Audio… (Amps, THD, channel separation) with analog equipment (scopes, etc.)

[Fried Chicken]

I’ve gone deeper into audio than is healthy; I know I’m not the only person who’s been afflicted by this.

I recapped an old amplifier (Yamaha M-65) and removed some corrosive glues.  I want to go in and actually test this work: channel separation, signal clarity, stuff like that.

My equipment to achieve this is limited: analog 2-channel scope, analog function generator, Fluke DMM, a smattering of coax adapters, and a couple power resistors.

The power resistors, I’m not sure can handle the full power of the amp.., I think I have two 4 ohm resistors, but the speakers I run are 8 ohm.  They are expensive!

Anyway, I’m just fishing for advice here, open to suggestions on what to do, what not to do.

I’ve heard claims that capacitors need to “break-in”; I’m not sure I believe that.  I can set the bias voltages (to center the A vs AB amplification I think this is?  What the signal rectified around?).  That and calibrating the VU meters is about all that’s offered in the service manual.

I think it’s here:
https://elektrotanya.com/yamaha_m-65_sm.pdf/download.html

Anyway, this is a fun project for me.  If anyone has any especially wild tricks, like figuring out the THD with all analog stuff, that’ll perk my ears!

Thanks in advance

[CaptDon]

Be aware of the subjective bullshit observations of 'Audiophools'!!! Remember that in a very real double blind test 12-2 NM house wire beat every other test sample wire including idiotic $30.00 per foot stuff you could use as a mooring line. And then the audiophools still insisted there must be a flaw in the testing because their expensive stuff MUST be better!!! Stop and think, how would you measure channel separation? Drive one channel to maybe 80% of full output and look for residual output on the other channel. Then calculate how many dB from the voltage ratio. THD is nearly impossible to measure with the gear you have. Audio THD meters are a specialized instrument. Look up the Potomac Instruments 'AA' audio analyzer or the older HP audio analyzers. Measure the THD of your generator and then measure the output of the amplifier. Cheap function generators and audio oscillators have terrible THD to begin with. As for 'A' vs. 'AB' vs. 'AB1' vs. 'AB2' setting of the bias, a lot of it comes down to "What sounds best to you at your listening level". So much bad information among tube amp guitar players!!!! How many amps I have seen burned up because some jackass said "set the bias to where the plates just begin to glow red". W.T.F., the whole idea of class B push pull is that the outputs really don't need to idle more than about 10% up the conduction curve. These red plate chuckle heads are around the class A push pull point of operation!! Audiophools think little wooden wire supports help the sound and expensive wall receptacles "Give a more open sound stage". Idiots!!

[Fried Chicken]

I appreciate it…

I was really really hoping not to encounter pro- or anti- audiophile bullshit here, and is a reason I asked here and not on audio forums.

Yamaha gives specific bias voltages in the manual, that’s what I intend to set them to.

How do I even know what sort of signals to apply to the amp?  Should I use the front meters?



I can find the difference between two signals on a scope by inverting one and adding them, right?

[CaptDon]

Many of the '50dB down' artifacts may be approaching the noise floor of your scope and test equipment. Many 'purpose built' audio analyzers can do THD, intermod IMD and other tests but often require the companion audio generator. Tektronix in their TM5000 (not the 500) series of modules had some audio distortion modules with matching generator. They still fetch big bucks on Ebay and NEVER buy from a seller who sells test equipment and lists something as 'untested'!!! That means they already tested it (after all, they are a test equipment dealer) and it is blown to bloody f--king hell!! I hate those crooks!!! May as well buy one laying in a bucket of rain water at a hamfest where it rained all day!! Anyway, best wishes in reaching your goals! Invest in good test gear as it will hold its resale value. Cheap junk won't!!

[Fried Chicken]

Surely something can be done with a normal scope?  I mean the scope samples at several orders of magnitude above audio signals?

Are there no analog tricks or similar?

[CaptDon]

THD and IMD are specialized measurements requiring filters and accurate clean frequencies incorporated inside the audio analyzers. I forgot about the Tektronix AA-501 audio analyzer for the TM500 series racks. Just looked at two on Ebay $500 and $840. If I were still making a living doing audio repairs I would consider that but I am retired these days from electronics. Still Master of a 140 passenger tour boat, guitarist and ham radio operator keeping me busy and somewhat sane in my 'golden years'.

