Quad 909 Technical details.

[Hodges]

Thanks for that - I appreciate the effort. The mechanism is still an extremely grey are - well for me anyway. But I will take the time to read some more historical articles which may make things a little less confusing. For example, Walker talks about Feed - forward and error signal, yet everything to date refers to it simply being NFB. It may well be, this simplifies matters for the authors, yet it still does not explain how the switching rate of the dumpers can be controlled, 4 times in one cycle and maintain distortion and noise figures approaching that of class A designs.

I reiterate however, the dumpers must be switched on sufficiently far enough from their non linear region, to ensure that the complex make-up of the audio doesn't provoke non linearity by hovering between on and off. This could occur where the drive (audio) is only just above the threshold of Dumping action. That would put it closer to Class C, which I previous said. I don't see how it can be Class AB as this would but it at a point where the alternative half of the output stage are at the threshold of conduction, virtually making the principle of 'Dumping' redundant.

Still, I am just chewing the cud, whilst I try to get a handle on it all.

Cheers,

 

[Arkless Electronics]

I think you need to spend more time with those articles. It seems you are not understanding how current dumping works. I gave the example of the Crown DC-300A as an amplifier that works in a simple and crude way (output stage operation anyway!) to use the output devices as "current dumping" in the way you appear to think the Quad works... they are VERY different though! It is unfortunately not easy to understand Quad's current dumping idea.

Whilst it may seem counter-intuitive the 405 and later dumpers do indeed have some forward bias on the output devices. It reduces distortion, especially at HF. I have designed a few Quad style current dumpers and can confirm that a little bit of forward bias really helps. The precision of the LCR bridge has a big effect on THD also.

The class A amp should not be thought of as driving the speakers really, and in fact there is a 47R resistor (part of the bridge) betwixt the class A amp and the load! It controls the dumpers and "fills in" the gap as they start to conduct. The magic is in the way the bridge and feedforward make this process theoretically perfect so there can be theoretically close to zero distortion.

Don't worry, even some of the worlds leading luminaries in audio electronics at the time couldn't agree on how it works, IF it works as claimed even and whether it really uses feedforward or feedback!

It is a very clever idea but even by the time it went into production was a bit of "a solution looking for a problem". 10-15 years earlier and it would have been much more important as germanium transistors were much more sensitive to thermal runaway and overheating issues and of much lower performance than silicon in many ways.

Note that the patent on current dumping ran out some years ago now but there hasn't been a slew of rival current dumping amplifiers coming to market!

 

[Hodges]

If you are not going to admit that the dumpers are a Class B amplifier (they are), then we can't move on.

Peter Walker's own 1975 paper describes the principle: "...feed forward carries the promise of reducing to zero the distortion of that part of the amplifier over which it is applied. If this is the class B stage, then not only does the distortion itself disappear, but all the paraphernalia of quiescent current adjustment and thermal tracking disappears with it." That is a description of the current dumper stage.

The amp is a low power Class A stage, followed by a high power Class B stage with unity voltage gain (ie current amplification only) that only turns on when needed, and the latter has this very clever configuration that corrects distortion, self-maintains bias (at exactly 0 mA, ie Class B) and auto-tracks for thermal drift.

A bit of background reading on current dumping amps might help to clarify that they are invariably Class B...in fact, if one wants to be really picky, one could argue they are Class C because the conduction cycle of the output transistors is not quite 180 degrees. But if Walker is calling them Class B, then let's go with that, since the practical use of the term Class C is for amps with much less than 180 degrees of conduction cycle.

Walker's 1975 paper in Wireless World, link.




cheers

If you are not going to admit that the dumpers are a Class B amplifier (they are), then we can't move on.

Peter Walker's own 1975 paper describes the principle: "...feed forward carries the promise of reducing to zero the distortion of that part of the amplifier over which it is applied. If this is the class B stage, then not only does the distortion itself disappear, but all the paraphernalia of quiescent current adjustment and thermal tracking disappears with it." That is a description of the current dumper stage.

