Yes Nige, you did read me correctly. For he same effect, and less GNFB, the OLgain of the amp will also be lower, thus enabling the OL bandwidth to be wider.
And yes, op amps play a slightly different game, they usually have tremendous OL gain facors, and their OL badnwidth suffers, as it forcse you to use tons of GNFB, which is why I tend to avoid them at all costs. That said, I do use them for secondary functions, like DC servo (if present), protection circuits, display circuits (LM 3915 is a good example), etc., even for some regulation when using my own shunt supply (posted here somewhere).
Regardin the VAS, I always say - just give it lots of miliamps and watch it go. I mean like 20 mA and above, if necessary but it rarely is. By 20 mA, even the most hard case devices have reached their best range (e.g. 2SC3503, 2SA 1382, etc). But also make sure you add a decent sized local cooling fins/cases, since depending on the supply voltage, 20 mA could easily translate to sa 1.1+ Watts of dissipation. Fail to cool it and you will very likely get some thermal induced drift.
If you sould use some well known JFETs like 2SK170 (still available) in a BJT cascode circuit with say 2SC2240, with local FET voltage stabilization at say 10V, let it rip, make it work at a bias of 5 mA per trannie, and you will get an OL bandwidth of 130 kHz plus no problemo. Your slew rate will be ikely upwards of 300 V/uS. The limit I got to thus far is 145 kHz, but by then, the Miller C will definitely start to influence things, so if you want it to be relatively small (i.e. 10 pF or some such), you need to extra attention to the quality of your PCB board and its layout, else you'll run into some trouble. But even with a 22pF Miller cap, that still works out to (5:0.022) 227 V/uS. For the output devices to keep pace, you'd need a CL bandwidth of 350 kHz or some such.
But all this is old stuff, this has been done on many Japanese amps in the late 70ies, when the Slew Rate Wars were full on, by people like Sansui, Kenwood, Pioneer, at al. The only "trick" in it all is to make sure your Miller C cap is the first and determining filter in the lot. However, for the sake of stability, I would strongly advise you make room for small value caps collector to base for the predriver transistors, themy may become very convenient, but in the end, you may not actually even need them.
The way I look at this is that by increasing the system gain the feedback factor is reduced at HF reaching unity at a lower frequency.
The margin from that point to where the phase angle reaches 180 degrees would be increased and the amplifier would be less susceptible to signal or incipient load induced problems.
With regard to crossover distortion it would not be a bad idea to check the standing current and also check d.c. offset when altering the closed loop gain.
With a low GN FB design, you can bias the output stage at 100-130 mA per device no problemo, and that with a wide OL bandwidts takes care of crossover distortion pretty well. Let me add that my output stages have 3 or 4 pairs of ON Semi MJL 3281/1302 200W devices in series/parallel, so in fact the whole amp will run in pure class A by up to aroubnd 1.5 W/8 Ohms. That's where most of us spend over 98% of our home listening time.
Building a high power amplifier has always been a tempting proposition for me. I have a stash of 10,000 uF Sprague Powerlytic capacitors and a couple of transformers (a) a C-core with 47-0-47 ac. secondary windings weight 7.3 kg and (b) an EI core with copper strap with 40-0-40 ac. windings weight 8.9 kg. These have been on my shelf waiting for the day when... the motivation arrived.
Late last year I was the grateful beneficiary of a set of Quad ESL 63 speakers which are a more demanding load than most including mine. I have gotten out of the practice of building up projects and have tended to look for the simple approach to solving problems.
In this case I have a friend who runs a second hand Hi-Fi shop some 10 or more km away. His prices are reasonable and buying from him would be cheaper than building in most instances
My friend is an interesting fellow and has some interesting people who frequent his shop - one is a fellow countryman of yours. It is enjoyable to call in and look over the equipment from bygone days and hear it being played in the shop.
I am retired and the weather here has been lousy this week. I had some other business near my friends shop and so I took the opportunity to call in to see him.
I have read much of your comment on this thread about the wide bandwidth low feedback amplifiers and on Tuesday the equipment being played in the shop was through Hardon-Kardon PM655 amplification and with some Ohm speakers. Not that I have had such experiences before with other combinations I was rather taken by the sound.