[vk6zgo]

Surely something can be done with a normal scope?  I mean the scope samples at several orders of magnitude above audio signals?

Are there no analog tricks or similar?

Well, you can measure channel separation (aka as "crosstalk" in comms), Power output, pretty roughly with the 'scope, although DSOs do have a voltage readout), maybe guess on Noise, but it will be down near the bottom of the scope's range, or depending on the DMM, that may work.

If your 'scope has FFT, perhaps you could maybe.use that for THD, but what you really need is either a "Noise & Distortion Test set", which incorporates a sharp tunable filter, which you can use to notch out the fundamental frequency leaving the harmonics (plus noise), or as an alternative, a selective voltmeter, which can show you the levels of each of the harmonics.

There should be quite a number of N& D sets around, though they still maintain their price.
Selective voltmeters are somewhat rarer.

[Fried Chicken]

Surely something can be done with a normal scope?  I mean the scope samples at several orders of magnitude above audio signals?

Are there no analog tricks or similar?

Well, you can measure channel separation (aka as "crosstalk" in comms), Power output, pretty roughly with the 'scope, although DSOs do have a voltage readout), maybe guess on Noise, but it will be down near the bottom of the scope's range, or depending on the DMM, that may work.

If your 'scope has FFT, perhaps you could maybe.use that for THD, but what you really need is either a "Noise & Distortion Test set", which incorporates a sharp tunable filter, which you can use to notch out the fundamental frequency leaving the harmonics (plus noise), or as an alternative, a selective voltmeter, which can show you the levels of each of the harmonics.

There should be quite a number of N& D sets around, though they still maintain their price.
Selective voltmeters are somewhat rarer.

Yeah; I'm not getting into all of this.

I guess I should have rephrased the question: what *can* I do with my basic equipment?  I'm not against rigging up some clever circuit, but I'm also not wanting to go to unreasonable levels

[edit]

I found this nice playlist on youtube by a guy named "Blueglow Electronics":

[vk6zgo]

Surely something can be done with a normal scope?  I mean the scope samples at several orders of magnitude above audio signals?

Are there no analog tricks or similar?

Well, you can measure channel separation (aka as "crosstalk" in comms), Power output, pretty roughly with the 'scope, although DSOs do have a voltage readout), maybe guess on Noise, but it will be down near the bottom of the scope's range, or depending on the DMM, that may work.

If your 'scope has FFT, perhaps you could maybe.use that for THD, but what you really need is either a "Noise & Distortion Test set", which incorporates a sharp tunable filter, which you can use to notch out the fundamental frequency leaving the harmonics (plus noise), or as an alternative, a selective voltmeter, which can show you the levels of each of the harmonics.

There should be quite a number of N& D sets around, though they still maintain their price.
Selective voltmeters are somewhat rarer.

Yeah; I'm not getting into all of this.

I guess I should have rephrased the question: what *can* I do with my basic equipment?  I'm not against rigging up some clever circuit, but I'm also not wanting to go to unreasonable levels

[edit]

I found this nice playlist on youtube by a guy named "Blueglow Electronics":

Oscilloscopes can be used to measure (for example), distortion , just by looking at the shape of a sinewave, but the distortion has to be quite profound before aberrations become visible.

You could compare the sinewave signal at the input with that at the amp output by using the "A" minus "B" function of an analog 'scope, or equivalent on a DSO, but you will run into problems with differing phase.

Any "tricks"  delivering usable results become increasingly complex to perform, as they require additional circuitry which you will have to design yourself

I earned "my daily bread" measuring distortion in Broadcasting for years, & later in testing hearing equipment, & believe me, if there were easy tricks to satisfactorily measure THD using a standalone Oscilloscope,my various employers would have used them!
The two "filter" methods I mentioned give repeatable results, so are used, the N & D set in Broadcasting, & the Selective voltmeter in the testing of equipment used in hearing testing.

An interesting alternative people have often suggested was that of using the Sound Card on a computer.
I have no experience with Sound Cards used in this manner, but there may be "how to" descriptions on the Internet.

