Solid State Testing- Tug of war

[jcx]

I've mentioned from time to time Cherry's formulation of this test - put 2 different frequency signals at each end of the resistor between amp channels and you sweep the output over a large region of the I,V plane

same frequency, differing phase simulates manyaspects of complex impedance load

Originally Posted by jcx

How about some more gain?

Graham’s Class A Imagineering articles spend considerable space discussing amplifier output impedance and interaction with “back EMF” from the complex loudspeaker load, with the possibility of nonlinear back EMF components tossed in as well

Suspicion over feedback control of the output impedance with standard dominant pole miller compensation is suggested as a source of poor performance with the complex/nonlinear speaker load

Cherry has addressed the same issues in “Feedback Sensitivity and Stability of Audio Power Amplifiers” and “ Output Resistance and Intermodulation Distortion of Feedback Amplifiers” JAES V30 #5 May, April 1982

Cherry’s prescription is more gain, specifically more hfe in the VAS and more loop gain to address output stage current and voltage nonlinearities respectively

The JLH output stage Graham uses naturally provides voltage gain, borrowing another of Hood’s ideas (“Gain Stage Investigations” Electronics World july 1998) I changed the Q2 driver transistor (~= VAS transistor) to a small signal mosfet for increased “hfe” at audio frequencies

2-Pole compensation that includes the output stage is implemented with Graham’s component values, slightly rearranged, (Cherry’s nested feedback around the output stage)

I’ve put my thread opening sim of Graham’s circuit together with my modded circuit in the following sim with a complex excitation, 3:1 2 and 20 KHz input at ~ ½ max swing let you look at intermod skirts as well as harmonics, a 1 A, 7 KHz current source in parallel with the speaker load shows output impedance and intermodulation interactions with the 7 KHz output disturbance that clearly show the advantage of increased gain



Increased loop gain at audio frequencies has improved 2 KHz harmonics ~ 30 dB and the 7 KHz output current induced signal is ~ 40 dB down with the mods in this sim

Speaker cable impedance will prevent the low level of the output impedance ( ~ 10 uOhms! ) from helping at the loudspeaker end but the amplifier’s distortion should show greater immunity to Graham’s “back EMF” concerns

[jez]

JCX. Do you have those articles by Cherry and would you be willing to PM them to me? They sound like interesting reading but not being an AES member I can't access them. I had thought about joining the AES but heard too many things about it being "not what it used to be" and not worth joining... or maybe in having to ask you for the articles I'm proving myself wrong!?

[jcx]

I don't have digital copies - the most active years for SS amplifier design info was in 1970's-80's JAES articles, tapering off thru the 90's, very few later articles are amplifier circuit design oriented

[sumaudioguy]

What a great thread with wonderful suggestions!

Have had great success with this unraveling faulty designs. Many amps have great phase margin under no load conditions. Put on the complex load and that phase margin magically disappears and the amp breaks into oscillation on a portion of the wave indicating gain is more than one at with 180 degrees of phase shift. Very bad for audio.

This can also show well common mode problems in stereo designs and lack of power supply rejection. Most amps I have tested show line frequency spikes on the virtual ground channel.

Enjoy

[unclejed613]

Originally Posted by jez

It's an old idea. IIRC Tim De Paravicini described this method for testing damping factor many years ago.

i've actually built a device for doing the current source method. it's a Howland current source with an added output stage. i need to do a bit more testing with it, but i need to replace the TL072 with something that has a bit wider power bandwidth first.

[sumaudioguy]

I noticed there is discussion of "complex speaker load" and would really like to add:

In stereo speakers the left speaker produces sound which drives the right speaker acoustically and thereby produces an electrical signal in the transducer which then is applied to the inverting input of the feedback amplifier. Loudspeakers are more than loads. Loudspeakers are sources of signals. Generators if you will.

Further, any RF around is picked up well by most speaker cables and likewise applied to the inverting input.

All these sources of noise and signal somehow must be controlled by the amplifier. Driving the output of the amplifier with either real or simulated signals and using the amplifier as a virtual ground is sound engineering practice with this thread adding many techniques to this sound practice. Of course one does not find many consumer amplifiers which are designed this way.

[jcx]

in domestic setups typical speaker electrical-acoustic efficency is order of 1% - so coupled electrical-acoustic-electrical efficiency will be <<0.1%

if you don't terminate the "microphone" speaker you can measure the terminal V directly - then dividing by the speaker's impedance give a estimate of the current the amp would have to supply to to keep the speaker terminals at the commanded V

the number is trivial for any ampifier with low Z/high damping factor - less than many nonlinear terms from the speaker's own inductance/motor/suspension caused nonlinear currents


again Cherry's articles shows that distortion for these external error currents are equal or lower than the distortion from the forward signal causing the output stage to put out the same current - for low output impedance/high feedback factor amps their is no issue for any current within the output/supply's I,V limits

[sumaudioguy]

It is not gnats if you throw in room modes and consider woofers. I have seen over 0.5 volts come out of an inactive driver driven by the other channel. Further, many amps will pass a lot of current but make extremely lousy virtual grounds. To simplify, at the output terminals low output impedance and an amazingly high input impedance. Now I know this sounds out of balance and that is exactly how many amplifiers and other power supplies behave.

[jcx]

still gnat sized - 500mV/8 Ohm ~= 60 mA which is about 1/2 the Class AB bias current per output Q recommended by just about everyone excepting perhaps Doug Self

at low frequency a high feedback SS amp can have uOhm output impedance

my sim above is showing <10 uOhm output Z at 7 kHz - admittedly Class A baised - but not actually what I would call a high feedback amp


the thread is in the Solid State amp forum, the topic is testing Solid State amp designs, specifically those designed to approximate Vsource outputs

there is no need to get excited over "back driven by the room sound" effects on such amps - at SPL compatable with human life

[sumaudioguy]

JCX- believe what you want. Your "SIMs" are almost a complete waste of time and worthless unless that sim is accurately calibrated to a REAL circuit. I have the lowest output impedance amps made and none are micro-ohm scale in the output impedance which shows clearly your statement to be false. Go test a REAL amplifier and not some imagination on a computer and discover the FACTS and not your wishful sim. I do not happen to listen to simulated music on virtual or simulated speakers. In short, pretty much everything you say is incorrect. At least you are consistent. Further, I was talking input impedance at the output terminals and you are talking output impedance. Not the same thing in reality at all.

I would like to take this moment to say to everyone with a sim not calibrated to a real circuit with real result, you all mislead yourself and many others with your nonsense notions of how amplifiers work and performance of those circuits. All of you sim people need to build you "sim" with real parts and find out just how far off those sims are from reality. Indeed, I believe DIY audio should have some method for SIM heads to be segregated from people who actually build and test real circuits as sims are pretty much straight fantasy until accurately calibrated where a real circuit just cannot lie.