When I got home I was intrigued enough to download a service manual from HiFi engine to check out the design.
I found some aspects very interesting - the use of circuit elements comprised of complementary feedback transistor pairs in the voltage amplification and driver stages. Also interesting was the way the HF compensation was managed.
I did observe the PM655 has only one set of power devices - emitter resistor values of 0.22 ohms so it would probably be good for the levels of current you mention for one pair. I suppose that particular model to be due to a question of economics due to larger heat sink requirements and the larger case needed to accommodate multiple pairs. I agree with your thoughts on about these - which would give a better cushion in the " first watt" region whose importance you highlighted.
I also downloaded the service manual for the HK6550 which has a minor change to the HF compensation. I would be interested to know your thoughts in this regard.
Well, if they can pull it off with just one pair of power devices per channel, why not? Manufaturers generally shy away from multiple pairs not so much because of the added cost, more because of the problem of pairing them well.
As you say, my 6550 has one pair, but the 680 has two pairs (85/130W 8/4 Ohms), the Citation 24 has two pairs (100/170W 8/4 Ohms) and the PA 2400 has 4 pairs (170W 8 Ohms). Their and my favortites, Toshiba 2SC3281/2SA1302, although I use Motorola/ON Semi's version, MJL 3281/1302, which are rated not for 150W as the Toshibas, but at 200W. More grunt, and I like grunt.
But H/K does also carry on with their evil practices of using about the worst pots and trim pots you can find anywhere. The easiet way to get it working better is to change them for decent material, and they don't quite excell in filter capacitors, either, more room to improve. But they do excell in load tolerance, they will drive literally anything you care to call a speaker, and the impulse currents they deliver is often prodigious. For example, the 680 (in its time, their top integrated model) will push out over 530W of power into 2 Ohms, something not many can pull off. And that amp has just 12 dB of GNFB, yet still goes up to just under 300 kHz.
There are lessons to be learnt there. These people know what it's all about. 1 will get you 10 that if you were to change the 650's volume pot with something solid, like Alps Blue, it would come on song. That's what happened to me with my 6550, I still have trouble understanding how such a small chage can kick the sound quality up by at least 3 notches, space and air like you wouldn't believe, hell, I still have trouble believing it. IF that doesn't prove Frank's points, I don't know what does. From being a reasonable amp to becoming a top notch player is a gigantic jump.
Not thatthey are the only ones, mind you. My Marantz 170 DC power amd and 1152 DC Integrated amps wre fully refreshed (all electrolytics exchanged or new ones), but the main two filter caps were upgraded from 12,000 (170 DC) and from 15,000 uF (1152 DC) to BC Components 22,000 uF and really came on song. Crank the volume up and nothing changes, the tonality stays the same, only it plays louder, just as it should ideally be. A joy to listen to. Those were also built with lower than usual at their time GNFB factors, although not quite as low as H/K, but then, these were made in March 1978.
The current H/K philosophy calls for a single NPN input diff stage, but the VAS is fully complementary and is so util the very output. My own work follows that pattern, hardly a new one, they (H/K, Marantz, Philips) were using it from the early 80ies. Its best side is that you can use a cascoded JFET input stage with say 2SK170, and in fact, that and its complement do sit in the fully complementary input stage of the PA 2400 from circa 1996.
I strongly recommend you download and carefully look over the schematics for H/K PA 2400 and current integrated 990 amps, you will see a clear line of evolution and will hopefully pick up a few ideas for yourself. It's all very logical.
Got to run now, I'll add a few comments later on in the day.
As you say, my 6550 has one pair, but the 680 has two pairs (85/130W 8/4 Ohms), the Citation 24 has two pairs (100/170W 8/4 Ohms) and the PA 2400 has 4 pairs (170W 8 Ohms). Their and my favortites, Toshiba 2SC3281/2SA1302, although I use Motorola/ON Semi's version, MJL 3281/1302, which are rated not for 150W as the Toshibas, but at 200W. More grunt, and I like grunt.
