Review and Measurements of vintage Yamaha AX-396 integrated amplifier

[pma]

Large signal step into 4ohm

.... and amplitude and phase response into 4ohm at 25W

 

[pma]

Measurement into complex speaker simulating dummy load

This measurement was made into my dummy load that is used to simulate a simple 3-way speaker. I did the measurement as I can see a lot of misunderstanding regarding complex loads vs. resistive load effects.

First, this is the measured impedance of my dummy speaker load

We can see that the impedance magnitude never falls below 5ohm, up to 20kHz.

So, let's make a comparative measurement of the amplifier under test into 4ohm load and this dummy load, which never falls below 5ohm. This is the result:

Interestingly enough, though the dummy load never goes below 5ohm, the distortion of the amplifier under test is from 90 to 180Hz and from 1600 to 3000Hz higher into dummy load than into the 4ohm resistor. Why? Look at the impedance phase in the plot above this one. It is the combination of relatively high capacitive phase angle and impedance magnitude that makes the difference. Even if impedance magnitude is quite higher between 1600 and 3000Hz than those 4ohms, it is about 7ohm.

Edit: one can see that even with this light dummy load (>5ohm) the output pair SOA is exceeded at 80Hz and near this frequency

 

[restorer-john]

@pma what is your square wave source? You're not using the DSO cal point again are you?

 

[pma]

@pma what is your square wave source? You're not using the DSO cal point again are you?

DSO was used only for the reason it was "just there". As for square wave sources, I use either GAG-810 Audio Generator
https://www.gwinstek.com/en-US/products/detail/GAG-810_GAG-809
https://www.gwinstek.com/en-US/products/downloadSeriesSpec/555

or, if I need faster rise time, my own built generator based on MAX038, that goes up to 20MHz

 

[restorer-john]

DSO was used only for the reason it was "just there".

I figured as much.

The reason I mentioned it, is the shape is essentially identical to my DSO's cal test point and when I looked really hard at mine, I realized my DSO generates a really poor square wave. Not that I've ever used it for anything other than probe x10 comp. Not only is it slow (rise time) it's nowhere near as fast and symmetrical as the cal point terminal on an old 20MHz analogue CRO on my bench.

I think your DSO cal point is the limiting factor, not the amplifier. Show the same square wave from your faster gen or the direct cal terminal vs the amplifier on the DSO? Your amplifier may be better than you think.

Cheers.

 

[pma]

@restorer-john , you did not get it. Yamaha step response here above was measured with GAG-810 generator. I mentioned the calibrator for the reason it was on one of my older photos (I thought you were commenting on the photo, not on the step response here above). Please do not underestimate that much . If you looked at amplitude/phase plot, you would see 4-5 us rise time fits to amplitude decay at 40kHz. Please also note that the rise time somewhat depends on output amplitude (pot impedance acc to setting + capacitances) and the plot has 48Vp-p. Things are not straightforward and clear at 1st view .

 

[restorer-john]

Glad to see the Yamaha got a decent rise time to start with from the GAG-810. <200nS

 

[Cahudson42]

would appreciate if it had lower distortion in 1 – 5W range and less high order harmonics in the spectrum

Very interesting - Thank You!
My question: Many of these Yamahas seem to have a crude form of bias adjustment - a 1k resistor is paralleled across (as I remember) an approx. 300ohm. You are to 'clip out' the 1k if voltage exceeds a certain mv. across a test point, or 'unclip' (if already clipped) if mv. is less..

Do you think it might be possible to improve distortion in the lower-wattage range by driving it more into class A by increasing the bias? Say by replacing the 1k resistor with a series resistor of a 500ohm fixed and 500ohm pot?

This at least would allow you to more perfectly balance the channels, as well as increase the bias..

How much? Do you know what the actual bias current is - as factory set by the 2 fixed resistors? My Rx-396 seems to run quite cool as shipped..Doesn't seem increasing bias a bit might hurt?

Thoughts?

 

[pma]

Do you think it might be possible to improve distortion in the lower-wattage range by driving it more into class A by increasing the bias? Say by replacing the 1k resistor with a series resistor of a 500ohm fixed and 500ohm pot?