The amp is a low power Class A stage, followed by a high power Class B stage with unity voltage gain (ie current amplification only) that only turns on when needed, and the latter has this very clever configuration that corrects distortion, self-maintains bias (at exactly 0 mA, ie Class B) and auto-tracks for thermal drift.

A bit of background reading on current dumping amps might help to clarify that they are invariably Class B...in fact, if one wants to be really picky, one could argue they are Class C because the conduction cycle of the output transistors is not quite 180 degrees. But if Walker is calling them Class B, then let's go with that, since the practical use of the term Class C is for amps with much less than 180 degrees of conduction cycle.

Walker's 1975 paper in Wireless World, link.

A follow up 1981 paper in Wireless World by Malvar, expanding on Walker's early paper and leaving no doubt that the output stage is Class B, link.

cheers

Thanks Arg;

''If you are not going to admit that the dumpers are a Class B amplifier (they are), then we can't move on.'' - ''if one wants to be really picky, one could argue they are Class C because the conduction cycle of the output transistors is not quite 180 degrees.'' Now which do you want me to agree to? One knowledgeable individual pointed out that Walker believed he could get away without any bias, but discovered it was less than successful and added the diode - making it Class AB.

Whatever mode is irrelevant really - I just want to get a succinct paragraph that describes how it all functions - not necessarily a mathematical analysis, although that would be most welcome. If my memory serves me correctly, then the turn - on voltage of a Silicon transistor is 0.7volts. So an AC swing of 2 volts P to P is not going to place any great burden upon the Class A section. A little more for insurance purposes would help, but either way it is well within the rating of the Class A amp.

Now, for me, the whole question revolves around getting the dumpers, to switch on at a point where they proceed to conduct over the linear part of their characteristic. It is how that is achieved that remains unanswered. We can do an analysis of the static conditions - but dynamically, what is doing what? Add to the mix the fact that, whatever it is, it is doing it at 4 X the rate of the audio signal!

I accept that the Dumpers may not be fast devices - indeed it was my poor explanation that led to the confusion. What I mean is, the speed at which they are switched on has to be incredibly fast. So; whatever the Feed - forward signal is, needs to be damned quick! But how is it derived, shaped, so that it is very precise? Analogous to this, it seems it acts like a ''Trigger'', holding the Dumpers out of circuit, until the threshold is crossed, then swiftly switching on the base - emitter junctions- very quickly. Up to this point, the Class A amp can handle the output signal power; and probably more, until dumper switch - on occurs.

Thanks for continuing input.

 

[Hodges]

I continue to cogitate over the issue of CD, because I find it so fascinating. I doubt I will ever get to the root of it all, but continue to marvel at how well my 909 performs.

For the moment my Sugden Headmaster remains an ''Ornament'', just as my 33 has become, given I am driving the 909 from a Alps zero - gain/potentiometer, preamp, with a further 20db attenuation between the two. They are being driven by a Zen V2 and it seems a huge leap in quality over my 303/33 combination. I just never tire of listening to it, particularly as I don't think I have heard my Spendors perform at this level - ever. Ok, they have new ferrite drivers and clearly aren't capable of handling high power levels, but when in their comfort zone, they bring a realism to the music that never seemed to be there before.

What continues to surprise is the Bass response. Certainly, the 909's output is sustained better than either the 303 or the 405 and this, I think, gives whatever the Spendors are capable of; at the lower registers, a fighting chance of producing some decent bass. Ok, it's not going to crack the mortar between the brickwork of our lounge, but they certainly don't leave me feeling cheated out of anything - Neil Diamond is not made to sound like Olivia Newton John!

