Yes I did change all the transistors, they had all been previously replaced with different brands, probably some fakes and were almost all cooked. I went with the original references (except BC639 and BC640, I replaced those with BD139 and BD140).
I tried the 33pf across output and negative input of IC3 and it did look a little better but still not perfect.
I don't have enough caps to try it on the outputs (between base and collector) unless I can put 100pF directly.
I tried the 33pf across output and negative input of IC3 and it did look a little better but still not perfect.
I don't have enough caps to try it on the outputs (between base and collector) unless I can put 100pF directly.
You are asking the right questions. Keep trying and use your ears. Marginally unstable can sound better. Remember a sine wave overcomes charging qualities of devices. After one cycle it could measure better? Music is not the same when it matters.
The standard Darlington output stage would give 7 watts 8 ohms if this set up. It usually would have a feed-forward resistor to speed up the discharge of the TIP2955/3055 or whatever used. It also gives some feed-forward current. There are other way which Douglas Self takes time to explain. The simplest being the best with shared 200R resistor between drivers that swing below the cut off points. The D Self complimentary feedback output pair ( Cfbp ) also has discharge resistors. When my clone idea I used 100R. Both Self Cfbp and Darlington are gain of 1. The Brio takes the Cfbp and gives it gain. It's almost impossible to get a good marriage of op amp and Darlington, amazingly the Brio makes it no worse. The Cfbp has it's origins in the H C Lin amp of 1957. When that Darlington top Cfbp bottom. The 1967 French Gogny amp being a wonderful example of that with long tail pair input. 50 watts 1 ohm for ribbon tweeters. Class AB using 2N3055 driving 2N3055.
The standard Darlington output stage would give 7 watts 8 ohms if this set up. It usually would have a feed-forward resistor to speed up the discharge of the TIP2955/3055 or whatever used. It also gives some feed-forward current. There are other way which Douglas Self takes time to explain. The simplest being the best with shared 200R resistor between drivers that swing below the cut off points. The D Self complimentary feedback output pair ( Cfbp ) also has discharge resistors. When my clone idea I used 100R. Both Self Cfbp and Darlington are gain of 1. The Brio takes the Cfbp and gives it gain. It's almost impossible to get a good marriage of op amp and Darlington, amazingly the Brio makes it no worse. The Cfbp has it's origins in the H C Lin amp of 1957. When that Darlington top Cfbp bottom. The 1967 French Gogny amp being a wonderful example of that with long tail pair input. 50 watts 1 ohm for ribbon tweeters. Class AB using 2N3055 driving 2N3055.
Well this wasn't completely for nothing because I learned something (thank you Nigel Pearson) but almost. I jut now noticed (wanting to test maximum power output) that the connectors to my 8ohm dummy loads (yes both of them) weren't making contact so I was scoping without any load. With the 8ohm load correctly connected the square wave looks actually quite good, although the rise time isn't that great compared to other amps I've worked on but I assume this isn't that surprising with this amp.
We all do it. I was much more alert at your age and get worse with age. You are one of the few who questions these things, well done.
In my experience the Brio type is far better than would be imagined. I don't think it would impress Jean Hiraga much. Mr Hiraga liberated my thinking circa 1980. He approximately said the ear can not hear distortion if the relationship of harmonics is the same as music. That is 2nd higher than 3 rd and so on with exponential decay. Second can easily be 1% and yet totally inaudible. A valve design I built respected this ideal and sounded just like live music. It had almost no negative feedback and what is did have was minute cathode degeneration. It had feed-forward distortion cancellation by curve matching. It was 1% THD 6 watts and 0.2% 1 watt. 15 Hz to 50 kHz - 3db 25 Hz to 30 kHz - 1 dB. It could not produce a square wave above 2 kHz. In some ways it out performed a commercial design using 300B ( I lie, it was far better ), it used EL34. What the amplifier gave was layers in the music seldom heard. Even with modern speakers it was fine and ideal with Quad ESL63. My design is nothing like the RH34 as I built one and wanted more. RH84,34,88 etc are very good projects and cheap to do. I am not a valve fanatic except the SE types. I don't think 300B is a good choice.