But H/K does also carry on with their evil practices of using about the worst pots and trim pots you can find anywhere. The easiet way to get it working better is to change them for decent material, and they don't quite excell in filter capacitors, either, more room to improve. But they do excell in load tolerance, they will drive literally anything you care to call a speaker, and the impulse currents they deliver is often prodigious. For example, the 680 (in its time, their top integrated model) will push out over 530W of power into 2 Ohms, something not many can pull off. And that amp has just 12 dB of GNFB, yet still goes up to just under 300 kHz.
There are lessons to be learnt there. These people know what it's all about. 1 will get you 10 that if you were to change the 650's volume pot with something solid, like Alps Blue, it would come on song. That's what happened to me with my 6550, I still have trouble understanding how such a small chage can kick the sound quality up by at least 3 notches, space and air like you wouldn't believe, hell, I still have trouble believing it. IF that doesn't prove Frank's points, I don't know what does. From being a reasonable amp to becoming a top notch player is a gigantic jump.
Not thatthey are the only ones, mind you. My Marantz 170 DC power amd and 1152 DC Integrated amps wre fully refreshed (all electrolytics exchanged or new ones), but the main two filter caps were upgraded from 12,000 (170 DC) and from 15,000 uF (1152 DC) to BC Components 22,000 uF and really came on song. Crank the volume up and nothing changes, the tonality stays the same, only it plays louder, just as it should ideally be. A joy to listen to. Those were also built with lower than usual at their time GNFB factors, although not quite as low as H/K, but then, these were made in March 1978.
The current H/K philosophy calls for a single NPN input diff stage, but the VAS is fully complementary and is so util the very output. My own work follows that pattern, hardly a new one, they (H/K, Marantz, Philips) were using it from the early 80ies. Its best side is that you can use a cascoded JFET input stage with say 2SK170, and in fact, that and its complement do sit in the fully complementary input stage of the PA 2400 from circa 1996.
I strongly recommend you download and carefully look over the schematics for H/K PA 2400 and current integrated 990 amps, you will see a clear line of evolution and will hopefully pick up a few ideas for yourself. It's all very logical.
Got to run now, I'll add a few comments later on in the day.
Typical/standard pots are carbon composition pots.
Carbon comp resistors are well known to cause excess noise, so it may well follow that CC pots suffer the same/similar excess noise, and consequent subjective misbehaviours.
Word has it that the Alps Blue are conductive plastic but I have no proof and have also read that that they are in fact carbon comp ???.
TKD (not TDK) conductive plastic pots are reputed to be very good sonically.
Bourns 10-turn WW pots are reported as sounding particularly good also.
Quality of wiper contact and wiper loading resistance (following stage) are discussed as causing subjective misbehaviours.
diyaudio.com -some-collective-knowledge-about-sound-potentiometers
diyaudio.com -so-i-did-try-all-those-different-potentiometers
diyaudio.com -potentiometer-overview-alps-penny-giles-tkd-vishay-more-whats-your-fav.html
tkd-corp.com/product
Dan.
Dan, AFAIK Alps Blue is indeed a carbon pot, albeit a relatively precise one (but not uncommonly so, error through 270 degrees is +/- 2dB. This does not even begin to compare with mechanical switches using discrete metal film resistors, whose tracking error is claimed to be at just 0.1 dB.
I actually have two such pots, one from Australia, and another from Denmark, the Aussie one uses classic metal firm resistors from Dale, the Danish one SMT resistors. Haven't tried them yet, I'm saving them for ma preamp. 24 clicks each.
I actually have two such pots, one from Australia, and another from Denmark, the Aussie one uses classic metal firm resistors from Dale, the Danish one SMT resistors. Haven't tried them yet, I'm saving them for ma preamp. 24 clicks each.
During each positive half cycle the excess noise is modulated according to the instantaneous signal/current level, ditto the negative half cycle....Duh.
Dan.
bonobos-found-to-be-smarter-than-chimps
Last edited:
The 2 dB error is quoted as worst case, I believe the typical is much better, but of course not nearly as good channel matching as switched alternatives.