However, there are two other issues I am looking at, The first being the lack of tone controls, particularly when listening at very low volumes and my hearing drops off a cliff. I am reluctant to add anything by way of gain for the sake of getting a little bass and treble lift at low level, so my 33 will not be coming back into service any time soon. But equally, I am not all that confident about using some form of loudness control. I have bought a couple of loudness control circuits to experiment with, but I don't want the loudness to be continuously adjustable over a wide range. Just a pre-set amount of lift at a pre-determined volume will suffice, which can be switched in or out depending upon normal listening or back-ground level. I accept that tone controls can compensate - to a degree - for some ''Cackhanded'' recording imbalance, but rather than that, I am more inclined to ''Bin'' poor recordings than attempt to correct them.

The second issue relates to ''Burr - Brown'' Audio Processing. I understand the company was taken over by ''Texas Instruments'' about 20 years ago. My Yamaha CD player uses it, or so I am told. However, so too does the Zen V2. I have never had the opportunity to listen seriously to any top class D to A Streamers so I cannot fairly compare my system with any. However, what did become apparent to me was the way the V2 processes the LSB - Least Significant Bits, i.e. those bits which contribute the finer detail of musical content, the region where some words of a song could be confused for others, because of lack of definition - sibilance creating confusion. Clearly, the higher the bit rate, the greater the accuracy. However, the ability to resolve high bit rates, constantly is an imperative if one wants to achieve the greatest fidelity.

So, my question is: - What is so special about ''Burr - Brown'' that it seen as superior to other types of audio processing? And indeed, is it?

 

[Arkless Electronics]

The 33 and 303 are not really up to much, though very good in their day, and the 405 has excessive bass roll off. Various mods are available to soup them up somewhat from a variety of companies (when I get a delivery of round tuit's I may do some mods I worked out ages ago on my own pair of 405-2's and see what they are then capable of. I don't use them as I have designs of my own which thrash them. 303 could be modded to sound vastly better also according to CAD simulations of it I made).

Some older B & O amps had variable loudness compensation which adjusted in step with the volume.... unfortunately couldn't be switched off!

Burr Brown were a leading manufacturer of IC's of all varieties but specialised in analogue (low noise op amps etc) and DAC's/ADC's.
Sadly many more companies have been sucked in by bigger ones to form "mega corps" since BB were bought out and the first thing that usually happens is "consolidation" which means loads of fantastic parts are obsoleted over night. TI also bought out National Semiconductor who had just brought out a fantastic range of super low distortion and noise op amps for audio about a year before.... yep TI obsoleted about 2/3rds of the new op amp range within months.

 

[Hodges]

I took delivery of a ALPS RH27B Equal Loudness Control this week-end, having previously bought two others: -

A LG177 - 12 level stepped control, which claimed that as it was 400kohm would not load the preamp.

Another which I bought a few weeks earlier was quite clever, insofar as it used what looked like the playback head of a cassette player to form the inductive element of an LC Loudness control circuit. It was not very successful as it constantly needed adjustment unless one got the output level of the V2 set to match - which made it very fiddly indeed.

The LG177 only really worked at one end of its range and if one wasn't careful had the tweeters screaming. It is very well made but it seems to me, in this instance, ''The sum of the Parts are of Greater Value than the Device as a Whole''.

However, The ALPS RH27B, being of a stepped type, was much more effective. To be frank, I only bought it because of its case and volume control which are a lot better than having to make up some ''Heath - Robinson'' box to fit one of the former in. It is easier to find a setting which restores a fair degree of balance to low and high frequencies at low levels. It certainly helped out at the low frequencies but from the few tests I have done, perhaps the musical balance of the material I used wasn't particularly good. My point being, the treble lift seemed a bit excessive by comparison.

Bass however, seemed to work very well insofar as it was lifted without attendant resonance, to a degree that it seemed better than at normal listening levels. It suggested to me, the Spendors are capable of giving a decent amount more bass response than hitherto. This I believe to be a function of bass lacking in the material I was using.