I am absolutely sure that optimising bias (idle current) would improve distortion at lower power. Optimal bias is, as explained e.g. by Douglas Self
https://www.eetimes.com/distortion-in-power-amplifiers-part-iv-the-power-amplifier-stages/#
defined by Vq voltage thus by voltage across output stage emitter resistors.
AX-396 uses 0.22ohm Re resistors. Crossover distortion is minimised if voltage drop across one Re resistor is 23mV, i.e. idle current of the output pair is about 105mA. Considering 2x46V supply voltage, idle power of the output stage would be 9.66W. Yamaha AX-396 heatsinks would be too small to handle this continuous power. This is true for almost any cheap consumer class amplifier. Yamaha engineers of course knew that and they wrote in their service manual that voltage drop across the emitter resistor must not exceed 10mV.
The 2nd issue is that the amplifier has not very good idle current temperature stability and heatsink temperature thus affects the idle current. Another pain of similar amplifiers and another reason why they keep low idle current, so the output pair is underbiased, so it produces excessive distortion especially at low level.

So yes, you are right, however optimising idle current is not possible with this amplifier. My amplifier with similar power 100W/4ohm has heatsinks like shown below and you can run it hours at any power without going to thermal runaway issues.

 

[Cahudson42]

optimising idle current is not possible with this amplifier.

Thank you, @pma!
I am reminded English expression: 'Perfect is the enemy of the Good'..

If we can't provide the optimum bias, what is the amount we 'could' provide - and what would be the effect? For example, how non-linear is the distortion as a function of bias current? As we increase from 10 mv to 23 mv, does going to, say, 17mv reduce distortion, relative to 10mv, by 50%? 10%? Or?

We could also add external 'sucking up' quiet fans above the heatsink. I use a couple USB powered ones on a cheap Onkyo AVR... don't hear them at all..

In short, is any gain (reduction) in distortion going to be worth going to the effort to increase bias 'as much as we can', perhaps with forced cooling?

 

[pma]

I think even permitted 10mV across 0R22 would be OK, it would make 45mA idle current. I think that the amp has now much lower bias - the service manual says 0.1mV - 10mV is OK - which is of course too big tolerance field re crossover distortion. I have not measured the idle current yet, I will do it and try to set it close to 10mV. I guess now it is about 1mV.

P.S.: I have the amp in simulation now

 

[pma]

Simulated distortion for output stage idle current 10mA and 40mA. See rising distortion at lower power for too small idle current (10mA).
Distortion axis (Y) 1m = 0.001%, 10m = 0.01% etc.

Edit: added plot for 1mA idle current, which is still in the service manual limits. We can see how grossly the idle current affects the low level signals distortion, though it is seemingly hidden in amplifier or measuring system noise. However, we are able to hear sounds below the overall noise level, so this may make quite a difference in sound.

 

[Cahudson42]

Simulated distortion for output stage idle current 10mA and 40mA. See rising distortion at lower power for too small idle current (10mA).
Distortion axis (Y) 1m = 0.001%, 10m = 0.01% etc.

It would seem a nice improvement in what is most probably the usual listening power range of 1-5w. Great job - thank you! And only for the cost of a few resistors/pots..

But it leads to another question: Is there a more modern pair of final output complementary pair transistors that could be used in place of the existing ones? Higher wv/current capability so you can use the higher rail voltage safely at 4ohms? Perhaps ones with higher-efficiency/faster switching speed?

They look easily replaceable..

 

[pma]

Yamaha AX-396 bass + treble control, set to flat and to maximum values

 

[restorer-john]

Rarely does anyone use extreme maximum boost or cut values together. It certainly may look nice, but not a realistic operational metric is it?

The smaller incremental adjustments in combination (+/-) are also more useful, along with filter functions.

 

[pma]

Rarely does anyone use extreme maximum boost or cut values together. It certainly may look nice, but not a realistic operational metric is it?

The smaller incremental adjustments in combination (+/-) are also more useful, along with filter functions.

Yes, and IMO we can imagine what is between the extremes

However, John, there is a big issue with the simple tone controls circuits, in general. Not addressed by others than audiophiles, usually, and they are right on spot. You know, audiophiles have often complained that the tone control circuits, even if set to "flat", change the sound and sound worse than if they are bypassed (= direct mode). And I do support this opinion and I am bringing technical proofs.

