Recapping amps: facts, myths, legends etc

[Tony L]

Bias does seem pretty tricky to set as it seems to vary with temperature, how long the amp has been on etc. I tend to set just a little on the conservative side with the stuff I’ve done (Quad 303s, Onix, an old Sony) as I suspect it will rise a bit with the case assembled and the amp working. I tend to leave the amp powered up for an hour or two, connect the probes, give it a nudge, see where it is after 10-15 minutes and repeat the process until it looks stable a mV or two below the stated rating. If anyone has better advice I’m all ears as I’m no expert on this by any stretch!

 

[martin clark]

Nice safe approach Tony, esp. for amps that don't have the bias thermally-compensated (Naim - but it's a quasi-comp output stage, so can't be compensated even if you want to)

There's an article in the Reference area which shows the effect of bias adjustment on the Naim architecture, using a Nait2 as a model.
http://www.pinkfishmedia.net/forum/showthread.php?t=36057

As to Peter's pair of 180s experience - using LesW's excellent advice to aim for 36-38mA per amp channel total draw: note that since the NAP input gain and VAS stages are effectively constant-current, the 6-8mA under bias must have all come out of the output stage draw. No wonder it sounded different - it must have been running close to 0 bias current (full class B) - so lots of added crossover distortion, and well to the left of the distortion graph posted in my link above.

 

[peterm]

Thanks Martin for explaining why the newer amp sounded so much worse!
I've seen your post about biasing several times in the past via the accoustica link (and liked the "cheeky young rioja" addition to your test kit!) without thinking any more about it.
This explains why the "new" 180 had lots of shipping labels all over it and the fisher I bought it from appeared to have had it only 6 months, some of that boxed up under the bed. Most previous owners passed it on!
I note that Naim were just about to discontinue the 180 and replace it with the new look 200 in late 2001!

 

[Dowser]

Bias decides where in the performance envelope of your output devices they actually deliver sound to you. Only in my DIY hybrid valve SE mono blocks have I experienced lower sound quality by increasing it - general un-qualified rule; if transistor output, push it as high as you can. But note this can lead to instability, heat and magic smoke.

 

[Dowser]

And I should add - in my limited experience

 

[Tony L]

Is there an easy to grasp correlation between bias and DC offset at all? I suspect not as it doesn’t seem to impact that reading much if at all. Some of the old Japanese amps have a specific adjustment for that, I remember adjusting a big Marantz receiver but many don’t e.g. the Onix, Sony TA-1140 etc, so I guess you just have to live with whatever it is? FWIW the Onix is ok at about -15-18mV but the ancient Sony is rather high at about 35mV on one channel IIRC. The bias on that one was way high too, about twice what the service manual stated. I know DC offset is a good quick hint as to health of an amp (so I always check it before plugging anything into speakers I care about), and that distortion rises hugely the higher it is, but I don’t understand any more than that.

 

[martin clark]

Bias, and DC offset are two unrelated mechanisms.

DC offset essentially comes from the design of the input stage (usually a long-tailed-pair), the degree of 'matching' of the two devices involved, in combination with any other measures used in this stage to - for example - equalise the bias currents in both legs (current mirrors, degeneration etc). As such, it is one thing that essentially is not fully corrected by feedback. So it will move depending on the parts selection (or not) at build time, and a very, very small dependency on temperature.

That said - while it is natural to want DC offset as small as possible - up to 100mV will do no harm with any moving-coil speaker you can think of. 100mV into 8ohms = 1.2milliwatts, and so ~10mA or so through the voice coil -which means an offset of 1,2,3 tenths of a mm at most* It's not going to do any harm, to music or long term.


*the displacement is essentially down to the drivers 'BL' product x offset current, vs 'Cms' - compliance

 

[Arkless Electronics]

Probably no help for people without the requisite test equipment but bias is best set by monitoring the distortion residual on a 'scope, which apart from the scope means you also need a low distortion oscillator and a distortion meter. With zero bias big spikes of crossover distortion will be visible. If you slowly increase bias you should see the crossover spikes diminish until they pretty much disappear. Increasing bias further will usually increase distortion due to Gm doubling distortion so there is a "Goldilocks" area which can be seen to be optimum. Optimum ish anyway... with quasi complementary the upper and lower devices have different optimum settings so a compromise must be accepted.

 

[Andrew.B]

I was bored just so plotted my poor test amps dc offset from cold start as it's temp rose, then calibrated it back to 0 +- 0.5mv at operating temp. Quite a range, though I'm not sure which exact components temp change actually causes it, I suspect the 32v (?) + regulator which hit 90c (?!) Lowering the bias, which is at the max recommended for this amp, would probably put less strain on this bit, and less temp flux? I don't use this amp atm.

 

[favorit601]

I had my Naim cd3.5 & Nait 3 serviced a little while ago, completely recapped, both were working perfectly prior to this & as in your case, all looked fine,
I fell for the myth unfortunately, cost me a huge amount too, when they returned, the sound lacked that Naim quality I enjoy, I sold them on with the full knowledge of what had been carried out, I have since purchased both again, no service, working fine, back to that Naim quality many enjoy, never again would I do this, only if something is not working would I consider a service recap.

>>In my opinion, you should have waited some time for the caps to formate before selling your recaped NAIT/CD3.5. A freshly recaped piece of Hi-Fi need at least a quarter of a year to sound good again. The first week, the sound is absolute terrible: weak, distorted, small stage, hardly any low end, no air. The first month, you get back 70 percent of the old quality. Depending on the caps quality and the temperature at the cap, this procedure may take some 12-15 month. Usually, I use low-ESR long life 105* Panasonic FC/FM caps. For my designs, there is hardly a chance of testing in the first 3 month after building/ recapping... „Keep calm and carry on!“

 

[Lee T]

6 months of use on a daily basis did nothing to diminish the annoyance, the sound was lacking after the service, whatever they did, I didn't personally enjoy the resulting lack of detail, focus & definition which was present prior to the service.