It suggest, there could well be a need for independent Bass & Treble controls. It also confirms my faith in the 909 as if low frequency gain wasn't there, not only would the bass frequencies be missing, but the ''Tightness the 909 exercises over the loudspeakers at low frequencies would also be lacking. That remains an ''Outstanding Feature'' of the Quad 909.

However, I don't think returning my 33 to service is the answer; well not entirely the answer. I need to investigate, but I am forming the opinion that, were a tone control circuit, similar to that of the 33 be employed, but with bass control slope which is more acute, it would be much more effective in lifting bass response without encroaching upon the lower mid-range. Essentially, a Baxandall with an asymmetric response.

All in all, one could view this exercise as a bit of a disaster, albeit, a not to expensive disaster. However, I view it as having some positive aspects.

One, it re-enforces my faith in the 909.

Two, it has allowed me to see my Spendors in a different light, insofar as clearly they are not lacking in bass response - just simply have not been exploited properly in that region due; I think, to lack of decent bass components and a very average system which hitherto fell of a cliff at low frequencies with very little means of restoring it because of the lack of decent Bass - lift control.

Three, it has provoked even more curiosity in me as to how much more this little combo of mine has yet to give. At my age, that's no bad thing.

 

[Hodges]

I appreciate, this thread is taking on a life of its own insofar as current dumping appears to be taking a back seat. That is only for the moment as I still have to digest a lot of information about how it works. But in the mean time, I want to see how much more I can get from my system by way of small additions. And so, for the moment, I seem to have levitated in the direction of Loudness Control - relevant to me when listening at low volume.

The LG177 Loudness Control, mentioned above, seemed the ideal guineapig, given how ineffective it is over most of its range, until it gets to the point where it causes the tweeters to scream for mercy. It appears to be a very well made piece of junk, constructed using close tolerance components and a very expensive ALPS 12 position rotary switch.

However, the components are tightly packed around the switch and it is difficult to see their values or trace out the circuit diagram. At under fifteen quid. it may well be close tolerance but it still appears its function centres around one frequency. A few rudimentary checks indicates it is essentially a series of band - stop filters, the centre frequency nominally being around where the ear is most sensitive at low level.

Position One of the ALPS is where the device is switched out, followed by 10 identical Band Stop Filters, all of which have identical response. Each section adds 6dd attenuation at the centre frequency of the band stop filter. Rotating the switch, ''Overlays'' each successive filter over the previous one. Position 2 adding 6db attenuation, 11 giving a total of 60db attenuation. Position 12 being infinity. My rudimentary tests were greatly helped by the legend printed on the PCB, albeit very small.

This is where assumption was made on my part, as to what the circuit looks like. Plus, as I was concerned about the potential for damaging the Spendor Tweeters, I decided to look for/at the through - path for the High Frequency component of music/speech. This was found to be - taking just a single filter section, given they are all identical - a series 25kohm/1% with a B102k capacitor in parallel - I am not going to chase up the value as it is, what it is, and at this stage have no intention of modifying the device. The other components are 50kohm/1% + poly caps the value of which I can't tell without removing one from the PCB.

Reasoning, that if I removed the B102K caps, the high frequencies would suffer the same attenuation as the centre frequencies. Leaving the Poly's in place with their associated 50kohm components, surely; I reasoned, would effectively create a ''Low Pass Filter'', with 6db attenuation at the centre frequency. ''I Removed All 20, B102k Capacitors'' and proceeded to test my theory.

The Quad 909 is a powerful amplifier and the Spendor power rating is extremely low by comparison. So much so, that I have fitted 20db. attenuation at its input to ensure it could not overdrive them. So, I reasoned, ''If I fit the ALPS device, switched to position 3, which brings into circuit, three identical filters one overlaying the other - Cascade? - the insertion loss being 18db, is not too dissimilar to the fixed attenuation''.

