Null Difference Testing

[Circlotron]

I tried a different sort of null test last night on one one of my mosfet/LM317 choke loaded power amp follower thingies. I drove the input from a transformer so it was floating, then put the scope earth probe lead the live input and the scope probe live to the output. That means the only signal read was any difference between input and output. For a 24v p/p driving into 6 ohms resistive there was about 3mV difference just before the onset of clipping, but it was way down in the noise so it was a bit hard to tell accurately. Way better than just trying to compare input and output though.

[Tube_Dude]

. That means the only signal read was any difference between input and output.

Also in my test...the signal between AC is the difference between input and output...

Way better than just trying to compare input and output though

But the first quot means that you are comparing the input and output.. as you are reading any diference...you are comparing...

[Circlotron]

Way better than just trying to compare input and output though

I meant by putting the input on one scope channel and the output on the other and putting the traces one on top of the other to try and see any difference.

Well... I knew what I meant anyway.

[nw_avphile]

Originally posted by mrfeedback
Hmmmmm, 100Hz-1.8mV, 1kHz - 2.0mV, 20kHz-9.1mV = rising error voltage/frequency despite DMM droop/frquency does it not ?.
So who is misunderstanding what ?.

I have no idea what you're trying to say. Your claim was DMM's don't have good response out to 20khz (which is certainly true of some of them). You pointed to my own results as "proof". Well my own results show a RISING null signal at higher frequencies. My only comment was that's not proof of anything as it's the opposite of what your comments would predict.

Originally posted by mrfeedback
In a conventional input diff-pair stage, if the current source is not perfect, and the transistor gains not linear, then a DYNAMICALLY varying input impedence will eventuate, and due to stored energy, stuff will spit back up the interconnect, and confuse the previous line output stage.

The above makes very little sense. First of all, your original comment was that the load on the OUTPUT of the amplifier affects what load the preamp/input cable sees. While load on the amp will change the feedback signal slightly, that will effect the other half of the diff pair very little even with a "psuedo" (resistive) current mirror. Then we have to consider that most of the input impedance is determined by the passive components in the input circuit which aren't affected AT ALL by loading on the amplifier. So you have a very small percentage (feedback), multiplied by another very small percentage (diff pair error), multiplied by another very small percentage (dominant effect of passive input components). The result, by any reasonable standard, is ZERO.

The input network to a typical power amplifier is a mostly resistive load with a small reactive component. This complex load DOES NOT vary in any significant way with loading on the output of the amplifier. If you want to convince me otherwise, show me the math.

[janneman]

Originally posted by mrfeedback
[snip]Hmmmmm, 100Hz-1.8mV, 1kHz - 2.0mV, 20kHz-9.1mV = rising error voltage/frequency despite DMM droop/frquency does it not ?.
So who is misunderstanding what ?.
[snip]Eric.

Eric,

The rising error with frequency most probably is related to uncomplete nulling caused by freq response and/or phase response differences between channels.
This, ultimately, is the limit of the available null even if there is no "traditional" distortion, as I am sure you know. Unless you want to go to really complex nulling correction networks.

Jan Didden

[mrfeedback]

A PERFECT null would show as zero reading on your meter.

Well my own results show a RISING null signal at higher frequencies. My only comment was that's not proof of anything as it's the opposite of what your comments would predict.

Your results show a rising LACK of null with rising frequency, do they not ?.

As you can see, the channel match is over 10db worse at high frequencies--probably due to component tolerances in the feedback loop and board layout issues in the preamp and amp. NULL TESTING IS VERY SENSITIVE!

And you even say so yourself.
Part of my point was, despite a rolling off meter response you are getting increased reading with increased frequency - IOW the matching error is actually worse than your meter indicates.
Go back to bed.

Eric.

[nw_avphile]

Originally posted by janneman
Eric,

The rising error with frequency most probably is related to uncomplete nulling caused by freq response and/or phase response differences between channels.
This, ultimately, is the limit of the available null even if there is no "traditional" distortion, as I am sure you know. Unless you want to go to really complex nulling correction networks.

Absolutely correct Jan. Thanks. And in the example I posted it was accurately measured as my Fluke 189 is essentially flat to 100khz.

If you load the amplifier, you make the null worse (because you bring out more of the sorts of errors you suggest), which could mask subtle differences between cables.

[fdegrove]

Hi,

From what I gather, Eric's refering to "spitting" in a previous post
points to reflections of the signal back into the cable at the receiving end.

Cable reflections can cause distortion of interface and backwall echoes, and in extreme cases (cables on the order of 30 meters/100 feet or greater) they can even result in large spurious signals following desirable signals at an interval equal to the electrical transit time in the cable.

No transmission line is ever prefect, hence the phenomenon can rear its ugly head in audio/video setup causing image ghosting (video) or said "spitting" or reflections in audio.

So, when setting up a test rig you need to be sure of what you're doing.
If not you're actually misleading yourself.

Nothing's ever easy.

Cheers,

[janneman]

Originally posted by nw_avphile

Absolutely correct Jan. Thanks. And in the example I posted it was accurately measured as my Fluke 189 is essentially flat to 100khz.

If you load the amplifier, you make the null worse (because you bring out more of the sorts of errors you suggest), which could mask subtle differences between cables.

Yes, but on the other hand, if you cannot hear any differences even with the imperfect nulling test, as has been stated by some, what are the chances that there is a real difference, masked by the residual nulling error, that you CAN hear (the subtle cable differences) in a normal listening session?

Jan Didden

[Tube_Dude]

Cable reflections can cause distortion of interface and backwall echoes, and in extreme cases (cables on the order of 30 meters/100 feet or greater) they can even result in large spurious signals following desirable signals at an interval equal to the electrical transit time in the cable.

Yes that can hapen in the Mhz region...but is not a problem in the audio band...

And using cables of 30 meters...is a little unusual!!....I think!!

No transmission line is ever prefect, hence the phenomenon can rear its ugly head in audio/video setup causing image ghosting (video) or said "spitting" or reflections in audio.

But interconects in audio and speakers cables don't work as transmission lines(if they work they will need matched impedances at the sending and receiving end)...only the digital interconect between two SPDIF interfaces...but that is another issue!!!