JC makes comment on the resistive element substrate material possibly being influential factor in the subjective sound imparted...IME I fully expect this to be the case.
Can you give links to the Aus switched pot ?.
Dan.
Not really, Dan. I lost that HDD the address was on. Look for a DIY company selling Real Time DACs (I have one of those also) and such pots. Something along the lines of DIYAustralia or some such. Sorry.
Not at all, Jonah. You seem to be doing it the way I learnt a lot, by looking how other people dealt with the problems, meaning looking at service manuals. It a good way to learn, pick up info, then perhaps copmbine that info with other little gems I picked up here and there, and especially here, on this forum.
For example, for a long time I was convinced that whatever was acting as my CCS for the input stage was just fine as long as it had a stable voltage and current, including resistors and zeners. That was before I tried the alternatives and sicovered that it was NOT all the same. Then, somebody here did a test of various CCS stages and showed clearly that their CMRR factors were all but the same, and could range from about 20 dB to over 80 dB. You can actually HEAR the difference as one stage being cleaner and clearer than another. THis singular differenece is, of course, lost in the overall sound, but if you can hear it, it must be worth a second look.
Don't get me wrong, I haven't designed many power amplifiers, only 5 or so, and soe were better than others, and my oly redeeming value is that they show a progression from poorer to better. Right now, I am self-engaged on two topologies and I wonder which one will end up sounding better, plus in making m second Otala/Lohstroh amp. I made one in the mid 70ies, but was enough of a cretin to sell it off later, hoping to meet another amp which sounded like it. I never did, that is one of a kind it seems, so now I'm back to making it again as a refence standard.
One problem the Brits had was making the amp use the volume control correctly. That is to work on the best part of it's range which would be above 50 %. Many would see this as underpowered as they were comparing with cheaper amps. Other amps clipped just over half volume. Naim copied the latter. I never saw anyone say they shouldn't !
Recently I upped the gain of my Quad 33 . Very sucessful. Now it will clip at No 7 using a DL110 medium output MC ( 1.6 mV ). Before it would have needed No 12 if it existed. It makes a big difference to use the volume control to it's best. This includes impedance matching.
Rotel uesd double high speed transistors since the early 1980's. I have a RA840 that has. RA820 was single transistor version with a smaller transformer. The RA840 was one I had been given. It was in the garage for years. One transitor was short collector to emitter. It had casued no other damage. Rotels are very robust and sound very good.
Recently I upped the gain of my Quad 33 . Very sucessful. Now it will clip at No 7 using a DL110 medium output MC ( 1.6 mV ). Before it would have needed No 12 if it existed. It makes a big difference to use the volume control to it's best. This includes impedance matching.
Rotel uesd double high speed transistors since the early 1980's. I have a RA840 that has. RA820 was single transistor version with a smaller transformer. The RA840 was one I had been given. It was in the garage for years. One transitor was short collector to emitter. It had casued no other damage. Rotels are very robust and sound very good.
Electronics has always been a hobby interest, the operative word when I did some amplifier repair work for the New Zealand agent of an English brand. In lieu of any payment he gave me a couple of pairs of audio cable and I passed one set on to another friend.
There were no problems with the cables until my friend replaced his amplifier with a Rotel. He later told me the amplifier had burned out the transformer and was replaced under warrantee. The replacement started humming loudly on power up and he hit the off button before this did the same.
He reported this to the agent who replaced the cables and took the suspect ones away.
Last edited:
Many Rotels have a thermal transformer fuse. They sometimes offered a test pin to do a test repair to bypass the fuse.
The funny thing about Quad ESL 63's is that 5 to 8 watt SE valve amps will drive them. No loop feedback would seem to be very bad news with the reactive load. Not so, no boom or weird frequency balance. Very OK on how loud they go. In truth louder is a risk. On paper this seems unlikely. The reality is different. 63's will sort out many valve amps. Most I wouldn't want to own.
Perhaps the tweeters have settled with the OB speakers ? They sound a bit nicer, Radio 3 is excellent. Perhaps to put a 1 st order crossover on the 12 Lta of about 250 uH and make the tweeter 5 kHz second order ( 2u8 and 360 mH ). In fact 360/2.8 and 2.8/360 should work ( Butterworth ). They are ballpark only. Active is better as I don't need to know the drive unit real values and reactive phase diagrames. An analogy. Speaker design old school and passive is like infering a colour balance and it correction and having a number of all or nothing goes to correct it. Active is like setting up a TV. My 3 drive uinits are if you like red the bass, green the mid, blue the treble. Unlike TV I don't dare go for white as the tweeters will last a few seconds. What I can do is pink. The cheap TV I just bought has none of the obvious defects of the past. That is grey changing it's colour cast, barrel distortion and fuzziness. The modern TV has lack of dynamic range and lake of punch mostly which is related to price. My OB speakers still do not come up to that standard. Nor do most. If TV was as diffficult as speakers I wouldn't own one.
I dare say the beefed up 303 clone I must complete should be ideal for driving ESL 63's. I said about fear factor building this amp. So many of the transistors I am using are different pin out. The PCB's looks fine but came without any info. I must proof read the circuit to be sure I have it right. I am running it's output caps and they do sound fine. I take the 33.5 V right up to the speaker. It works fine.
The funny thing about Quad ESL 63's is that 5 to 8 watt SE valve amps will drive them. No loop feedback would seem to be very bad news with the reactive load. Not so, no boom or weird frequency balance. Very OK on how loud they go. In truth louder is a risk. On paper this seems unlikely. The reality is different. 63's will sort out many valve amps. Most I wouldn't want to own.
Perhaps the tweeters have settled with the OB speakers ? They sound a bit nicer, Radio 3 is excellent. Perhaps to put a 1 st order crossover on the 12 Lta of about 250 uH and make the tweeter 5 kHz second order ( 2u8 and 360 mH ). In fact 360/2.8 and 2.8/360 should work ( Butterworth ). They are ballpark only. Active is better as I don't need to know the drive unit real values and reactive phase diagrames. An analogy. Speaker design old school and passive is like infering a colour balance and it correction and having a number of all or nothing goes to correct it. Active is like setting up a TV. My 3 drive uinits are if you like red the bass, green the mid, blue the treble. Unlike TV I don't dare go for white as the tweeters will last a few seconds. What I can do is pink. The cheap TV I just bought has none of the obvious defects of the past. That is grey changing it's colour cast, barrel distortion and fuzziness. The modern TV has lack of dynamic range and lake of punch mostly which is related to price. My OB speakers still do not come up to that standard. Nor do most. If TV was as diffficult as speakers I wouldn't own one.
I dare say the beefed up 303 clone I must complete should be ideal for driving ESL 63's. I said about fear factor building this amp. So many of the transistors I am using are different pin out. The PCB's looks fine but came without any info. I must proof read the circuit to be sure I have it right. I am running it's output caps and they do sound fine. I take the 33.5 V right up to the speaker. It works fine.
I think you just stated the key point why we must have at least some GNFB, Nige. Reactive loads. They can be very nasty and play some awful tricks on an unsuspecting amp. This is what I always bear in mind, I remember how a Densen zero GFNB amp reacted to a particularly nasty load in form of an AR 3 Improved speaker. He just plain hated that speaker and never managed to sound right with it, but did better with more average loads.
At the cost of being ridiculed, I said it and I say it again: so much depends on whar you are using for your outpute devices. Old style MJ15xxx and newer MJ 2xxx trannies, which are medium speed types (Ft >4 MHz) and do not switch on and off as fast as newer types do are in my view much more robust and hard to offset than the newer types. They are far more robust under dverse conditions.
For suspect loads like the Quad ESL, I would always prefer them to newer faster types.
At the cost of being ridiculed, I said it and I say it again: so much depends on whar you are using for your outpute devices. Old style MJ15xxx and newer MJ 2xxx trannies, which are medium speed types (Ft >4 MHz) and do not switch on and off as fast as newer types do are in my view much more robust and hard to offset than the newer types. They are far more robust under dverse conditions.
For suspect loads like the Quad ESL, I would always prefer them to newer faster types.
- Status
- Not open for further replies.