I digress for a moment to reflect on a Historical Bose Loudspeaker. It was a peculiar shape, with only one front facing driver, several others facing to the rear and angled sides. The Cabinet was relatively small and internally, not only did it have long fibre wool, but steel bracing bars, the purpose of which was to eliminate cabinet resonances. Compensation for the cabinet was in the form of an amplifier with the reciprocal response of the cabinet. They weren't cheap and sounded great. Only large floor standing transmission lines came close to matching the neutrality of those cabinets.

And the connection? Well, with the best will in the world, Spendor loudspeakers are not particularly well braced. But neither were many other loudspeakers of that era. And No Loudspeaker Manufacturer, to my knowledge, specifies the 'Q' of their cabinets - ''Q being the Magnification Factor at Resonance.'' And as the resonant frequency of an enclosure is approached any enclosure with a low 'Q' will start to vibrate - hence the low frequency ''Booming Bass''. So, conventional bass controls, which emphasise, not only bass frequencies, but lower mid-range, will exacerbate loudspeaker resonance. This not only creates a ''one note characteristic bass'', but also encroaches into the lower mid-range frequencies, creating considerable colouration.

Anyone who is sufficiently bored to have followed this, will know the direction I am going. Fitting this provisionally altered LG177, switched to position 3: 18db insertion loss; a provisional setting, results in incredible base performance, devoid of resonance without intruding upon lower mid-range audio. The clarity of instruments in the bass region is outstanding by comparison with what it was hitherto.

Being long retired and having disposed of my test equipment leaves me unable to look at the overall performance for improvement; or otherwise. However, this little device is going to be permanently fitted between my Sugden and 909. True, it no longer has any effect on the higher frequencies when listening at low levels, but that is another problem which needs addressing, separately. Not however, by the use of any Loudness Control. Perhaps a conventional treble control may be the solution, I'm not sure. What I am sure of though. this is the best Bass Control I have ever heard, outside professional circles.

 

[Hodges]

Completed the task this afternoon, the ALPS RH 27B innards being removed and the modified RG177 Low Pass Filter board is now fitted inside the case. The end result does not look out of place with the rest of my set-up. Now for some listening over extended sessions. It seems, other than being in the bypass mode/switched out, the first three settings amounting to 6, 12 and 18db attenuation work well at the levels I am currently listening. It's impact seems to be quite neutral in effect with regard to the music, albeit with the bass response lifted. As an exercise, I really enjoyed it. In pure Audio engineering terms, it is a very rustic solution. However, the devil drives when needs must and I am really delighted at the way the Spendors have responded to the bass lift.

 

[Arkless Electronics]

I've never found the need for any bass lift with my own Spendors.... a fairly neutral sounding speaker in the BBC tradition. Even dub reggae sounds great and this with the speakers around 1.5M from any wall.... and "only" 16WPC.

 

[Hodges]

I've never found the need for any bass lift with my own Spendors.... a fairly neutral sounding speaker in the BBC tradition. Even dub reggae sounds great and this with the speakers around 1.5M from any wall.... and "only" 16WPC.

I understand your point and in truth, at normal listening levels there is a very good balance. Neutral, Transparent, un-phased by most music, vocal or instrumental.

However, at the kind of level I am driving them, they lack quite a lot at bottom and are a bit thin at the top. Strictly speaking, it's not the speakers, but my hearing. This gismo really does beef up the Bass without taxing the cabinets. The components I removed have essentially brought the top end in line with the mid-range so they too could do with a boost, but not to the degree these loudness controls gave. So I would like to fit an independent Treble Control as well. Happily, all this is non intrusive, so it is easy to restore the system to flat - indeed the zero position simply bypasses the LPF with no losses. I wish I could get a plot of this filter to quantify the response but sadly all my test kit has gone.

I'll investigate what can be done about treble, not least because it keeps me occupied and out of the pub. Also, given the number of filters in cascade on the LG177, I may tempted to modify it further for smaller bass boost increments.

However, you are running 16 watts class A, which probably means the bass is as tight as hell.