AA501A analyzer needs a filter for testing switching amps

[davygrvy]

I can't seem to get acceptable results using a Tektronix AA501A distortion analyzer for testing a Class-D switchmode power amp. The 350kHz hash is all over it. I found this NatSemi FAQ that tells me exactly why the DistAnalyzer is getting it's front-end smashed. I want to build the Figure #6 circuit, but it doesn't look optimal with large signals such as this 500W amp I'm trying to test.

Anyone made a diff input filter like this before for such a need? AES-17 says 60dB brickwall, but that's for digital in general though, and this is a bit different.

[davygrvy]

Wondering if I could do this passive?

/me opens LTSpice to experiment

[davygrvy]

Audio Precision's AUX-0025 looks to be exactly what I need. Or from what I should copy. I can't think of a reason not to go passive on this filter except for the pass-band ripple tuning mess.

[hitachi_nut]

What's the AA501A THD+N reading with the 400Hz HPF and 30kHz LPF button engaged? I'm assuming you are testing with 1kHz test tone hence the addition of the 400Hz HPF..

The AA501A's 30kHz LPF is 3rd order meaning a decade(x10=300kHz) away from 30kHz is going to be close to 60dB attenuation. This should give you at least 60dB rejection of the 350kHz carrier relative to say the 1kHz test tone.

Lowest THD+N would be around 0.1% assuming the amp you are testing has a THD+N better than this. What is your amp's spec?

If you need more that 60dB attenuation, build an active single-ended high order LPF between the AA501A's "FUNCTION OUTPUT" and "AUXILIARY INPUT". It's a lot simpler than building your own coils.

[davygrvy]

Originally Posted by hitachi_nut

What's the AA501A THD+N reading with the 400Hz HPF and 30kHz LPF button engaged? I'm assuming you are testing with 1kHz test tone hence the addition of the 400Hz HPF..

Not believable would be the best answer. The scale is at the 200% range, yet numerically it shows 000.1% rather than 0.100%

The AA501A's 30kHz LPF is 3rd order meaning a decade(x10=300kHz) away from 30kHz is going to be close to 60dB attenuation. This should give you at least 60dB rejection of the 350kHz carrier relative to say the 1kHz test tone.

Yet the front-end is still seeing over 2Vrms of 350k which is messing up the scale. When I raise the scale so I get some numerical resolution, the AGC clips. disengaging the filters shows clearly why it's clipping. With the filters, I get very little function output level as well resolution.

Lowest THD+N would be around 0.1% assuming the amp you are testing has a THD+N better than this. What is your amp's spec?

Dunno... that's what I'm testing for

I have to go passive right at the front ala AP AUX-0025, but build it myself as I can't shell-out $965 for one. Thanks for your help anyways.

If you need more that 60dB attenuation, build an active single-ended high order LPF between the AA501A's "FUNCTION OUTPUT" and "AUXILIARY INPUT". It's a lot simpler than building your own coils.

Can't! gain structure not correct for it to be successful.

[Zero Cool]

you have to use an external filter! I will send you a PM on the details.

[davygrvy]

at the front-end passive or between function out and aux in?

Ok, but there's a gain structure issue going on with the built-ins and I'm clipping it with little indication

[hitachi_nut]

Originally Posted by davygrvy

Not believable would be the best answer. The scale is at the 200% range, yet numerically it shows 000.1% rather than 0.100%

You set it to 200% full scale, why would it switch to 2% full scale? 000.1% looks reasonable.

Yet the front-end is still seeing over 2Vrms of 350k which is messing up the scale. When I raise the scale so I get some numerical resolution, the AGC clips. disengaging the filters shows clearly why it's clipping. With the filters, I get very little function output level as well resolution.

The front end is suppose to see everything, that's why the range should be set higher than the maximum level coming in. The filters limit the bandwidth that you want to measure. These are basic THD+N analyzer topology.


Dunno... that's what I'm testing for

You don't specify your design reference so how am I suppose to know? I'm not a psychic and so is 99.99% of the people in this forum!

I have to go passive right at the front ala AP AUX-0025, but build it myself as I can't shell-out $965 for one. Thanks for your help anyways.

Can't! gain structure not correct for it to be successful.

Gain structure? Most active filters are designed for 0dB gain and it is trivial to measure filter response and take that into account in your measurements.

Read the AA501A manual and familiarize yourself with testing methodologies.

[davygrvy]

yeah, but the scale chosen is pre-filter by the AGC, apparently. So if it should be in the 0.2% range, it's stuck in the 200% range

[davygrvy]

You set it to 200% full scale, why would it switch to 2% full scale? 000.1% looks reasonable.

Actually, I don't set it there, auto-scale does. And it's doing it wrong because the gain structure determined by the AGC, again, is apparently pre-filter. And unluckily for me my switched carrier is rather large. So when the 80k filter is engaged, little resolution is available for measurement of the residual.

Also, I need at least 5V of audio signal from the DUT to even get the auto-null to lock. It should be fine down to about 70mV normally. built-in filters or not.

/me goes back to thinking about a 0.1dB ripple Chebyshev 7-pole passive filter network