Hiraga said approximately that a very low distortion design with harmonics unlike music will sound distorted even if the percentage is very low. Negative feedback risks that. What happens is distortion that should not be there is not corrected whilst more benign distortion that would have made a better spectrum is removed. If a bottle of wine what has been made is vinegar with little fruit. Class AB is not easy. I suspect 1 watt of class A is a possible to overcome this. Douglas Self would say it's an abomination to do that. I am not so sure. 5 watts would be better, alas far too hot.
Before you think Hiraga a fool you should factor in the distortion of typical microphones and loudspeakers. 2% would not be unusual. However most have OK spectra. The Quad ESL would be about 0.1% and some microphones also. The Quad ESL has no real need for damping factor which helps. However my amp always sounded very impressive in the bass with them. I could imagine the Brio would be OK with them. The old test was 8R+2uF. That's very nasty.
In my experience the Brio type is far better than would be imagined. I don't think it would impress Jean Hiraga much. Mr Hiraga liberated my thinking circa 1980. He approximately said the ear can not hear distortion if the relationship of harmonics is the same as music. That is 2nd higher than 3 rd and so on with exponential decay. Second can easily be 1% and yet totally inaudible. A valve design I built respected this ideal and sounded just like live music. It had almost no negative feedback and what is did have was minute cathode degeneration. It had feed-forward distortion cancellation by curve matching. It was 1% THD 6 watts and 0.2% 1 watt. 15 Hz to 50 kHz - 3db 25 Hz to 30 kHz - 1 dB. It could not produce a square wave above 2 kHz. In some ways it out performed a commercial design using 300B ( I lie, it was far better ), it used EL34. What the amplifier gave was layers in the music seldom heard. Even with modern speakers it was fine and ideal with Quad ESL63. My design is nothing like the RH34 as I built one and wanted more. RH84,34,88 etc are very good projects and cheap to do. I am not a valve fanatic except the SE types. I don't think 300B is a good choice.
Hiraga said approximately that a very low distortion design with harmonics unlike music will sound distorted even if the percentage is very low. Negative feedback risks that. What happens is distortion that should not be there is not corrected whilst more benign distortion that would have made a better spectrum is removed. If a bottle of wine what has been made is vinegar with little fruit. Class AB is not easy. I suspect 1 watt of class A is a possible to overcome this. Douglas Self would say it's an abomination to do that. I am not so sure. 5 watts would be better, alas far too hot.
Before you think Hiraga a fool you should factor in the distortion of typical microphones and loudspeakers. 2% would not be unusual. However most have OK spectra. The Quad ESL would be about 0.1% and some microphones also. The Quad ESL has no real need for damping factor which helps. However my amp always sounded very impressive in the bass with them. I could imagine the Brio would be OK with them. The old test was 8R+2uF. That's very nasty.
I think this is what mjona referred to earlier. In this application, the 072 opamp is right at its gain-bandwidth limit for 20kHz signals and will normally show sloped and rounded waveforms for squarewaves before that point. Op amp Characteristics
Try MC33078 if you can test as you have. It is slightly more forgiving against NE5532 and I suspect sounds better. It's cheap and sometimes better than any other choice. 16MHz device. OPA2604 was a drop in replacement for TL072 and the others. These days I wouldn't know. I always use DIL sockets because these questions always come up.
CFP power amp. design
I came across this link recently: (https://s9f05a937bddab529.jimconten...dule/6674127354/name/lacav8-supplement-v1.pdf) There are some interesting comments on the Brio's and similar CFP topology by member Ian Hegglun there and it is supplementary material for his article in Volume 8 of Jan Didden's publication "Linear Audio". That series of bookzines so far, is an amazing collection of current audio authors with topics of interest for both DIY and professional folk. They're not cheap with postage from Jan's location, but I can highly recommended them for anyone serious about our hobby.
The point of this link is to give some introduction, a little scope and a starting point for own experiments with some clear indications of the benefits and problems here. It's an unusual design choice but obviously not so bad as some might assume - perhaps a lot better than we might think if we actually get down to trying the ideas out. Happy tinkering
I came across this link recently: (https://s9f05a937bddab529.jimconten...dule/6674127354/name/lacav8-supplement-v1.pdf) There are some interesting comments on the Brio's and similar CFP topology by member Ian Hegglun there and it is supplementary material for his article in Volume 8 of Jan Didden's publication "Linear Audio". That series of bookzines so far, is an amazing collection of current audio authors with topics of interest for both DIY and professional folk. They're not cheap with postage from Jan's location, but I can highly recommended them for anyone serious about our hobby.