[Kleinstein]

A twin T notch fitler for THD test can be build relatively easy. One may want some trim (2 pots or 1 pot and adjusting the frequency).
Builing a reasonable low THD generator is possible, but needs some knowedge and not all plans found hold the promisses.
For just hearing there is little need for THD much better then 0.1% - one can hardly hear 1%. For checking the amplifier specs it can get a bit tricky though.

Using a soundcard is a real option, especially as a low THD generator.
Quite some of the modern audio measurement systems are essentially a good sound card with a few input dividers and software.

[CaptDon]

There is a program called spectrum lab and several others, perhaps RoomEq that use your sound card and software FFT to show spectrum analysis of your signal. When I compare 'those' results to a real purely analog 'swept I.F.' style analyzer the results differ and there must be at least four types of FFT which all show different results. I think my Tektronix TDS644B scope has at least three FFT calculation algorithms. Anyway, the simple point being if you analyze the input signal to the amplifier and note its cleanliness and harmonic content and then analyze the output most forms of distortion show up as addition harmonics or elevated harmonic levels. If the sinewaves look good on the scope and sound good to your ear then you are in the ballpark. As you raise and lower the output level of the amplifier look for three important signs of problems. 1. Zero Crossing Distortion, Shows up as little flat segments where the trace crosses zero. Usually at low volume levels and they sound really bad!! 2. Flat Topping Clipping, especially if it occurs well below the max power output level. 3. Weird Kinks in Sinewave. This is sometimes found in so called 'Class G' type amplifiers where there are multiple voltage rails in the output stage and the BJT or MOSFET devices are 'stacked'. Many of the Bob Carver amplifiers are built this way and the transfer of voltage output level can see a small anomaly as the next higher device in the ladder begins to source current and voltage (P=I X E). Remember, even harmonics are musically pleasing, odd harmonics sound horrible and are very evident to the ear!!

[Fried Chicken]

This classic A/AB amp should reference to ground across everything, including the speaker terminals, right?

I can safely probe around bing bing bing bing bing?

[Fried Chicken]

Oscilloscopes can be used to measure (for example), distortion , just by looking at the shape of a sinewave, but the distortion has to be quite profound before aberrations become visible.

You could compare the sinewave signal at the input with that at the amp output by using the "A" minus "B" function of an analog 'scope, or equivalent on a DSO, but you will run into problems with differing phase.

Any "tricks"  delivering usable results become increasingly complex to perform, as they require additional circuitry which you will have to design yourself

I earned "my daily bread" measuring distortion in Broadcasting for years, & later in testing hearing equipment, & believe me, if there were easy tricks to satisfactorily measure THD using a standalone Oscilloscope,my various employers would have used them!
The two "filter" methods I mentioned give repeatable results, so are used, the N & D set in Broadcasting, & the Selective voltmeter in the testing of equipment used in hearing testing.

Well... what other "tricks"?

Does the A minus B work off the displayed signal or the fed signal?  Can I set the voltages so that the two signals match on screen and then do A minus B?  I can alter the phasing by altering the trigger, no?  Or do I need a tunable RC circuit to change the phase?

[vk6zgo]

Oscilloscopes can be used to measure (for example), distortion , just by looking at the shape of a sinewave, but the distortion has to be quite profound before aberrations become visible.

You could compare the sinewave signal at the input with that at the amp output by using the "A" minus "B" function of an analog 'scope, or equivalent on a DSO, but you will run into problems with differing phase.

Any "tricks"  delivering usable results become increasingly complex to perform, as they require additional circuitry which you will have to design yourself

I earned "my daily bread" measuring distortion in Broadcasting for years, & later in testing hearing equipment, & believe me, if there were easy tricks to satisfactorily measure THD using a standalone Oscilloscope,my various employers would have used them!
The two "filter" methods I mentioned give repeatable results, so are used, the N & D set in Broadcasting, & the Selective voltmeter in the testing of equipment used in hearing testing.

Well... what other "tricks"?

Does the A minus B work off the displayed signal or the fed signal?  Can I set the voltages so that the two signals match on screen and then do A minus B?  I can alter the phasing by altering the trigger, no?  Or do I need a tunable RC circuit to change the phase?

