testing doa's

[pucho812]

Have several doa's that we have been testing at work. Some are the obvious ones like a 990 and a 2520 while a couple others are discrete designs that are not common(well known).  What I am trying to figure out is this. Using the AP system one and our five fish audio opamp test jig, the opamps tend to perform similarly to one another.  There are differences in things like THD and noise on the different opamps which would be expected. What I found interesting is with the same settings the opamps behave very similarly to one another in terms of frequency response having differences that are minor, less then a tenth of a DB usually. In fact the big differences are THD and noise as one had a spec of -90db while another had a spec of -80db. This is with sweeps of around 0dbu.  However when taken off the test jig and used in circuit, a 312 style pre, they all behave much differently and an audible difference can be heard. Again same settings all around once using the same pre amp channel and once using multiple pre amp channels with matched settings.  No one knew which was which but they all agreed there was a sonic difference between opamps.

considering the AP says they close enough in terms of behavior, what would explain the noticeable sonic differences? Or is it all like psycho acoustics and perception?

 

[usekgb]

I would be curious to know how they would compare with a modern IC op-amp as well.  Have you thrown any of these in to your test rig?

 

[JohnRoberts]

Reality...

Opamps are 5 terminal devices, so noise or AC voltage on PS rails can affect output.

Load, matters, so heavier loading in use than as tested can make a difference.

Also loop gain margin or noise gain in use vs as tested.  If you test it at unity gain but use it at 30 dB expect metrics to degrade.

Finally topology as used vs as tested. While this is not as apparent in commercial opamps, DIY op amps may have different performance when used inverting or non-inverting. (non0inverting  at low gain may be harder for some DOAs).

If you can reliably hear a difference between units that do not show up in your tests, your tests are not measuring the right stuff... You must test the opamps as they will be used worst case...

JR

PS: I believe Samuel tested some of the more popular DOAs on his bench so he may have some useful data.

 

[pucho812]

hmmmmm so testing with the AP is moot? Would it be best to retest and measure using the 312 preamp circuit as the benchmark?

 

[JohnRoberts]

No the AP is fine, but you are not testing the raw opamp with no feedback network, but configured for some fixed gain, and loading. For the bench test to be representative, it needs to mimic the way it the device is used.

I am not familiar with the 312... but testing the DOA in a 312 will be the benchmark for 312 applications.

Sorry to be circular, but that is the nature of circuit design, how parts are used can matter.

A DOA that is good inverting may not be so good non-inverting.

Bench testing is useful for abstract grading of parts vs others. but the bottom line is always about how they behave in actual applications.

JR

 

[pucho812]

Thanks John...  well worth noting...

 

[JohnRoberts]

I do not mean to diminish value of bench testing, if anything IMO this is more about some DOAs being a little sketchy.

If you look at a full data sheet for a modern high performance opamp you will see pages of charts and graphs.

Unless you bench test the DOAs to that level of rigor, there may be snakes in the woodpile that you can hear in some applications, that didn't reveal in limited bench testing. 

JR

 

[pucho812]

understood. The purpose of this whole thing was that the company I work for makes a DOA for some of the gear we use. The DOA does sound quite good in our gear but is not the a-typical 2520(API doa) foot print.  TO be honest I like our pin out better as it is a space saver. But anyway, I converted a few over on vero board to see how they would perform in place of the API or 2520 foot print doa's like the 990.

 

[gridcurrent]

Have several doa's that we have been testing at work. Some are the obvious ones like a 990 and a 2520 while a couple others are discrete designs that are not common(well known).

However when taken off the test jig and used in circuit, a 312 style pre, they all behave much differently and an audible difference can be heard. Again same settings all around once using the same pre amp channel and once using multiple pre amp channels with matched settings.

It's been discussed ad nauseam, but the 1:8 ratio input transformer on a 312 card is a poor match for a 990.

 

[dfuruta]

The introduction/methodology in Samuel Groner's op-amp distortion paper has some example test circuits, and a lengthy discussion of what one might hope to learn.

 

[abbey road d enfer]

Using the AP system one and our five fish audio opamp test jig, the opamps tend to perform similarly to one another. 

How have you set-up the 5 Fish test jig? What topology (non-inverting, inverting, follower)? What gain? What load? What level?What voltage?
One of the major points of DOA's is the capability of driving heavy loads, and that's probably one area where there are differences. If your test jig has minimal load, you may not be in a situation to discriminate.
In a 312, the 1:2 iron-core output xfmr is a challenging element.