency sweeps from 20Hz to 100kHz will verify this - with the instrument set to notch a variety of spot frequencies in the audio band.

The notch should be sufficiently narrow at each test point that little attenuation exists at half and double the centre frequency and beyond up to a -3dB point at 100kHz.

... Phil

And

1. Measure distortion accurately.

The square wave/attenuator trick does all of that.

(I actually own an HP 339A but haven't used it for anything serious.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

** Like hell it does.

** That is very obvious.

... Phil

Really? Why not?

If the indicated distortion gets down to the noise floor, the fundamental has to be gone to that level, which satisfies #1.

If the indicated reading is correct and stable over frequency, then the notch isn't attenuating the third harmonic. Some more asymmetric rectangle would get you the second harmonic, but those are a bit more difficult to calculate, especially for a technician. So #2 could be improved, but is probably fine if the filter is actually working.

And nothing in your method addresses #3, the accuracy question, whereas the THD of a square wave is known very accurately, and an attenuator allows you to make accurate distortion standards down right into the mud. What's not to like?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

The 339A manual is on the BAMA site at edebris.com

** Passing test 2 is essential, it shows the instrument complies with the d efinition of THD+N used in audio.

Unlike your test, it identifies if the notch has the required narrow bandwi dth and that the high frequency response is correct. Your test, if done at 8kHz or above, will show a significant under reading error on a faultless instrument due to progressive attenuation of the harm onics.

BTW:

A *far simpler* test of overall THD reading accuracy is to mix a small sine wave signal of known amplitude from another oscillator with the internal o ne. Long as the new frequency is set clear of the notch filter's effect, th e correct reading in percentage is easily calculated.

.... Phil

Seems you have answered a question I have not yet asked. Like, if the readi ng is to be considered valid there has to be a specified bandwidth otherwis e you are excluding the meaningful or including the useless. Like, you do n ot need to know the content at 4.8 GHz.

So 100 KHz is the cutoff then ? I never checked. Seems reasonable, that wou ld be the fifth harmonic of 20 KHz. If anyone can hear that we should all l ine up to kiss their ass and give them time to advertise the tickets.

Anyway, my buddy took it home today. It is his actually, but we just umm, d on't care which place a certain piece of equipment is at all the time.

But now I want to get more specific on this mixing a square wavee into the sine wave thing to do an ersatz calibration check. Right now it is reading pretty high distortion on its internal oscillator, and actually reading low er on my Wavetek. I haven't been to all the ranges of the internal oscillat or yet, nor on the Wavetek. At 1 KHz, the Wavetek read like 0.48 %. the int ernal oscillator was like 27 %. This does not seem right, maybe a ground pr oblem or bypass caps or something because it just doesn't seem all that lik ely. I would like to have a distortion meter to check the oscillator on thi s distortion meter. Yup.

The wavefoprm looks alright by eye. I did not try nulling it with the Wavet ek nor doing the Lissajous thing to see if I oculd get a decent circle. but it looked alraght and back then I did see a decent sine wave. the high dis tortion reading came after it had an all night burn in.

Then, unfortunately it was time to go.

He has to learn how do run the thing. Hey, I didn't know initially. I know the theory behind it but operation is another thing. I got it figured out p retty much and next will come the finer points. Like, yeah it measures, but does it measure accurately ?

Next question. This thing FILTERS, not really nulls out the fundamental. th is allowws it to accurately measure through a system that has phase shift. Otherwise an OP AMP would make the best distortion meter. So it gets lovcke d to that frequency by something that resembles a PLL at least remotely.

Now, if I have 1 V P-P sine going in there at 1 KHz, if I took another gene rator, not locked or anything to the fundamental one, and injected a 2 KHz sine wave into it, the distortion meter should read that accurately, correc t ?

I mean I am putting in a controlled amount of the second harmonic, but it i s not locked in any phase relationship and in fact is not even frequency lo cked.

WILL THAT WORK ?

That thing wiht the mixing a square wave with the sine wave, I want to try that. But the ratio, is it RMS voltage or P-P ? A sine wave makes it easy b ecause it is the same as the fundamental. Adding it at the second harmonic at 10 % the amplitude clearly should read 10 % on the meter, correct ?

In fact, now that I think of it I shold be able to add ANY frequency to it not at the fundamental and the meter should read that. Read the dB scale or the percentage, whatever snokes ya up. (yeah, I hear some are starting to describes THD in -dB instead of percent)

Anyway, now that I think of it, all I need is a second generator.

Interesting.

** It's the typical -3dB point of the metering circuit. ** Correct - 2nd, 3rd & 5th harmonics plus noise are read by the meter when if 20kHz is the fundamental. By convention, this has long been is deemed adequate to quantify non-linearity in audio circuits. ** A notch filter IS a filter, adjusting the centre frequency to match an incoming sine wave is called " nulling". The 339A has a notch filter. ** The auto nulling feature tracks the incoming frequency, long as it varies slowly. Without it, small temperature drifts in the oscillator and notch filter makes manual nulling very tedious or impossible when dealing with low THD percentages. ** Yep - a pair of say 10kohms in a Y arrangement allows the signals to combine at the input to the 339A. The level of both will be cut in half, of course. ** Yep.

** Best set the level of the second generator to not more than say 10% of the main one, so the auto-nulling system is not confused.

.... Phil

Now I notice the onboard oscillator goes up to 110 KHz. Also, I notice that the bandwidth is only 110 KHz. This should mean that it really cannot meas ure any distortion on the 110 KHz from the oscillator. At 55 KHz it would o nly pick up the second harmonic which is pretty much useless.

Now that the thing seems to be acting right, when it comes back it will be time to see about accuracy. It actually belongs to my partner but we kinda don't care who has what as long as we got the basics. He has frequency coun ters, and this. I have everything else. He lacks a really good function gen erator, well he decided to tinker with it. There's another story that might warrant a thread. I did a workaround that made it work almost perfectly bu t this guy is a fanatic about measurement and things like that. I can start the thread but any real working on it will have to wait until I figure ou t what went wrong. I think he has some wires connected wrong. They are all individual with almost no markings, though some of them have a little paint on them. Few.

So basically this thing will be back. It was here a few days. Probably in a couple of weeks after he tinkers with it.

Anyway, it is good to know that I can check this thing. If the readings are within reason I should be able to trust it enough for the kind of audio I do. See I will work on high end and vintage stuff, but I do repairs, not re stores. I don't go changing every capacitor in a unit with those high resol ution types. I have tried to explain that to some people and some understan d yet others claim thy can hear the difference.

The way I see it is if you can hear it I should be able to measure it. In f act that claim was made and proven by Bob Carver. He, in one act (well a se ries of the same) not only proved that the regular IHF specs don't tell the whole story, but also put some of the audiophoolery to rest. It IS measura ble, and what I say is that if it IS measurable then what is not measurable should not be hearable.

Then it gets to the realm of psychoacoustics. Then it also gets to the poin t where people might WANT some distortion. Electric guitarists are certainl y not afraid of distortion.

Anyway, back to this. Since it is a notch filter based device we know it is measuring THD + noise. Anything outside of the notch counts.

I have contemplated a way to measure that without such a device. (yeah, now we got one and won't need it HA !) I have, in the past, taken and nulled o ut the input and output of an amp using the add function on the scope, and usually having to invert one channel. The scope will display THD+N but will also indicate if there is any phase shift. You start getting up near the u pper limit in frequencies, many audio amps have considerable phase shift.

To solve that, first of all I do not want to tamper with the signal coming back. But the sine wave going in, where it goes either to the DUT or the sc ope, I could shift the phase around a little with a cap and resistor. Botto m line, in an amp the phase is going to lag of course. So all I have to do is lag it a little bit where it goes into the scope or whatevr actually doe s the nulling oin this case.

That still leaves the problem of deriving an RMS measurement out if it whic h means nulling it externally and using a true RMS meter. I have a Fluke do wn there. I just checked and it says it is down -3 dB at 200 KHz.

So I can use the scope to get the best null in both aplitude and phase with a pot and a cap, and then what the Fluke sees on there in RMS should be co mparable to the total amplitude and then we can derive percentage easily. I n fact using the KISS theory just keep it at like 10 volts RMS or something that makes it easy to divide. (when possible of course) Make it read 10 on the Fluke and then switch the source to the nulled signal.

