Review and Measurements of vintage Yamaha AX-396 integrated amplifier

[restorer-john]

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.

There's so many different ways tone controls were implemented and also the hardware the manufacturers used.

Some tone circuitry is completely bypassed at 0 due to the pot wiper having a small "dead spot" of direct metal contact.
Some used switched position wafers and the tone is bypassed at 0.
Others use normal pots and put in a tone bypass function which skirts the entire tone amplifier stage.
Others put the tone in the amplifier feedback loop.
Most bypass/pure/direct functions offer slightly better performance, but not always. Usually, the direct switch is deliberately louder, so people perceive it's "better"...

The quickest and easiest way to see whether the tone controls are affecting the amplifier response is viewing a square wave. If you want to see what tone controls do to a pristine square wave, you'll never use them again.

 

[Cahudson42]

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

I'd check the value of the 'cut off' resistor - it makes more sense for it to be 1k and the 'fixed' resistor to be 270 ohm.. I personally would replace - if it's 1k - with a series pair of a multiturn 500ohm and fixed 500 ohm - with the multiturn preset to 500 ohm.

The reason for the series pair is to prevent 'fireworks' should you accidently set the multiturn to zero, or if it fails with a short...but I would like that proposed solution confirmed ok by @pma or another of our hardware experts before you implement..

Using a pot both let's you adjust the magnitude of the idle current, and balance out that current so it is the same in both channels..

 

[xr100]

I'd check the value of the 'cut off' resistor - it makes more sense for it to be 1k and the 'fixed' resistor to be 270 ohm.. I personally would replace - if it's 1k - with a series pair of a multiturn 500ohm and fixed 500 ohm - with the multiturn preset to 500 ohm.

Thank you--you are correct, my mistake, the cut-off resistor is indeed 1kΩ.

 

[xr100]

There's so many different ways tone controls were implemented and also the hardware the manufacturers used. [snip...] Most bypass/pure/direct functions offer slightly better performance, but not always.

Would you care to enumerate any examples of "slightly better" performance?

The quickest and easiest way to see whether the tone controls are affecting the amplifier response is viewing a square wave. If you want to see what tone controls do to a pristine square wave, you'll never use them again.

Do you mean as a result of the tone control's implied transfer function, i.e. impulse response, or because of "misbehaving" circuitry?

As I intimated in the previous post, for the most part, I wouldn't touch tone controls with a barge-pole.

64-bit floating point IIR or FIR filters on the other hand...

 

[pma]

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.

.

That's right, however the tone control circuit in AX-396 introduces additional noise and additional distortion, which is not a case or issue of a perfect/imperfect flat amplitude response. You may change bass controls around flat to +/- and distortion is still there. Interestingly, less distortion with bass increased than in a position near flat. Once again, nothing in common with perfectly flat setting. It is a non-linearity of the tone control circuit, non-linear means creating frequency components that were not present in the input signal, in other words, distortion.

 

[restorer-john]

Would you care to enumerate any examples of "slightly better" performance?

The 'slightly' better performance usually is a result of:

a) lower noise from removing (often) active componentry from the signal path
b) flatter response
c) lower distortion

Sometimes, the bypass/pure/direct switch makes things worse due to impedance mismatches, cable routing, noise introduction and/or loss of separation.

Here is an actual example of a tone defeat switch on an integrated amplifier making the frequency response worse. It was so unexpected, I checked the schematic and the switch contact wiring to ensure I wasn't imagining it. Ran the plots several times for sanity. The flattest response was achieved through the active tone stage.

The orange (top plot) is tone defeat engaged. The purple and yellow are offset channels (deliberately) through the active tone stage. The bottom plot (white) is low and high filters engaged.

 

[xr100]

That's right, however the tone control circuit in AX-396 introduces additional noise and additional distortion, which is not a case or issue of a perfect/imperfect flat amplitude response..

I understood that; I just meant that it was another possible mechanism (generically, not just in relation to the AX-396) for reported differences with "direct" mode engaged. Obviously, the frequency response could be matched (but that surely would be atypical for "audiophiles?")

 

[restorer-john]

I understood that; I just meant that it was another possible mechanism (generically, not just in relation to the AX-396) for reported differences with "direct" mode engaged.

Since tone defeat/dire/pure direct etc switches were introduced, the gain staging (in the amplifier/preamp) have been generally arranged to ensure the "direct" button provided a louder (=better) signal.