The point of this link is to give some introduction, a little scope and a starting point for own experiments with some clear indications of the benefits and problems here. It's an unusual design choice but obviously not so bad as some might assume - perhaps a lot better than we might think if we actually get down to trying the ideas out. Happy tinkering
Last edited:
Thanks Ian , I found some of the same things when I built a version. I assumed the thermal stability was because I was slightly under-biased ( 85% ). It seems not so. This was a sub woofer type application. Distortion was already excellent. I played with the output gain a lot and think the Texan was set a bit high. Three to four seems about right. Past a certain point a conventional design makes more sense. The Texan ( Brio ) is that point. I guess if one of Douglas Self's Cfbp designs was adapted to this way of working it would make no sense. I suspect it would also be at about an output stage gain of 4 if the sweet spot found. One has to suspect higher order distortion would slightly increase. The big deal with Cfbp ( NPN,PNP - PNP,NPN ) when at unity gain they have I guess 100% local negative feedback. When I measured the same in class A distortion is very nearly zero on equipment measuring to - 80 dB. I think 0.8% mentioned open loop for some Texan types. That's about the same as simple emitter followers from memory. Whilst it's not great it has been good enough to make well thought of amplifiers. Saying the Cfbp was perhaps 0.03 % THD open loop is not exactly true as it has it's local loop. All the same that's a nice thing to have as it has the same sort of current gain as a Darlington.
One day when brave enough I plan to build a 320V rail version. The only real problem I see is a P type device. A better idea is two 200V versions in bridge. I want 115 Vrms. If we say the op amp can give 8 Vrms that still is inside the Texan design limits (60/8 = 7.5 ). For a low current MJE340/350 could be used for 200V. The op amp would be between 2 x 15 V zeners. The load I have in mind is 1/4 amp peak. More than ever protection diodes the the output devices if so. 1N4007 might be OK if only 1/4 amp peak.
One day when brave enough I plan to build a 320V rail version. The only real problem I see is a P type device. A better idea is two 200V versions in bridge. I want 115 Vrms. If we say the op amp can give 8 Vrms that still is inside the Texan design limits (60/8 = 7.5 ). For a low current MJE340/350 could be used for 200V. The op amp would be between 2 x 15 V zeners. The load I have in mind is 1/4 amp peak. More than ever protection diodes the the output devices if so. 1N4007 might be OK if only 1/4 amp peak.
Ian, you are right. On paper it could work for my example, only just. A better idea might be to use MJE15034/35G for all functions. These would be both driver and output to give circa 55V rms at 0.25A. Gain is very good at this level, it tails off at 1A. 350V and low price. To be frank a nice device for any use. 30 MHz also. If ever a paire of transistors need to be in the useful box these do. I seem to remember people think they sound good. That is the people who dare think that way. T0220FP would suit me, these are standard T0220.
A friend said FET's. I'd rather not. The Brio/ Derek Skinner amp is enough of a mystery without a drain driven load added. That is unless someone has done it? The 1980 JLH did do it using BUZ900/905 type devices. I would need to use IRF/IRFP ones. I did see the FET example in the analysys PDF. It had the look of a concept design.
Valhalla for a 401.
http://www.farnell.com/datasheets/1918751.pdf
A friend said FET's. I'd rather not. The Brio/ Derek Skinner amp is enough of a mystery without a drain driven load added. That is unless someone has done it? The 1980 JLH did do it using BUZ900/905 type devices. I would need to use IRF/IRFP ones. I did see the FET example in the analysys PDF. It had the look of a concept design.
Valhalla for a 401.
http://www.farnell.com/datasheets/1918751.pdf
Aha!...I thought as much. A friend...well actually, a Naim-aholic friend, was having this AC motor drive installed in his Sondek back in the 1980s when I was making plans to buy a Sondek myself. The used price of a standard LP12 with entry level MC cartridge/arm was about AU$1,000. That was roughly equivalent to ₤490 then - not a decision to take lightly.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.