Repairing the Musical Fidelity A220

[NickK]

I bought this 24 years ago in 1998 and I thought it was about time I investigated some obvious hot traces for the mute control. I hooked up the scope and switched on good but needed to change the parameters. Switched off and on again and I had a light show. I was half expecting something like this at some point - it's getting old and if it's totally dead I have a nice case and metal hardware to make a brand new tube amp



That fuse is rated at 250V T6.3A and the transformer is providing around 30Vac. So quick bit of maths, R=V/I so 30/6.3=4.7ohm, and P=I*I*R = 6.3*6.3*4.7 = 189 watts before that fuse goes pop. The amp has two - one for positive and one for negative rails. So something on the positive rails sucked a lot of power.

Here's the board being stripped down - can you see the hot spots? (discoloured PCB). The fuse holder at the bottom left side is the one that blew. Above around that are 400V caps, and above that BR1 is the bridge rectifier - a 400V 10A device that converts AC to DC. Looks like it got a bit hot down one side.



Sure enough a test of the AC to the +Vdc rail portion of the bridge rectifier showed a full short. So no rectification and lots of current flow.. especially when 15000uF + 2x 3300uF 50V caps are connected directly to that along with the rest of the amp. So the fuse and bridge rectifier need replacing - easy.. but what caused the over current condition?

The amp is simple - power comes in, goes to the power transistors attached to the big black heatsink at the bottom of the picture - 2 for left and 2 for right for the positive and the same for the negative rail. Well the speakers are still work fine, so after desoldering the n-channel positive side mosfets - these test ok. They should do - they can cope with 8A at 160V each. Phew - these are obsolete and have no replacement and even the substitutable pieces need some playing to get to work.. So what else?

The amp is packed with electrolytic caps - all rated at 85degC. This amp is class AB and is biased so hot that you can't keep your hand on the top of the case. That's on the outside.. the inside of the amp turns into an oven.. 24 years of cooking and they are prime suspect. Even when they test at 9-24Mohm it's unlikely they're still good.. well they do hold charge but given they just got connected in reverse thanks to the BR failure.. that's likely to cause a massive current flow.
Option two for failure is that the amp went into oscillation at a high frequency but we've already tested the output mosfets and they're ok.

I've since tested the transistors in the driver section, the mute etc by desoldering them and testing them - all ok. Given nothing else in there is really voltage or polarity sensitive.. that leaves the caps and the old bridge rectifier as being the main actors in this. There doesn't seem to be any further burning of the PCB. After investigating the mosfet bottom right second in that is current removed, that tests ok but there was no real thermal connection to the heatsink! That would explain the rather toasting via the heat going down the legs and burning the PCB.

I've also noted some rather dumb parts of the design - first the signal goes through an electrolytic cap rather than a film cap hence you get less detail and the cap acts like a filter, so I have a couple of 0.22uF film caps I can use to test out any improvement (WIMA FKP1, WIMA MKP series 1uF and a Panasonic MPP 0.22uF) that I will try once it's back running. Also the designers use polystyrene caps which if you know about them.. they change with temperature and once change stay like that forever.. a bit crazy and unpredictable given the oven heat of the amp during operation.

I've ordered new caps, a new bridge rectifier, fuses, some thermal pads, and two 100W 8ohm resistors to act like speakers for testing. I've also been modelling the amp in a simulator and there's some additional stability mods to remove some 2MHz noise that causes problems with the amp and I can make that may help make the amp sound better too. Just need to order some more caps.

I'll start the amp with a light bulb in the power line - this sounds odd but limits the current that the amp can draw from the mains. I can then essentially power up the amp but limit the current being drawn through it so if something is going pear shaped then I can see it in slow motion and stop it before it goes bang.

 

[NickK]

First uncontrolled start OUTSIDE


Voltages all good, and I left it running for 30 minutes to test it and leave the caps to fully form (I'd preformed for precaution with 13V earlier). The green blocks are 100W 8ohm resistors to act as speakers.

I did make a minor boo-boo in that I ordered the four caps with pins winder than the holes. No Biggie, some creativity for the initial configuration with some wire and heat shrink.



Marked are all the bits replaced/upgraded and/or both. You can see the fat boys don't quite fit the original spots for the caps on the topside lol. Next steps are to get it hooked up to the scope and try some music before braving running with a real pair of speakers.

 

[NickK]

Sounds great, just needs the MPP cap to settle down (my WIMP FKP and MKP caps took a while to loose a little harshness)

In typical nick style, I've also modelled the amp power section in LTSpice (a circuit simulation tool). The simulation shows that the system is unstable and prone to a 2MHz oscillation. At worse this can destroy amps/speakers etc at minimum it can add some fogginess/harshness.



That shows a 10KHz sound but you can see a mass of noise (oscillation noise) that is caused by the feedback mechanism the amp uses to correct any errors. That's well outside of the audio spectrum in the RF spectrum at over 2MHz.



Interesting - that the real amp also picks up RF at 8.9MHz which is short range radio RF on the scope at the speaker jack whilst playing music.

So a future enhancement would be to add decoupling caps to compensate and remove oscillations.

 

[NickK]

GOT NOISE? The amp does not have a mains noise filter..



Yes that may be in the MHz frequency but any noise messes up the sound. The ~100KHz bumps are SMPS switching noise from elsewhere in the house.

I had a IEC mains filter in the 'high' but not 'highest' attenuation that I was going to use for the tube amp but this is a good time to try it.



I think that nailed virtually all the noise... and severely attenuated the 1.57MHz noise... ahh peace on the audio spectrum.

A while back I wrote/adapted a software signal generator.. now comes it's first proper use Bode plot time to show the frequency response. There are caveats - the linux VM, the Mac sounds system and the Mac analogue filter all put their own sonic stamp and restrict testing the fill frequency range... but it's good enough for the test..



The scope tells the software what pitch to play, the scope then samples the input and output from the amp.. giving what you see here. That crazy pitch shifting is the Mac audio filter kicking in at 20KHz.

 

[Feek]

Enjoyed this, thanks Nick.

 

[NickK]

Enjoyed this, thanks Nick.

No problem.

I just did a full volume run. The 8ohm 100W load resistors and the amp smell hotter than a hot thing on a sunny tarmac day. The caps were getting such a work out that you could hear the pitches of the bode plot through them. It's akin to running your speakers at 50W full volume, without irritating the neighbours.



Not bad. It appears you get a butt load more bass (and I assume that's at a higher THD). I could really do with a proper sound AWG and a proper sampling card for spectrum analysis (I have my eyes on a 1MSPS sampler that that operates at -140dB ).

The reason for this additional testing is it see if the amp is unstable as it is in the sim. It appears it's more stable, although the scope's Bode plot doesn't go down far enough in dB due to the scopes noise. I would add some pico-farad caps to reduce the instability based on the frequency that the amp is oscillating at. It's oscillating it's doing it quietly so there's still some possible gains to be made with some pF surface mount caps across some of the transistors.
MF amps are known for their instability..

 

[NickK]

Today I should have all I need to put in the second order. Being a little slow because I want to reverse engineer the schematic to select the right components based on function. So currently I have recapped the main rails used by the power amp stages - there is plenty I can do to this amp that goes beyond the fixes.

First this..





The above schematic contains two building blocks - the regulated power supply at the bottom and at the top is the moving magnet phono pre-amp section.
It's important to know what the caps are there for to make a considered replacement. The power section caps on the right can be replaced by Panasonic FC and I may increase the capacitance to provide more smoothing, thus reducing noise for the pre-amp and MM sections. The resistors pointed at will also be replaced as they've been cooked and should really have a better/higher wattage component used to provide lower noise (noise is related to current draw and the temperature of the component so lower current and lower temps the better).

Now the left of that first image has the phono MM section and the caps with the purple arrows are actually electrolytics on the signal path. It turns out these are input coupling and output coupling caps. So given these are audio signals and low level then Nichicon MUSE non-polar caps with a film bypass cap would give a really good result. I could use boutique caps such as audio note etc but the size and the amp isn't "top notch" means they're likely to be cooked whilst not returning the performance for the investment. I've used Nichicon MUSE ES caps before with a FKP bypass and they're decent as long as the voltages are line level. The two remaining red arrowed caps are filter caps so 105degC Panasonic FC caps would work here.
I could also improve the performance of the MM section by replacing the resistors (such as the green arrowed photo loading resistors) with better quality resistors but for now I'll simply do the changes for the caps as they need it anyway.

I tend to favour Panasonic given their tolerances are typically very small and the quality is high but can favour higher frequencies, Nichicon, CDE (Cornell-Dubilier), also Vishay too.



Above is the main preamp section that feeds the power amps from the sources and the MM pre-amp. It's powered by the regulated power supply above.

This is where I would target more changes including resistors - simply because all the sources go through these:
* Caps C25 & C25 are miles away but they're electrolytic.. sure.. but look at the schematic - they're signal caps!! so I think these will end up be nichicon MUSE ES with FKP bypass.
* I've already changed the C2&C213 to Panasonic PC as they're power decoupling caps. An improvement could be to add a bypass FKP so that noise can be reduced. Sinking the noise to ground away from the Preamp MC33079P opamp package.
* Resistors R40,R140 could be upgraded along with R41 and R141. These could also be paralleled to reduce noise.
* Resistors R46 and R146 can also be upgraded/paralleled to reduce noise.

Now I could go nuts and create a daughter board with the switch for the tape monitor, here's why:



You can see the signal take as road trip around the PCB. Two options to improve noise would be:
a) a daughter board on the selector switch could then implement the tape switch via relay then use a shield coax to get the signal to the pre-section.
b) The volume could be carried by shielded coax from the pre-amp itself.

This would be invasive as I would need to cut the tracks to prevent the remaining tracks acting like receiving aerials.

I have a couple more changes for the power amp stages:
1. The resistors carrying the current can be uprated/paralleled so they don't overheat (there's a couple which have discoloured).
2. Targeting the RF hash noise by adding small 27pF caps and bypassing some to improve the stability performance which results in lower noise and better sound.

Now a fellow MF owner got in touch and he's done many of the changes (such as bypassing and upgrading the caps etc) he's also seen the addition off caps to the mosfet pins but given the heat levels I the caps would need to be 125degC to allow temp derating to maintain performance.
Lastly the option of replacing the quad opamp MC330979 could also be done.

Now I'm trying to keep the character of the amp but just add some additional detail which I'm starting to see. The danger with going at the board with a parts cannon is that you replace everything and end up with a completely different sounding amp.
If I wanted a completely clean amp then I would have bought an Audiolab 8000 I listened to when I bought the A220. The 8000 was stark and cold, unlike the A220 warm sound.

 

[NickK]

I'm just highlighting one channel here but the opposite channel component needs the changes too.

Frequency compensation - this makes the amp more stable. However it's a little "try it and iterate" given the complexity of the system.

• TDK 27pF C0G 5% 100V caps for Cadded1 on the LTP around C124 and Cadded4 around the TR106
• TDK 10pF C0G 0.5pF 100V caps for C104
• CPF3 10K 1% 3W for Radded3 - way over the top wattage but if this doesn't work I can use it for the tube amps.

Power stage fixes

• 24y/o C103 100uF Jamicon 'lytic - this will be bypassed but I ordered Panasonic FR 25V as a slot in replacement should I back out the freq compensation.
• C101 0.22uF MPP -> WIMA 1uF MKP10 5% I have already that's burnt in.
• discoloured R106 1K is being upgraded to CPF3 3W 0.1%
• discoloured R110 6.8K is being upgraded to CPF2 2W 1% 6.81K nearest match
• R108 being upgraded to CPF2 2W 1% 56K to match closer to the 10K characteristics - not strictly required.

Preamp fixes
* 24 y/o C25/C125 Jamicon 'lytic signal path replaced by WIMA MKP4 10uF 100V I may bypass with FKP1 caps.
I resisted the urge to replace all the signal resistors with larger ones at this time.

Preamp/MM regulated power supply

• discoloured R204/R203 220ohmª wattage upgraded to CPF2 2W 249ohm 1%. Edit: I made a oops and misred the resistor so I will be a little more resistance. Not a biggie.
• discoloured R202/R201 1.5Kª wattage upgraded to CPF3 3W 1.5K 0.1%
• 24y/o Jamicon C206/C205 47uF 50V replaced by 56uF 63V Panasonic EEH AZA
• 24y/o Jamicon C204/203 1000uF 16V replaced by 2x 560uF 25V Panasonic EEH AZA to give 1200uF 25V. (resulting in 5mOhm ESR)
• 24y/o Jamicon C202/C201 100uF 16V replaced by 220uF 25V Panasonic EEH AZA.

The AZA caps take a hit for frequency derating for low frequency up to 0.15 but given their ripple is in the order of 4000mA this isn't an issue with the MC33079 taking ~12mA from the supply rails. I'll admit I wanted to see how these Al organics compare and I hope that the low ESR isn't going to be too much of a problem. I'm expecting iron solid supply rails!

MM stage

• 24y/o Jamicon C17 100uF 16V replaced with 100uF 25V Panasonic FR
• 24y/o Jamicon C14/C26 10uF 16V signal path replaced with 10uF 25V Panasonic FR but is likely to be changed to MUSE ES (although their lifetime is 1000h!).

Last but not lest the Schaffner FN9262-6-06 IEC power line filter. For now I've given the benefit of the doubt to the diodes as they test ok..

The only other caps are Panasonics for the 24y/o Myriad MC100 CD player power board (2x1000uF+4700uF and a 100uF).

That will do for now. It should solve the power problem, all the Jamicons will have been replaced along with any damage they've been causing.

EDIT: Corrections(ª)
EDIT: I took one of the 1000uF 16V caps out of the pre/MM-PS and it read 2Mohm.. that's bad! So I can't realistically use the amp until the new parts are here.

 

[SenTineL]

Thanks for the write up - was a good read

Reminds me that I've got a boxed Musical Fidelity A1 in the loft over at my parents house in need of some attention.

 

[NickK]

Well.. and then there was none.. well no old Jamicon caps anyway!

The caps are now replaced, and so are most of of the discoloured resistors.



The flip side.. only a subtle use of capacitance here..



Now I found that the pre power stage had an imbalance (++ve to -ve being different) and I was going to double check one of the resistors but I pinged it behind the monitor whilst attempting to test it. I'll find it tomorrow and measure it. if it was 1.5K then one of the old caps was causing an imbalance, if it measures different and the strips show non 1.5K then I need to adjust that resistance and perhaps that is how it gets some class A second harmonics into the system.



Frequency response shows it's ok, but I've not applied any stability caps etc and I still need to re-adjust the power section bias. I will look tomorrow at measuring stability.

Listening to the amp with music - it's far faster than previously but it's very fast on the bass, clearer in the mids and highs. The preamp power supply is 5mOhm which is bananas and probably explains the speed.The preamp IC bandwidth is 16MHz whilst the output mosfets are in the order of 100MHz.

I think before making further changes I need time to measure things.

 

[NickK]

So brief update.

I decided to try out the the frequency compensation and switch out the Panasonic ECW 0.22uF coupling cap between the pre and power stages. Pure dead-bug style wiring going on


So after the changes - note that it looks like there more of a bass bump but look at the scale and you can see it's actually flatter:


The audio listening - more bass and initially shrill treble which gave way to clearer treble that will only get better - this is the same behaviour when I have used the combo for headphone amp. In comparison the other channel with the Panasonic has a definite mid frequency but lacking highs and some depth on the low end. ECW is more tube sounding but not without some clear treble.

The results of the frequency compensation aren't audible as such, but this give a good idea between the left (the mods) and right channel (ECW):


You can see the two super imposed - same scales and settings. The dark blue is the modded left channel, the light blue is the right channel (no mods).

The flatness of a square wave output shows how the amp is coping. In this case the left channel slope tells me that there's low frequency attenuation (small 0.22uF PCW) vs the more balanced larger 1u+0.22uF WIMA caps paralleled. This confirms the listening.

The difference in scale may be explained by the difference in bias of the two channels - left is 21.5mA and the right is 19.33mA or there's a resistor or capacitor that's not closely matched with it's mirror on the other channel. The additional noise may be down to the dead bug style circuit as this has lots of bare metal to act like an antenna to receive noise.

I think that will do - now it's a case of make the changes good (ie not be so fragile etc) and then fit everything back into the case.

 

[NickK]

Back together again and I had a little listen yesterday - sounded very good. There's plenty more changes I could do but in reality I don't want to pour money into a design that's not the best in the world.

ON reassembling and sorting out the final positions I managed to rip a couple of PCB trace pads from the board. Not surprising really given it's been cooked/cooled repeated so the traces of any old PCB are fragile to say the least. Even with putting a pin through and soldering resulted in dead pads.

So the next best thing - wire patches some of the broken connections under the board and then do some direct to pin point-to-point soldering. This meant that the new signal caps are connected above the board and the signal is pulled from the tape out connector which is directly connected to the source selection. The black cable is shielded to ground.

The amp now has more clarity by the bucket load but also deeper extended bass (the larger cap between the pre-amp and power-amp stages means less low frequency roll off) but that bass has more detail too.

It exceeds the sound quality of the original amp when new but retains the musicality which is what I was looking todo.

The green tape is simply to add some additional coverage - there's a lot of heat shrink tubing (600V isolation) and I should sort out some cap support at some point. I'll take some measurements with the case on (protects from RF) later on in the week but for now that does nicely!

 

[NickK]

Thanks for the write up - was a good read

Reminds me that I've got a boxed Musical Fidelity A1 in the loft over at my parents house in need of some attention.

http://www.exicon.info/news/end-of-an-era-for-to3-lateral-mosfets/

If you need to replace the lateral mosfets using a similar metal TO3 can variety - they're being discontinued by Profusion. If the mosfets are ok then a recap and it should be fine. The A1 has quite a number of recap and fix blogs due to the heat.

 

[SenTineL]

http://www.exicon.info/news/end-of-an-era-for-to3-lateral-mosfets/

If you need to replace the lateral mosfets using a similar metal TO3 can variety - they're being discontinued by Profusion. If the mosfets are ok then a recap and it should be fine. The A1 has quite a number of recap and fix blogs due to the heat.

Cheers for the heads up on Profusion discontinuing the mosfets. I'm not 100% sure they okay, so it's probably worth grabbing some, even if I end up saving them to use at a later date

They certainly do get toasty - gotta love class A! Years ago I borrowed an A1000 for a home demo over a long weekend and didn't need the heating on the whole time

 

[NickK]

So I finally got around to re-biasing that right channel. 1 hour amp and scope warm up, scope self-calibrated. Here's the max waveform at max volume, you can see the clipping of the right channel:


Now, it appears that the 24 years of baking hasn't done the trim pots much good. The left channel shows 95mV across one of the output resistors (0R22 big green ones at the top of the last PCB image). This is giving me the bias for that channel. Next is to attempt to get the right side to 95mV.. would it? nope.. 80mV ... turns screw a fraction.. 100mV.. nope... so finally I managed to get about 6mV between them.

It still shows a the right side is clipping less than it was before. I suspect this may be differences in resistors/capacitor components as they can shift over age/temperature damage.


However that is at the top of the volume dial. The amp is putting out a good 70V peak-to-peak and the 100W 8R load resistors are getting hot. with 89dB/W sensitivity that's more than enough to be rather deaf at this point (not to mention the THD probably is probably high). Turning down a touch results in this:


That will do. There are some differences I suspect between left and right but none that I'd notice. I could improve the biasing but again, it's an old amp and is working better than it did when brand new. I would prefer to make something new rather than continue to pile lipstick on the pig.

I have four matched Profusion EXC10N20Ps lateral mosfets on the way.. but that's for a different thread