Now this is nulled electronically, not with a tuned circuit.

Except for the notch filter, that 339A is just a generator and an oscillato r. Question is how hard will it be to get a comparable accuracy as opposed to something like the 339A ? One thing about it is that any distortion in t he source oscillator is also nulled and is taken out of the picture. This should lower my measurement floor if it works.

Another question, should you decide to field, and able, is what about PURE THD ? I mean pure THD without the noise. The only way I can figure is to ex ploit the randomness of the noise by somehow recording the output and synch ing it somehow to get an average reading from a number of cycles.

The problem with that is the only real way to do it is digitally or BBD. If I want a low measurement floor there, that is going to have to be one hell of a BBD.

But is there another way ? They routinely say THD, not THD+noise. Either th ere is a way to null out the noise, or they are using signal levels that ma ke the noise insignificant. I don't see how this would be easy on something like a moving coil phono preamp. Very few, but some of them are compliment ary push pull and in those you might want THD measurements at the lower lev els. In the ones that just use single ended stages not so much. Mostly you would want to use higher levels anyway because the THD levels will be highe r then. It is just that it is not true of the fully complimentary ones.

Sorry this is getting so damn in depth but I guess that's the way the rock rolls.

** Digital scopes can do that job, by averaging a number of screens.

When the scope is synched to the test frequency (using ext trig), it is also synched to the harmonics. So averaging 8 or 16 screens removes much of the random noise and hum.

... Phil

the main one, so the auto-nulling system is not confused.

.... Phil "

Hmmm.

I just sent an email to my partner on this, thinking about going up to 25 % . If I let it lock, or tun or whatever, would I gt away with those higher l evels ? Or are you talking about different units or something.

This is askable because this thing has a scale to read up to 100 % THD.

The way I am reading this fromm you is that the tracking circuit may be ups et by frquencies too cloes. Would phase locking them help ? I can fuly unde rstand a 1.4 KHz square ave assauling the tracking circuit, that is obvious .

But then, IF that freqquency rsponse of the detector is realy flat it doesn 't matter what I put in there as ong as it is far awat fro the fundamental. What if I use 1 KHz for the fundamental and 5 KHZ for the induced signal ? (from the external oscillator) Surely that should not confuse it.

What's more, ou probably don't want a test " pollutin" frequency any closer than second harmonic. I can see that if you are testing the veracity of th e filters themselves, but for most THS measurements doesn't matteer. Howeve r I intend to beat on the thing.

We got scopes, and I might just be using them all when this happens. In fac t it might be a nice to have a high Q 1 KHz tuned circuit to eliminate any THD in the HP oscillator. I am not holding my breath.

We'll see how it goes. I am finally up on working the thing and next we scr utinize. For the audio business, when I test something on it, usually I can just compare channels. But it is nice to have at least a semi-accurate set of absolutes.

Update :

Now have calibrated distortion in via a matched set of resistors and an ext ernal generator. It measured 0.055 % on its own oscillator. I had a 10K rom that nd a 10K from the HP generator with 600 ohm output impedance.

I calibrated, which is to set the level so the needle goes to 100 % and the n you switch the switch.

We found that even moving around the room affected the reading when we were reading that 0.0055 %. Also, because of my slight (only one order of magni tude) error, we were measuring 10 KHz, not 1 KHz. Of course all the filters were off, but I do want to try those filters.

I started to get close to the null frequency with the external generator. I saw how the response dropped.

The scope on the left was connected to the monitor, where you can see the T RMS meter input. I am about to chew through these ropes and get a spectrum analyzer to hook up there. The scope to the right has the original HP 339A output going to one channel and the output of the other HP generator sittin g on top to the other channel. Thus it is easy to compare the amplitude of their waveforms.

When the amplitude was 25 %, the distortion meter read 25 %. I mean when th e internal HP was 3 volts, and the external was 1.5 volts it read 50 % dist ortion. Since the TRMS convertor is after the input attenuator, is it not r easonable to assume that it will be accurate at different ranges ? that is determined by resistors just like every, and I mean every other thing on th e planet.

Well there are exceptions...

THD is a ratio of RMS voltages, so 1.5V vs 3V is 50% THD, not 25%.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net