A notable exception I can think of (owning a few of them) is the Harman Kardon PM-660's main direct which is -7dB below the non-direct position.

 

[xr100]

Since tone defeat/dire/pure direct etc switches were introduced, the gain staging (in the amplifier/preamp) have been generally arranged to ensure the "direct" button provided a louder (=better) signal.

In the case of the Arcam Delta 290 integrated amplifier that I mentioned as having a far inferior noise floor performance with "DIRECT" disengaged...

With "DIRECT" engaged, the preamp is in fact passive, with the input signal passing to the source selection switch, then the volume control potentiometer, and straight to the power amp.

Disengaging "DIRECT" adds of 2 stages of NE5534's (post-volume control pot,) the first being a buffer, the second used for the tone control filters.

So, other than the raised noise floor, perhaps the performance would actually be better with "DIRECT" disengaged; it all seems like a terribly source output stage dependant arrangement.

(Rear panel pre-outs are available which are in fact buffered by a TL072, so it's not as if their "op-amp budget" ran out...)

 

[lopchu]

Dusted off (slightly!) my old AX-396. Would be interesting to know how much better (or if better at all) a e.g. JDS Atom sounds. Source is a RPI3 with Moode Audio.

 

[Cahudson42]

I've an RX-396 - similar, but not as good heatsinking. It's driving a $160 pair of Infinity R162 (thanks to tipoff here at ASR) - works great for that.

But is not replacing my Liquid Spark for HP. Almost certainly, your headphone jack - like mine - is simply fed by drop-down resistors off the main amp finals. Not by its own amp. This presents a high output impedance which is not optimal for your phones. Likely thin bass, maybe noise and distortion, compared to Atom

 

[AnalogSteph]

Note that this model (and the '496 and '596) has two "Direct" functions:

• PURE DIRECT, which bypasses the tone control section (and clearly for the better)
• CD DIRECT (CD input only), which reshuffles gain to before the volume control on top of that, and should yield very low output noise as a result.

@pma's Direct measurement would seem to have been the first. Noise level should drop several dB more in CD DIRECT (spec is 35 µV(A)).

Ironically, the CD input may be best used with anything but CD players and their ilk, as the poor little NJM2068LD that makes up the +10 dB amplifier has to deliver 7.1 Vrms for a 2Vrms input (from a nominal +14.2/-14.3 V), while loaded with about 1.64 kOhms. The O2 gain stage has shown that the part will break a bit of a sweat under such conditions, showing increasing crossover distortion in the highs. Plus, this stage is just super low noise for no good reason - its effective input noise level (0.65 µV!) would be equivalent to 129.7 dB of dynamic range ref: 2 Vrms. Absolutely nobody needs that in a high-level source, and only the best DACs even get into this region.

Some improvements I can therefore think of:

• Plan A, turn the stage into a unity-gain buffer by removing R127, R129 - CDs have gotten much louder over the years anyway, so you're not likely to run out of gain any time soon. Those with more sensitive speakers might appreciate having their volume control in a somewhat sensible range for once, and the opamp will have an easy life.
• Plan B, replace NJM2068LD (IC102) by NJM5532L. (Not too much choice among SIP-8 case parts.) Dynamic range has now shrunk to a measly 127.7 dB, boohoo. Distortion should be better in return.
• Plan C, rescale the feedback network to be an easier load, e.g. R127/129 = 1k, R125/130 = 2.2k (all metal film), C121/123 = 47p (film MKP or ceramic C0G/NP0). Estimated dynamic range = 128.7 dB still. HF distortion is expected to go down by up to 6 dB.

The Volume Amp (IC107) shares a similar design with 510R/220R (I'm not really sure what they were expecting out of their 100k volume pot - I mean 40 dB down = 1k, 46 dB down = 500R). May be easier to get away with at the low levels encountered there - with the 150 mV input sensitivity this only has to output 0.5 Vrms max. Clearly a much less stressful position. Distortion is probably swamped by the power amplifier entirely.

Note that the AX-396 and 496 share much of the same circuitry and rectifier, but the former shipped with 6800µ/56V reservoir capacitors and the latter with 10000µ/63V. For full power bandwidth into 4 ohms, at least 8200 µF is recommended. An upgrade should be quite feasible.

@pma's AX-396 is likely to need its speaker relays cleaned (or replaced). These are, unfortunately, to be considered expendable in this position. Standard relays take a lack of current long-term about as well as standard electrolytic capacitors take being shorted, which is to say not very.