There may be tricks you can do with a DSO, but with an analog 'scope the RC phase change circuit would be the way to go.
The difference in shape of the two signals would show up as the result of A minus B, which may require some analysis to work out the relationship between that display & distortion in % (or dB).

It will, in fact be a result of THD + Noise, but usually the latter is low enough to be ignored.

In TV testing, we expressed both THD + Noise & Noise by itself in dB, making it easier to mathematically remove the latter to give pure THD.

[TimFox]

This classic A/AB amp should reference to ground across everything, including the speaker terminals, right?

I can safely probe around bing bing bing bing bing?

Usually, that is true.  It is the exceptions that might kill you, for example an analog amplifier with an H-bridge output.

[TimFox]

Oscilloscopes can be used to measure (for example), distortion , just by looking at the shape of a sinewave, but the distortion has to be quite profound before aberrations become visible.

You could compare the sinewave signal at the input with that at the amp output by using the "A" minus "B" function of an analog 'scope, or equivalent on a DSO, but you will run into problems with differing phase.

Any "tricks"  delivering usable results become increasingly complex to perform, as they require additional circuitry which you will have to design yourself

I earned "my daily bread" measuring distortion in Broadcasting for years, & later in testing hearing equipment, & believe me, if there were easy tricks to satisfactorily measure THD using a standalone Oscilloscope,my various employers would have used them!
The two "filter" methods I mentioned give repeatable results, so are used, the N & D set in Broadcasting, & the Selective voltmeter in the testing of equipment used in hearing testing.

Well... what other "tricks"?

Does the A minus B work off the displayed signal or the fed signal?  Can I set the voltages so that the two signals match on screen and then do A minus B?  I can alter the phasing by altering the trigger, no?  Or do I need a tunable RC circuit to change the phase?

Note that "A minus B" mode of an analog two-channel scope is not a perfect subtraction:  the tolerance on the analog combination can be comparable to the THD that you are trying to measure.
I do THD measurements (technically, THD plus noise) using an -hp- 339A analyzer that combines a low-distortion generator with a nulling filter and auto-scaling of the remainder to the input level.
The nature of the distortion products can be viewed by an analog output of the remainder to an oscilloscope.

[macboy]

With your equipment you listed, you can do some useful things.

You can determine the clipping point of the amplifiers by feeding the amp a sine wave (around 1 kHz, 1 Vp-p), and adjusting the volume up until you can clearly see the flat tops on the output sine wave, observed with the scope. Back off the volume control until the peak is just a little below that, and there you have the maximum unclipped power output. The distribution may already be a little high at this point even if the output isn't technically clipping. The load (e.g. 4 vs 8 Ohms) will alter the clipping point somewhat. Calculate power by P=V²/R . Obviously use a dummy load resistor for this, not speakers.

Now without changing the volume control, remove the function generator and short circuit the input jack (I have built shorted phono plugs for this purpose). The input is now zero. Measure the RMS AC voltage of the amplifier output. It will be low but won't be zero. Take the ratio of this compared to the full output power from earlier and you have the signal to noise ratio. Convert to decibels: SNR(dB)=20*log(Vmax / Vnoise).

You can repeat the SNR measurement at lower volume levels as well, but the noise will become increasingly difficult to measure accurately as it drops to mV levels.

As a rule of thumb, your multimeter's AC voltage reading is only sensible when it is at least 1% of full scale. So at 200 mV scale : 2 mV minimum input, regardless of how many digits it shows. Your fluke DMM should also have decent bandwidth covering the audio range, but you can verify that by feeding it a constant level from your function generator and sweeping through the audio frequency range. The AC RMS reading should remain relatively constant (remember +-10% difference is actually less than 1 dB).

If you have a computer with a decent sound card, you can measure distortion and other performance parameters with software such as RightMark Audio Analyser RMAA or others (some free, some are even free open source). With a scope and signal gen, you just can't do it, but you can see gross distortion such as clipping or zero cross distortion. The latter might result from a bias setting much too low.

[Fried Chicken]

There may be tricks you can do with a DSO, but with an analog 'scope the RC phase change circuit would be the way to go.
The difference in shape of the two signals would show up as the result of A minus B, which may require some analysis to work out the relationship between that display & distortion in % (or dB).