The amp returned & was initially over heating so needed to be returned for another go, I was hoping this was the issue but alas not.

 

[RustyB]

6 months of use on a daily basis did nothing to diminish the annoyance, the sound was lacking after the service, whatever they did, I didn't personally enjoy the resulting lack of detail, focus & definition which was present prior to the service.

If it ain't broke.....

 

[favorit601]

6 months of use on a daily basis did nothing to diminish the annoyance, the sound was lacking after the service, whatever they did, I didn't personally enjoy the resulting lack of detail, focus & definition which was present prior to the service.

The amp returned & was initially over heating so needed to be returned for another go, I was hoping this was the issue but alas not.

>>> Hmmm... What you tell about sound quality is almost the opposite of my experience. In most cases, new caps (good ones) have problems with the original tone balance, sweetness, "musicality" and so on. This can be a result of lower ESR causing esp. less linear impedance/frequency/phase response. Focus and definition with caps of newer technology of electrolyts (made for computer and switching power supplys) in my experience was always increasing.
Did the service really recap with trusted caps of high quality (Nichicon/Panasonic/Rubycon)? Or did they change the main caps (this is not neccesary at all...)? I think it has something to do with the type of caps. Otherwise no Idea!!

 

[favorit601]

6 months of use on a daily basis did nothing to diminish the annoyance, the sound was lacking after the service, whatever they did, I didn't personally enjoy the resulting lack of detail, focus & definition which was present prior to the service.

If it ain't broke.....

Thought of your problem: I think there was some kind of oscillation in one of the regulation circuits, lack of definition always tells you principle basics are not stable. For instance, the famous LM317/337-regulators (used 20 times in a normal Naim Electronics from the 1990s /2000s) were developed in the 1970s, when the word „low-ESR-Cap“ was defines as <1 ohm (values under 1000uF). In the 1990s a 10uF Panasonic FC has some 200mOhms, a 1000uF Solid Polymer nowadays is under 10mOhms. This leads some circuits to behave not the way they were designed.

Some more suggestions:
I made some simulations with the LM317-Spice-Models. When you do so, you find out why Naim, Exposure... always use 10-22 uF of about 800mOhms ESR to output pin. The LM317 has some 15mOhms from 20 Hz to 500 Hz and rising to the values of the output cap Obote. The impedance is smooth over frequency with a really small resonance/impedance high at about 25kHz with a 10uF up to 1Ohm, with a 22uF Cap this is somewhere at 12kHz but with some 600mOhms you got, this was a smooth impedance, making a small footprint to the sound, the developer dealt with in audio circuits. It‘s a really good compromise. When you use low-ESR-Caps, you sharpen the impedance-peak, when you also increase the value of the cap you lower the resonance frequency. The sharp peak results in a harsh sound.

I realized, that some German companies use more than 3000uF low ESR-Caps after the regulator. They did so, because at this values you reach about 100Hz with a smooth, low impedance of some 20mOhms and almost no peaks In the audible frequencies. 30 times lower than above.The LM-317/337-footprint is not audible. In my opinion, you get a very solid, neutral sound.

 

[Jim Audiomisc]

How many fine Armstrong 621 sold in the 70s are still out there in the field today-compared to 12/120 pairings? I don't think I've ever seen ONE.

That in a way illustrates some of the points/comments made. Although before adding some of my own, I'll point out that the main seller tended to be the 625/626 tuner-amps rather than the 621. (The *521* was somewhat different as a result of a 'Which?' report and promotion by Comet!)

*Early* 600 range items tended to use (electrolytic) caps made in the UK/Europe, whereas later ones used caps from Japan. Had an effect on the reliability as well as the performance.

My own experience with the 700 amps followed on. As yet (touch wood) I've not needed to change any of the caps in those, despite about 35 years of daily use. All good Japanese caps. No doubt they'll need replacing eventually. But they've done well so far.

The obvious difficulty for users is - if your amp 'seems OK' how do you tell if it is time to replace any of the caps? If no signs of bulges, leaks, etc, matter of guesswork. I tend to leave alone, but I'm lazy. YMMV.

 

[davidsrsb]

My own experience is that capacitance value does not change until long after useful life. The ESR rises earliest, but is hard for most people to test.
Seal failure on small capacitors is hard to see without unsoldering

 

[PigletsDad]

In reality, offset is not that big a deal, unless terrible. With most drivers, even 50mV uses up very little of the linear displacement range. Keeping at much lower levels is not a sound quality thing, just a nice thing.

 

[david ellwood]

Lower ESR is not necessarily a good thing.

Where capacitors are used as filters or noise bypass, the inherent ESR is likely to be useful in widening the band over which noise is grounded.

 

[Arkless Electronics]

Lower ESR is not necessarily a good thing.

Where capacitors are used as filters or noise bypass, the inherent ESR is likely to be useful in widening the band over which noise is grounded.

Nope. Pretty much the only time a low ESR can be a problem is at the output of some voltage regulators. In all other applications that spring to mind either ESR doesn't matter (such as coupling caps between stages) or the lower it is the better...

 

[PigletsDad]

Lower ESR is not necessarily a good thing.

Where capacitors are used as filters or noise bypass, the inherent ESR is likely to be useful in widening the band over which noise is grounded.

Sorry, when used as a bypass, ESR just makes it less effective.

As Arkless says, about the only case where ESR is needed is on the output of some regulators.