 

[Interference]

This thread is extremely instructive, thanks to all the contributors. It's interesting to see how achieving the best performance out of a setup is not trivial.

As well, it's interesting to see which kind of compromise goes into building this kind of product: I have always wondered what can make an amplifier "better" compared to this traditional class AB design from a reputable manufactured. Idle current and heat dissipation are part of the answer.

I am measuring the AX-497 in my possession.

 

[Interference]

The amplifier performance is very good. Some mains noise but overall SINAD goes nicely beyond 85 dB.

I noticed gradually increasing distortion on one channel, not with a sharp step as shown here but more of a smooth curve. However it disappears when switching to output "B".

Dear @restorer-john and @pma , would you kindly care to elaborate on the "relay" issue? Can it be solved with some easy maintenance?

 

[pma]

In my experience, output relays are the biggest issue of the ageing power amplifiers. There is no simple cure other than replacement by the new relay. Contact material and current rating will say if the relay will deteriorate sooner or later. Every break of DC failure stresses the relay contacts.

Few years ago I lost my patience with mechanical relays and in my designs I have replaced them by solid state MOSFET realys.

 

[Interference]

In my experience, output relays are the biggest issue of the ageing power amplifiers. There is no simple cure other than replacement by the new relay. Contact material and current rating will say if the relay will deteriorate sooner or later. Every break of DC failure stresses the relay contacts.

Few years ago I lost my patience with mechanical relays and in my designs I have replaced them by solid state MOSFET realys.

Thanks for the answer. Luckily mine is perfectly fine on output B. Guess replacing the other realy would not be worth the time needed to take everything apart.

 

[P_M]

Edit: one can see that even with this light dummy load (>5ohm) the output pair SOA is exceeded at 80Hz and near this frequency
View attachment 43340

View attachment 43341

Hi @pma
can you please explain these SOA charts how to interpret them ? I can get some of it but not quite sure. Here's my long list of questions -

I suppose SOA means safe operating area marked by the blue lines ? but I dont follow what the multiple areas of blue lines on the same chart mean. Multiple SOAs ? didnt quite get that.
Are the axes the voltage and current of one of the output power transistor ? If so why are both charts showing different soa areas/blue lines ?

And is the red line the actual measurement ? Meaning at 20Vce and 50Vce for 80hz it was outside the soa ? At what power level does that happen ?

 

[AnalogSteph]

The red one is transistor datasheet SOA limit, I think. (Not sure which one, would have to look it up - you usually get a whole range of curves, from steady state through very short-time pulses.)

I'm not sure I would fully trust simulation for this kind of thing - it wouldn't be accounting for power supply sag for one. The AX-396 doesn't exactly sport the world's biggest power transformer.

If you really wanted to know, you'd probably have to measure actual transistor Vce (from test tone onset to steady state(, and judging which SOA curves to use in each case. Not a trivial task.

 

[pma]

Yes, the red one is a relevant part of the transistor datasheet SOA (Safe Operating Area) curve. This defines maximum Ic current at the defined Vce transitor voltage for a defined period of time. I think I took a 1s curve. So the collector current must be below this red curve at the Vce chosen. Vce and Ic differ with signal and load.

In my example, I used the simulated speaker load as attached. The amp input signal was a 80Hz sine turned on at time T=0. The blue plot is a trajectory of amplifier's output power transistor with respect to Vce, Ic (into the complex load) and time. It will settle as an ellipse, after the initial transient is over. Everywhere the blue curve is above the red curve, there is a potential risk of the output transistor breakdown. This is usually fixed by the amplifier's current protection circuit, which should follow the shape of the SOA curve. As we could see in this review, the protection circuit has shut down the amp several times. This is a price of the cost-effective amp design (only 1 pair of output transistors, which is insufficient re complex load and DC supply voltage used).

 

[gino1961]

Review and Measurements of vintage Yamaha AX-396 integrated amplifier
This review shows measurements of Yamaha AX-396 integrated amplifier, ser. no. Y399871TV, made in Malaysia....
Yamaha AX-396 output noise A weighted.
It is -81.6dBV(A) and this is very good, the only pain is the PSU ripple throughput. S/N to full power (100W) is 107.6dB(A).... .

Hi ! thank you very much indeed for the excellent test. Looking inside i see two quite small psu caps ... i wonder if replacing them with two higher in grade and uF could lower the 50Hz spike in noise ... and its harmonics. That little issue aside the amp is very impressive.
Thanks again, gino