1) Comparing "flat tone control" vs. "direct" mode, direct mode has of 15dB better S/N in this amplifier (shown earlier) and also less hum/buzz components. This directly reflects in higher usable dynamic range.

2) The "flat tone control" mode introduces excessive distortion. Yes, audiophiles used to say so, but I am bringing the proof. Below is the measured comparison of THD vs. frequency of the "flat tone control" vs. "direct" mode, at 10Vrms output voltage. Amplifier unloaded, so the effect of crossover distortion is eliminated. One can see strange THD hump due to tone controls circuit insertion. This was repeated at many output voltage levels and pot settings. No clipping here. The result was always the same. Tone controls circuit considerably increases distortion of the amplifier.

Green = flat tone controls, Grey = direct mode

 

[restorer-john]

there is a big issue with the simple tone controls circuits, in general. Not addressed by others than audiophiles, usually, and they are right on spot.

Simple tone controls definitely do cause distortions, especially the inexpensive integrated amplifiers which put the tone controls in the power amplifier NFB loop and there are lot more of them than you would think.

We can't lump them all in the bad category, but even some TOTL models from highly respected manufacturers make mistakes in the tone circuitry that should never have seen production. But not all "direct" functions are well implemented either. You only find out what is wrong by comprehensively testing them hey Pavel?

I've got one for you, I'll dig out.

 

[pma]

Simple tone controls definitely do cause distortions, especially the inexpensive integrated amplifiers which put the tone controls in the power amplifier NFB loop and there are lot more of them than you would think.

You only find out what is wrong by comprehensively testing them hey Pavel?

Not only by comprehensively testing, John, but also by listening, always, as mentioned in the post #1 .

Regarding traditional tone controls circuit, I am quite sure that with a proper design and parts selection even such circuit may be well workable, with low distortion.

 

[xr100]

However, John, there is a big issue with the simple tone controls circuits, in general. Not addressed by others than audiophiles, usually, and they are right on spot. You know, audiophiles have often complained that the tone control circuits, even if set to "flat", change the sound and sound worse than if they are bypassed (= direct mode).

A problem is that the "unity" setting may not measure flat, so there might be a frequency response change to take into account for perceived differences.

On my old Arcam Delta 290 amplifier, switching off "DIRECT" mode would result in a very obvious increase in the noise floor.

Back in "pre-Internet" days when I used to read the hi-fi press, it would be suggested that tone controls were "bad" due to additional components in the signal path and "phase shift"--never mind that the vast majority of contemporaneous recordings were tracked/mixed on a desk with parametric EQ on every channel* (not to mention VCAs!), and using tone controls to "correct" the frequency response would of course mean "correcting" the phase...

(*Granted, it may well have been defeatable, but suffice it to say, they were there to be used... not to mention "outboard" gear...)

Talking of which, basic tone controls are fairly useless at "correcting" anything; it's quite unlikely that the shelving curves available would be exactly what is needed. Possibly, they might work to provide a "loudness"-type curve at low listening levels. These days, they could be considered somewhat anachronistic, when comprehensive ultra-high performance DSP options are available.

 

[xr100]

Simulated distortion for output stage idle current 10mA and 40mA. See rising distortion at lower power for too small idle current (10mA). [snip...] Edit: added plot for 1mA idle current, which is still in the service manual limits. We can see how grossly the idle current affects the low level signals distortion, though it is seemingly hidden in amplifier or measuring system noise. However, we are able to hear sounds below the overall noise level, so this may make quite a difference in sound.

Measurements on my own Yamaha A-S501 (not under any particular conditions, it had been on for a while) were ~1.3mV and ~1.9mV. Neither of the "cut-off" resistors are open.

Hmm. It would therefore seem to be a good idea to raise the "idling current" level.

Being that I have zero experience of modding gear (but have some experience of soldering etc.) and am not keen to remove the main board from the A-S501 (given that it is still a current model), the most expedient way of doing this would seem to be to, say, parallel a resistor across each "cut-off" resistor?

Left channel "cut-off" resistor R139 - 270Ω, connected across R132 - 1kΩ -- so the "stock" range is between ~210Ω-1kΩ -- question is, what sort of values to try? (Or is a potentiometer really needed? That would ideally be something to try once I've got the above working...)

(EDIT: Oops, that's the wrong way around -- R139 (cut-off) is 1kΩ, R132 is 270Ω.)