Author Topic: Power amplifier testing (frequency response,etc.) – what kind of load?  (Read 730 times)

0 Members and 1 Guest are viewing this topic.

Offline frozenfrogz

  • Frequent Contributor
  • **
  • Posts: 927
  • Country: de
  • Having fun with Arduino and Raspberry Pi
Lately I have been repairing a lot of hifi and bass guitar amplifiers.
Since I would like to start a small web series comparing various bass amplifiers, I am interested in plotting the frequency response (primarily, but maybe also getting into more detail regarding THD etc.). I do not own a dynamic signal analyzer and am going the sweep generator + oscilloscope route to get a visual representation of the frequency response.

In terms of "best practice", what kind of load should I be getting / building to get decent results?
My current setup is:

Rigol DG812, 10 Hz - 20.000 Hz sweep into amp input
connector box with SpeakOn input, parallel probing jacks, parallel SpeakOn output to connect a load / speaker
Micsig DP10013 differential probe/amp
Rigol DS1054Z (trigger synced from DG)

For transistor amplifiers the difference in frequency response does not really matter in terms of connecting a speaker load (as it seems). However, I will need a suitable load for tube amplifiers.
I am considering to build a load from resistive heating cartridges as it is proposed here: https://www.prosoundtraining.com/2010/03/11/big-dummy-load/

Also, I thought about splitting the signal from the generator into a second osc. channel to be able to plot phase differences. Do I need an active splitter or is a t-connector just fine?

Can you please advise if that is a sane route or if I am better of approaching it differently?

Thank you and kind regards
Frederik
He’s like a trained ape. Without the training.
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 20637
  • Country: nl
    • NCT Developments
I'm not sure whether a guitar amplifier should be quantified based on it's response. AFAIK most guitar amplifiers are highly unlinear by design to produce their 'sound' and therefore they are more like musical instruments than pure amplifiers.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline NANDBlog

  • Super Contributor
  • ***
  • Posts: 5174
  • Country: nl
  • Current job: ATEX certified product design
https://www.digikey.com/product-detail/en/te-connectivity-passive-product/TE1000B4R0J/TE1000B4R0J-ND/5591336
I would just order 4 of these resistors. Or something similar. No mocking around with oil or heatsinks. Direct into the air. Or the 8 Ohm version. Aliexpress seems to have some chinese made ones for insane prices.
 

Offline frozenfrogz

  • Frequent Contributor
  • **
  • Posts: 927
  • Country: de
  • Having fun with Arduino and Raspberry Pi
I'm not sure whether a guitar amplifier should be quantified based on it's response. AFAIK most guitar amplifiers are highly unlinear by design to produce their 'sound' and therefore they are more like musical instruments than pure amplifiers.

I am talking bass amplifiers, those are usually very clean with a lot of headroom. Distortion / overdrive is another story of course.

I would just order 4 of these resistors.

Thank you for the suggestion. So apart from liquid cooled vs. air cooled there is no objection to using "just" a resistive load?
He’s like a trained ape. Without the training.
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 20637
  • Country: nl
    • NCT Developments
I'd use a load which is more comparable to a bass loudspeaker (IOW: also inductive). An amplifier may be perfectly stable with a resistive load but may behave badly with an inductive load.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Online 2N3055

  • Super Contributor
  • ***
  • Posts: 3334
  • Country: hr
Tube amplifier will be HIGHLY sensitive to complex impedance of the speaker.

Only way for this to make any sense (to have some quantitative dimension) with tube amps is to either measure amplifier response with actual speaker as the load.
Or to take several representative speakers (or one reference one) and measure speaker complex impedance (with box, crossower and everything)
and than design/make a "simulator" complex impedance analog for that speaker (complex network built from resistors, inductors and capacitors) and use that for measurement.
Doing that is quite complex project itself, and will yield a rough approximation anyways.

Modern solid state bass amplifier will typically have a preamp that will do all of the sound shaping, and a power amplifier that will be wideband, pretty much like the HiFi one, but with less stringent requirements on distortions.  Any excessive frequency response deviation from flat will be defect. It won't and shouldn't have any real impact on the sound.
If you want to measure preamp frequency characteristics, tone shaping and such, for that PC sound card and software is way better tool. Software will take care of everything, doing measurements automatically. There are many free and paid for ones to choose from. And voltage levels are not high, you can make small attenuators with normal metal film resistors if you don't want to overdrive sound card inputs.

With an 8 bit scope, you cannot measure less than something like 0,12-0.15% distortion, probably a bit worse. With sound card you can measure way better than that, even with the one built in on the motherboard. Easily 0,002 %, or better, depending on card.

So for just testing power amps to full power and such, there are cheap wirewound resistors (inductivity here is not a problem, quite the opposite) , and you buy few of them so you can combine them for 4,8 16 OHm and so you can test stereo at the same time... If amplifier is connected directly to the speaker, an air coil in series with resistor will be better "simulation" of speaker. If you want to be close, again, measure the speaker and you can than simulate it quite easily good enough.

Just to make it clear: speaker is freakishly complex load. It not only has a coil in there that is resistive and inductive, it is an electromechanical system (it is basically a linear drive DC electric motor/actuator) so not only the coil will have it's electric impedance in stationary state, but when actually moving, speaker will create counter EMF force, that will be dependent on resonant frequency and acoustic impedance matching to the surrounding air/box.. That will "mix" with electric impedance of the coil to make speaker impedance parameters.

It gets really complicated real quick.

So you have to figure out what exactly do you want to accomplish and than go from there.
 
The following users thanked this post: Someone

Offline graybeard

  • Frequent Contributor
  • **
  • Posts: 272
  • Country: us
  • Senior III-V RF/mixed signal/device engineer
    • Chris Grossman
I would use these low-inductance thick film resistors mounted on a big heat sink.  If you need more power use series/parallel combinations of the and values.

I have used these as loads up about 100MHz in the past and they have worked out well.  At audio frequencies their connection will not be highly critical, but making sure you have an adequate heat sink will be.

As others have stated here real speaker loads can be complex, but having a good non-inductive load for testing is good thing to have when working on and testing audio power amplifiers.
« Last Edit: June 04, 2019, 02:12:54 am by graybeard »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 12421
  • Country: us
  • DavidH
For more demanding applications, you can make a load which simulates a loudspeaker like that shown below.  For small signal tests, an actual loudspeaker works fine.

I might also include a simulation of the cable with a test load to test the local stability of the output stage.
 

Offline Neomys Sapiens

  • Super Contributor
  • ***
  • Posts: 2060
  • Country: de
At those frequencies, you can work with a simple tee for the second channel. If you have any doubt about that, you could also use your generator's trigger/sync output (I hope it has one), which has a fixed relation to the output signal.
Then first make a measurement of the input signal and then the output.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf