[davidw]

I'll occasionally post hints and tips for users of my various kits and modules, this is the first installment. I double check everything here to ensure accuracy but even I have been known to make the odd mistake. If it doesn't seem right check with me first.

If you've got a MOS100 you can reduce the, already low, hf distortion by connecting a 100k resistor between pins 2( - input ) and 6( output ) of IC1. This extends the "open loop" bandwidth of the opamp stage and allows the global negative feedback to remain fully operative at higher frequencies than previously.

If you've got a MOS100, MOS125, or MOS250 you can replace the dc blocking capacitor C5 ( 220uF, 50 or 63V ) in the global feedback loop with a non polar type ( I used a Nitai 470uF, 50V ) type. The non polar capacitors introduce less distortion than standard electrolytics.

While you're at it you can replace the protection zener ZD1 ( 2.7V 1.3W ) with an 8.2V or 15V type. The exact voltage isn't important as long as its greater than about 7.5V and less than the working voltage of the non polar electrolytic. Why do this? Because the leakage current of low voltage zeners is quite high ( around 1mA at 1.5V for a 2.7V zener ) and non-linear. Of course the normal working voltage across the zener is 0V but you may as well make everything as good as you can.

Remaining with C5; you can remove it altogether and jumper over its position on the pcb with a straight wire if you want. ZD1 and C6 are redundant if you do this. The output offset is so low, provided you're using a fet input opamp like the OPA134 or OPA627 supplied, that a dc coupled feedback loop causes no problem. This tip was suggested to me by a user who reckoned it made the amp sound noticeably better.

[Raj]

Hi David,

What sort of difference in sound did you percieve using the Nitai bi-polar cap?
Have you tried putting a wire across this position, if so do you think it's better than the non-polar cap and increased zener voltage tip?
Also what the safe maximum for a cap in this position, is a bigger cap better?

Thanks
Raj

[davidw]

Raj,
Putting a wire in place of C5 has to be the best option - no cap is better than any cap. You have to listen hard to hear the difference a bipolar cap makes compared with an ordinary electrolytic, but overall the sound is marginally smoother.

I personally haven't tried a wire in place of C5 although this must give lower distortion than any other option. A few of my customers have modified their amps in this way and all reckon the improvement is well worth while.

Out of interest one has also reduced the gain of his power amps from 23 down to four because his preamp can easily deliver 10V rms. Again this will result in lower distortion and a small improvement in sound quality. Be careful if you try reducing the gain of power amps - it can make them unstable, with catastrophic consequences for the output devices if you don't have adequate overcurrent protection.

[Raj]

David,

If using just a cd player with the buffer amp is it possible to dispense with the 4.7uf polypropylene cap in the signal path?

Thanks
Raj

[davidw]

Raj,
      You can usually leave out the 4.7uF polypropylene coupling capacitor at the buffer amp input provided that the unit feeding it either has a capacitor on its output, or a low output offset voltage (<10mV). The vast majority of CD players have a capacitor coupled output so you should be ok

[davidw]

If you've experimented with the mos125/250 you'll have found that they don't like very fast opamps ( SR>80V/uS ) in the ic1 position. This is mainly due to a lack of drive current from the bipolar transistors feeding the output mosfets. If you increase the drive to the mosfets you'll find that you can use just about any fast ( or slow if you like ) voltage feedback opamp for ic1, and also the increased drive current allows you to use more pairs of output devices if you want to.
Here's how you do it for a mos125.  Change R17 and R20 from 470R to 180R; change R23 from 1k to 390R (3W at least ); and change R24 from 220R to 82R ( 1W )
For a mos250.  Change R19 and R22 from 220R to 180R; change R24 from 470R to 390R (3W at least ); and change R18 from 820R to 75R ( 1W ).
If you make these changes, use a faster op amp than the OPA134 or OPA627, and leave out the 680pf capacitor at the input ( which I always do anyway ) you'll be rewarded with a much faster and cleaner sounding amp with slightly lower intermodulation distortion.
Got one or two new products coming out soon too. Details later.

David

[Raj]

Hi David,

1) I've been rolling the idea of adding seperate regulated supplies for the opamp section of the mos 250. Given the fact that there's already onboard regulation to the opamps, - albeit from the main power supply which also drives the output stage, is this worthwhile in terms of cost and benefit?

2) Adding extra output transistors was somethig I thought about whilst I was increasing the quiescent current, but I always assumed that the cleaness of the amp would suffer as the mos only uses a single pair to provide the power, my limited knowledge has believed this to be an advantage over using multiple pairs of ouput devices.

.... Is there any correlation between the number of output devices and overall sound quality/texture?

3) The pot thats used to trim the quiescent current is a high quality sealed type. Is there any benefit to reading the value of resistence accross the pot and using high quality resistors to the same value instead, if so what would be the required wattage?


Thanks
 
Raja

[davidw]

Raja,
       Yes its worth adding a separate supply for the opamp. What I recommend is to replace the 2k2 2W resistors that feed the onboard regulators with 1k 2W types but to only solder in the opamp end of each resistor. The other end of each resistor is joined to the corresponding output of a cascode power supply set for +/-24V operation. You must join the 0V line of the cascode psu to the star earth at the main power supply 0V connection. This completely isolates the opamp from the main power supply by providing two stage regulation.
  As for multiple output devices the pros are that the devices run cooler so they last longer and produce slightly lower distortion. The cons are that unless the devices are perfectly matched ( i.e. all nch matched and all pch matched, you dont have to match all nch with all pch ) there will be a small rise in intermodulation distortion, and of course the cost is higher. My take on this is that multiple output devices are a better, if more expensive, option.
  Its always better to replace the trimpot with fixed resistors once you've settled on a value for the quiescent current. Quarter watt types are fine, and the Welwyn RC55Ys are ideal.
  Apropos of my last post I tied a MOS125 with an LM6171 opamp, these have 3200V/uS slew rate. Interestingly enough the amp was just about stable but more of a radio transmitter than an audio amp! So I don't recommend opamps as fast as this but anything up to 200V/uS will be fine.
These slew rates are gross overkill but I always reckon fast amps sound very clean - that's probably just because I know they're fast.

Regards, David

[davidw]

Please bear in mind that nearly all of the transformers you buy in the UK are wound to give their rated output voltage assuming a 230V input. In most of the UK the mains voltage is around 240V so that you'll get a higher than expected output, particularly at quiet times of the day when the voltage rises. So be careful if your power supply gives close to the maximum rated voltage for the amplifier that you're using.

David

[Raj]

Hi,

I've now removed c4 entirley, and was quite suprised by the difference in sound......... So the desire increased to see if I could remove c1 also, which is the dc blocking cap at input, I removed this and measured the dc offset at ouptut which was around 20mv (with 2 different sources connected and checked), the result is a far more natural presentaion.... Well worth it. A word of caution though I use opa 627's in my mos 250's  which David has pointed out have a low dc offset. So perhaps caution is to be advised with different opamps and also different sources too. I'll let David warn of any possible problems in this department as this is definatley his forte and most certainly not mine.
As you pointed out David I will check new sources for dc offset when and before I connect them to my sytem......

So now my entire system including cd player,pre amp and also power amp does not have any dc blocking caps at all, or dc servos (as far as I know), the cdplayer uses output transformers, which negate the need for dc blocking caps. The pre amp again uses opa 627's, so the dc offset is again low. Again I don't want to be responsiblle for any problems due to misadventure, so I'll let David warn of any possible problems and dangers........ Again testament to David's excellent design............


Thanks
Raja

[davidw]

Raja is definitely doing the right thing - the best capacitor is no capacitor! You'll find that if you use fet input opamps like the OPA134, OPA604, OPA627, AD744, AD825, AD8610, etc then the output offset produced by the amplifier itself will always be low - often a bit less than the 20mV Raja quoted. A dc output offset of up to 100mV is tolerable, particularly if you have power on/off muting. DC servos are to be avoided at all costs, they're far worse than poor capacitors in the signal path because most of them really screw up the lf response of the amp. Setting the dc output offset to near zero with a trimmer is a far better option than using a servo.


David

[davidw]

I was just reading the latest edition of Douglas Selfs power amp book when I chanced across a section on relays for ls protection units. He found that some high current relays use an iron alloy frame which is part of both the magnetic and electrical circuits. Relays of this type can introduce large amounts of distortion, up to 0.1% at high currents, because the iron alloy is an easily measureable non-linear conductor.
 The relays I use have silver contacts set in copper plates and the contacts are actuated by a nylon bar. The coil and magnetic parts are at the far end of the relay from the contacts. I took one apart to discover this. This type of relay has no effect on sound quality whatsoever.
 Unfortunately I think that the relays shipped as standard with Vellman loudspeaker protection units are of the distortion inducing type, although I don't have one on hand to take apart. If you have a Vellman unit it might be wise to change the relay if you can get a physically compatable type with a "good" construction.

David

[Ian M]

If you've experimented with the mos125/250 you'll have found that they don't like very fast opamps ( SR>80V/uS ) in the ic1 position. This is mainly due to a lack of drive current from the bipolar transistors feeding the output mosfets....

I am interested in trying faster opamps in my MOS100's, can a similar mod be implemented ?

Incidently, as a test, I have just hooked up separate battery power supplies to the opamps and I can report that the improvement is significant in pretty much all departments but especially in the bass. I wouldnt recommend batteries as a permanent solution though, as the consequences of battery failure could be serious.

[davidw]

Ian,
 The output stages of the MOS100 and MOS125 are virtually identical so you can make the same mods to the MOS100. You'll find that using a faster opamp will reduce the intermodulation distortion so that instruments and voices are better resolved. The effect isn't massive but is easily audible. Let me know how you get on if you do try some fast opamps.

David

[Raj]

Hi David,

As you know my mos 250 uses opa 627's....

Which opamps are available easily as a suitable replacement for these? I know quite a few people like the ad825's, which I have heard are to be discontinued, the only problem with these though, is that they are surface mount so the right sockets are required, any suggestions for possible sources etc or other good alternatives?????

Thanks
Raja

[davidw]

Raj,
    The AD825s work pretty well in the MOS125/250. I've got a few of them with some soic to dil adapters. I can make up one or two of them for you if you like. Didn't know that they were being discontinued though - I'll check.

David

[davidw]

I haven't tried this tip myself but one of my customers has and he claims that it improves the measured performance ( he has a commercial distortion analyser ) and makes the MOS125/250 sound better. What you do is reduce the gain of the power amp from about 20 to around 5 -  this will undoubtedly lower the measured distortion. Don't go any lower than 5 or there's a risk that the power amp will oscillate. To do this you need to change the value of one resistor - details later if anyone's interested. The power amp will now need around 4V rms input for full output. If your preamp can manage this then there's no problem. Otherwise you need to increase the gain of your preamp or add an amplifier stage with a gain of around 4 between the pre and power amp. The customer who suggested this mod had a preamp which produced 10V rms output. Happy modding.

David

[davidw]

Amplifiers with a fast slew rate and hence wide power bandwidth will, in general, always have lower intermodulation distortion than a slower amp. For example make a simple heaphone amplifier with an OPA551 and listen to massed orchestral music. Then swap the OPA551 for a LM6171 and repeat the listening test with the same music. You don't need golden ears to realise that the LM6171 separates the instruments far better than the OPA551. The only substantial difference between the two opamps is their slew rate; around 20V/us for the OPA551 and around 3000V/us for the LM6171.
You can experiment with faster versions of the MOS125 by lowering the value of the 22pF capacitor, C14, across the 2k2 feedback resistor. In my experience the power amp is still unconditionally stable with a 10pF capacitor; and I have run a MOS125 with no capacitor at all, but I wouldn't recommend this. If you lower the value of the capacitor too far the amplifier will begin to oscillate. This isn't usually a major problem, the amp will typically draw 0.5A and doesn't seem to fry the zobel network, but you should always try these experiments with a dummy load rather than a loudspeaker connected to the power amp. Remember too that a power amp with no input and no output load may be stable but can become unstable when a load is connected and/or an input signal is present.
I wouldn't recommend trying this with a MOS250 as the phase margin is lower because of the increased transconductance of the output devices.

David

[Raj]

Hi David,

with the mos 250, if using a passive preamp with either a 10k or 50k signal resistor in the attenuator, how important is the presence of r1 and r2 at the input of the amplifier?

also do you know where I can get a metal oxide to change r24 down to 390r?
I've had a look around but can't find a 3w metal oxide of this value. I managed to obtain the 180r welwyn's from farnell, and 1w 75r resistors are easy enough to find.

Is there any harm in trying the resistor changes with the opa627's still in place, I understand that the true potential for slew rate won't be realised, but would like to see if the difference is still worthwhile.


Thanks
Raja

[davidw]

Raj,
      You can jumper over the 2.2k resistor at the input as it only forms part of a low pass filter ( which I personally never use anyway ). In theory you could omit the 47k resistor too if you're feeding the power amp directly from a stepped attenuator or potentiometer - but I definitely wouldn't do this if I were you. The 47k resistor sets up the ground reference for the opa627 and as a consequence the dc conditions for the rest of the power amplifier. If the stepped attenuator switch momentarily loses contact with its resistors, or the potentiometer slider loses contact with the track, or you unplug the input then the dc offset at the output of the power amp will be all over the place - extremely undesireable - at best it will trip the ls protection at worst it will put your woofer at risk. In addition the dc offset at the output would vary slightly with the volume setting, even if everything else were ok.
  You can get 390R resistors in 3 or 5W ratings from Farnell - they call them power oxide or metal oxide types. You can also get 390R by paralleling up the 2W metal film types that Maplin sell.

David