It will, in fact be a result of THD + Noise, but usually the latter is low enough to be ignored.

In TV testing, we expressed both THD + Noise & Noise by itself in dB, making it easier to mathematically remove the latter to give pure THD.

I feel like just looking at and dicking with the signal will start making things obvious to me.
Gah; I wish I had a more intuitive understanding of math, the relationship between Distortion and dB.  I used to be really good at this, but then I chose to study science instead of EE.

This classic A/AB amp should reference to ground across everything, including the speaker terminals, right?

I can safely probe around bing bing bing bing bing?

Usually, that is true.  It is the exceptions that might kill you, for example an analog amplifier with an H-bridge output.

Why would someone put an H-Bridge on an amplifier output?

With your equipment you listed, you can do some useful things.

You can determine the clipping point of the amplifiers by feeding the amp a sine wave (around 1 kHz, 1 Vp-p), and adjusting the volume up until you can clearly see the flat tops on the output sine wave, observed with the scope. Back off the volume control until the peak is just a little below that, and there you have the maximum unclipped power output. The distribution may already be a little high at this point even if the output isn't technically clipping. The load (e.g. 4 vs 8 Ohms) will alter the clipping point somewhat. Calculate power by P=V²/R . Obviously use a dummy load resistor for this, not speakers.

Now without changing the volume control, remove the function generator and short circuit the input jack (I have built shorted phono plugs for this purpose). The input is now zero. Measure the RMS AC voltage of the amplifier output. It will be low but won't be zero. Take the ratio of this compared to the full output power from earlier and you have the signal to noise ratio. Convert to decibels: SNR(dB)=20*log(Vmax / Vnoise).

You can repeat the SNR measurement at lower volume levels as well, but the noise will become increasingly difficult to measure accurately as it drops to mV levels.

As a rule of thumb, your multimeter's AC voltage reading is only sensible when it is at least 1% of full scale. So at 200 mV scale : 2 mV minimum input, regardless of how many digits it shows. Your fluke DMM should also have decent bandwidth covering the audio range, but you can verify that by feeding it a constant level from your function generator and sweeping through the audio frequency range. The AC RMS reading should remain relatively constant (remember +-10% difference is actually less than 1 dB).

If you have a computer with a decent sound card, you can measure distortion and other performance parameters with software such as RightMark Audio Analyser RMAA or others (some free, some are even free open source). With a scope and signal gen, you just can't do it, but you can see gross distortion such as clipping or zero cross distortion. The latter might result from a bias setting much too low.

These are some clever tricks.  I know the old analog guys had some wizardry that probably surpasses what the digital people can do.

I suspect my function generator doesn't have a linear output on a sweep... at least it didn't when I hooked it up to my scope, I think... low frequencies it was very high p-p, and then it went down as I went up the frequencies... I think.
Why measure with the multimeter and not the scope for the sweep?

I don't have a soundcard in my computer, but I think I might get one as my USB headphone amplifier thing from Behringer caused bugs in software (I suspect).


I look forward to getting some free time to hook up the amplifier.  I suspect once I start dicking around something obvious might appear.  I guess I should mention the model, it's a Yamaha M-65.  I recently did a recap on it, and while I heard a difference, I'm not convinced it was an improvement.  I feel like the right speaker is fuller and better, but only across certain frequencies.  I should have measured the caps before installing them... possibly I screwed something up, but I don't think so, I worked carefully.


If I want to measure a crossover, do I also need a dummy load?

[TimFox]

"Quote from: TimFox on Yesterday at 01:03:44 pm

    Quote from: Fried Chicken on Yesterday at 10:19:59 am

        This classic A/AB amp should reference to ground across everything, including the speaker terminals, right?

        I can safely probe around bing bing bing bing bing?


    Usually, that is true.  It is the exceptions that might kill you, for example an analog amplifier with an H-bridge output.


Why would someone put an H-Bridge on an amplifier output?"

To work with a low DC voltage.

[Fried Chicken]

"Quote from: TimFox on Yesterday at 01:03:44 pm

    Quote from: Fried Chicken on Yesterday at 10:19:59 am

        This classic A/AB amp should reference to ground across everything, including the speaker terminals, right?

