Testing Freq Res of audio amp w/ voltage divider and things got weird...

[Shredhead]

Hey guys, I'm stuck here. I made this voltage divider so I can turn the amp I'm testing to a little over a watt and use the out of the divider to send back into my frequency response software. (The voltage divider is going to an RCA connector in case you can't tell)

I've tested about 5 other amps using this circuit and the results were what I expected but I'm testing a new amp and I got this screwed up looking response.    There is nothing wrong with this amp that would make me expect to see this drop off in the higher frequencies. 

Any idea what the deal on the higher end is? Is it because the 8 ohm load is non-inductive or is something weird happening because of the divider circuit or what?

The amp is a clone of a lab gruppen FP series (pro audio). (I can't find the schematic)

[PChi]

Were the other amplifiers all linear?
If it is a 'tracking class D' is any residual switching noise is corrupting the test equipment?
What does the amplifier outoput look like on an oscilloscope?
A Sound Card could get confused with swithing noise causing aliasing.
A passive filter could be added to attenuate the switching noise. Audio Precision make a Class D amplifier measuring filter for a price.
Possibly the frequency response measurement is true. I would have thought that the challenge in getting 'TD' to work would be getting a power supply to behave well when having to track a signal.

[Shredhead]

Here is an example of another amp that I tested.  It is a Crown X1000 which is all analog.  The Other amp with the messed up high end has a tracking SMPSU but otherwise it is a complimentary class A/B amp.  The noise floor was around -85dB with no input signal.  I don't have an O-scope. 

Do you think the PSU is injecting some kind of noise into the signal?

[jlmoon]

Impedance matching looking back into output circuitry from the divider?  -10 db is alot of drop, is the amp scaling back due to feedback error into current mirror / bias system?

[Paul Rose]

Ground?

Is it possible that the  - output of that amp doesn't like you to ground it ?

[PChi]

Here is an example of another amp that I tested.  It is a Crown X1000 which is all analog.  The Other amp with the messed up high end has a tracking SMPSU but otherwise it is a complimentary class A/B amp.  The noise floor was around -85dB with no input signal.  I don't have an O-scope. 

Do you think the PSU is injecting some kind of noise into the signal?

Yes I think that the PSU will be injecting some noise into the signal. I don't know if thie is causing the measurement problem though.
It could be a failure of the tracking SMPSU so the class A/B amplifier clips because the power supply fails to respond quickly enoungh. It's possible the power supply needs a few audio cycles at high frequency to catch up. The ear is insensitive to brief clipping.
Are the frequency response measurements performed with steady state sine waves?

[DanielS]

You could start with a loopback test at various frequencies to verify that your input agrees with your output when nothing else is connected - check the waveform for any obvious defects and the FFT for unexpected additions.

If your sound card checks out ok, check the waveform and FFT at the same frequencies as above with the amplifier patched in and compare with the original results without the amplifier. If attenuation at high frequencies is the only difference then the "weird frequency response" is what the amplifier does - if there are any major non-linearities introduced, they should be fairly obvious in the FFT as extra THD+N.

[Shredhead]

Are the frequency response measurements performed with steady state sine waves?

Log sine sweep.  Also, I am only driving the dummy load to 1 watt.  The amp is rated for peaks of 14K watts so I can't imagine it would be clipping without melting my 20 watt power resistor. 

[Shredhead]

Ground?

Is it possible that the  - output of that amp doesn't like you to ground it ?

The amp shares the input and output grounds before anything is connected so I don't guess so.  Noise was pretty low with no input.

[Shredhead]

Impedance matching looking back into output circuitry from the divider?  -10 db is alot of drop, is the amp scaling back due to feedback error into current mirror / bias system?

I am too dumb for this question, sorry. 

[Shredhead]

The input impedance on my soundcard is 4.4K ohms if that helps anybody. 

Could this have something to do with the output impedance of the amp as it varies with frequency and the non-inductive load is loading it down higher up much more than an actual high range speaker would? 

[PChi]

Are the frequency response measurements performed with steady state sine waves?

Log sine sweep.  Also, I am only driving the dummy load to 1 watt.  The amp is rated for peaks of 14K watts so I can't imagine it would be clipping without melting my 20 watt power resistor.

Can you slow the sweep speed down or up and check for frequency response differences?
If the measurement problem is due to switching noise can you add a low pass filter (something like 1 kohm and 10 nF) at the output of the voltage divider and compare the frequency response of the Crown with the Lab gruppen?
The lab gruppen fp series is sold as an amplifier to go touring so some compromises may have been made.
I don't like their web site, it uses BS:-
Regulated Switch Mode Power Supply (R.SMPS™), pointless acronym, nearly all smps have regulated outputs. Nearly all switch mode power supplies use ferrite core transformers. Where is the innovation?
Class TD. Difficult to get a Class D stage supplying a Class A/B stage and is likely to still have switching noise on the output.
Multiple Position Gain Switch (M.PGS), it's got a gain control, wow.

[jlmoon]

Impedance matching looking back into output circuitry from the divider?  -10 db is alot of drop, is the amp scaling back due to feedback error into current mirror / bias system?

I am too dumb for this question, sorry. 

Shred,
Sorry, didn't mean to sound confusing regarding your issue. 
How about applying a good low distortion oscillator (I use a SG506) to the input and using good resistive load on the output of that amp.. then attach a good level / distortion meter such as a (I use a Tek AA501 properly terminated) and actually plot the performance curves manually so you can isolate the amp from your existing pc test setup.  I would think this should point you towards the right direction ... either the test setup or that amp .. one way or the other.  I would think in 10 minutes you could obtain some good data.  Only a suggestion!

[PChi]

Impedance matching looking back into output circuitry from the divider?  -10 db is alot of drop, is the amp scaling back due to feedback error into current mirror / bias system?

I am too dumb for this question, sorry. 

Shred,
Sorry, didn't mean to sound confusing regarding your issue. 
How about applying a good low distortion oscillator (I use a SG506) to the input and using good resistive load on the output of that amp.. then attach a good level / distortion meter such as a (I use a Tek AA501 properly terminated) and actually plot the performance curves manually so you can isolate the amp from your existing pc test setup.  I would think this should point you towards the right direction ... either the test setup or that amp .. one way or the other.  I would think in 10 minutes you could obtain some good data.  Only a suggestion!

I agree with the suggestion.

[salbayeng]

Can you be 100% certain the amplifier doesn't use a "bridge tied load" output  "BTL"
With a BTL , the negative speaker terminal is connected to an amplifier output and not connected to ground or chassis, i.e. both outputs are live.
Check with a DMM if negative is connected to chassis.

Its sometimes difficult to get sensible results with BTL amplifiers.
Use the headphone output if it has one, these are single sided, and tapped off the main amps.

[salbayeng]

Ok I've found a FP4600 datasheet.
It says it is class D but without PWM??
Confusing,  Possibly it is using some kind of phase shift modulation? ,  this modulation method is used in high power industrial power supplies, it has slightly higher efficiency than normal "class D" with PWM.
 

[Shredhead]

Sorry, I've been away from my post lately.

Thank you all for your suggestions, you've given me some really good ideas.  I'm going to take another look at this problem Monday and see what's what.

I'll let you guys know how things went.