        I can safely probe around bing bing bing bing bing?


    Usually, that is true.  It is the exceptions that might kill you, for example an analog amplifier with an H-bridge output.


Why would someone put an H-Bridge on an amplifier output?"

To work with a low DC voltage.

How is that different from simply flipping the polarity?  (Red --> Black, Black --> Red)?
How does an H-Bridge give a low DC voltage?




Dumb question: if I'm measuring/testing a crossover, do I need a dummy load?  Dummy load on all outputs or just on the output I'm testing?

[TimFox]

With an “H”, the load is located on the horizontal line between the two uprights, where each upright has a “pull-up” and a “pull-down” device.  Neither end of the load is grounded, so you can’t put the ground lead of the ‘scope on one side of the load.  The two sides move up and down with respect to the power supply, with 180 deg phase difference.
My “low DC voltage” comment refers to the power supply.  The peak-to-peak output voltage is twice the DC supply voltage, while the usual circuit gives a pk-pk voltage equal to the DC supply.
This is handy for annoying other drivers when blasting audio from a 12 V automobile supply.

[Fried Chicken]

I got my amp out of the system and hooked up.  I carefully set up my signal generator and at its lowest output it's within a safe regime, but still slightly high lol.

I turns out I have two 8 Ohm 200W dummy loads, so perfect!

I carefully got everything wired up, here's my setup (apologies for the out of focus picture):







Immediately I notice the right channel has a lower p-p than the left channel, which is unusual because I swear the soundstage is shifted to the right.  I can now start dicking around.  I'm just getting my feet wet here, so I'm being stupid cautious.

[Fried Chicken]

It's pouring rain outside (actually thunderstorming so I've had to pause my dicking around).

The left channel is definitely slightly stronger than the right channel, as reported by both the scope and the lights on the front of the amp (left rises first).  I swapped the scope probes as a test, and indeed, the left channel on the amp reports stronger.  I didn't measure the resistance of the loads I'm using, maybe I should do that.  This is weird because the soundstage is off to the right when listening.

I drove the amp into clipping, and.. oh boy those power resistors got hot fast. (I really am doing this for the first time)
The front of the amp reported somewhere between 30 and 90 watts into clipping.

The p-p voltages measured on the scope are ~110-120V since I have the probes attenuating 10x...  This doesn't sound right to me, that's almost line voltage??!







I guess this would make sense for power at a single frequency?

[Fried Chicken]

So now I'm dicking around and things are starting to get interesting.
I inverted the B channel signal and set the scope to add... and the difference was quite stark.  This difference changed significantly as output on the function generator was increased, but remained almost identical as the frequency was changed: it was like a sine-wave squishing and expanding apart.



I disconnected the power resistors and measured their resistance with the Fluke.
I came up with 8.22Ω and 8.20Ω so they're decently close to one another, then I swapped them around Left/Right, cut off a bit of wire and reattached the cleaned wire to the binding post yielding these new different results:





Weirdly, after cleaning and swapping around the resistors, the outputs look almost identical, Vp-p wise.  The behavior of the difference of signals remains the same (changes on output from signal generator, squishes with changes in frequency).


The probes are on 10x for this.  The I'm thinking it might not be stupid to open the amplifier and start probing to find where that difference of signal might be coming from, but first I think it would be smart to set the biases appropriately.



One final note worth adding (I am writing notes to myself as well as everyone else, I'm not expecting everyone absorb and interpret every detail on my behalf!), depressing the "Auto Class-A" button makes no difference in the signal.  None.  I tried at different power levels, low, high, different frequencies, measuring the subtracted signals, I saw zero change to the signal.

I think I'm going to open the amp up and set the adjustments according to the manual... One final thing I've noted, I need longer probe cables for my fluke multimeter!

[Kleinstein]

The observed 120 V peak to peak are some 43 V RMS. So still quite a bit lower than mains voltage. Higher power amplifiers, especially when made for 8 ohms can indeed provide quite high voltages. They may have a warning about potentially hazardous voltages at the speaker output.

The amplifier output is normally low impedance. So the difference in the loading should not be that relevant.

An amplitude dependent effect would point to a linearity problem - maybe also loudness enabled.