Agreed . Dvv does make a filter unit like that . Choke input fed PSU's look very interesting on the scope . Maybe it could work ? I suspect the choke allows a bit of wriggle room for the AB output stage . How the bus pumping of class D works with that I don't know .
"Bench test" ... ?? Analysis, yes - I've been using Spice for about 10 years, it certainly speeds up things - and gave me insights along the way ...Stand back guys , Frank shifted gears , " bench test" and "analysis", wooha....
Cant recall Seeing a production amplifier with CLC , maybe Accuphase , my understanding only high bias class -a benefits from the choke, where the inductor is large enough to handle the bias . Demian's response is inline with what most designers think about using inductors .
What say you RM ...?
CLC - what about Viola amps, my understanding is that they use such ...?
Last edited:
With the Philips HT setup I looked at doing the filtering completely on the primary side of the transformer, very little done after the secondary winding - and this in fact worked out well.
I used the CRC concept in the gainclone - Spice indicated a definite advantage in using the technique, and I was happy with the result.
No reason it has to be either/or, I'd say its both/and.
@Frank - yeah the Viola is a true choke input supply AFAIK, one huge lump of iron in that.
As many know I remove pieces from amps until they stop working . It should be said amps must not have crossover distortion and be below 0.1% THD at a sensible power output ( 100 watts ) . Usually I get to 0.01% without trying .
Recently reading the Blomley amp I released exactly why I like MOSFET's . They allow a text book class B to work . If you feel rich you can cure the disadvantage compared with bipolar of Ron . If low biased even that is not a problem . 3 mA bias ( ie zero volts ) will allow descent 10 kHz performance . A full 100 mA to make the rest look nice . 20 mA bias might satisfying those who like the bipolar sound ( incisive and slightly metallic ) .
I built a complimentary feedback pair as a buffer . This was done Deadbug which I think was a master stroke . The measurements were better than FET and so was the sound . It was the silky smooth sound I like . So I thought lets put the MOS FET's in the VAS compensation loop . Now I had what measured the same and guess what , they sounded the same . To my astonishment most people fail doing this . I reverted to the feedback pair and tried the VAS to output compensation and to my surprise is wasn't improved or troubled , Deadbug helped I feel . I forget what I used , very fast bits usually for Rotel repairs .
Looking back to Blomely I realized the MOS FET to be a better candidate for applying negative feedback . The Hitachi graphs look identical to the Blomley ones . That is two amps divided by a 0 volts datum line , they don't quite touch the line . I supect why this matters is the 0 V is determined by the LTP and not a resistor divider ( there's a thought ) .
The feedback pair amp will start to conduct at 0.5 V yet have massive current gain . This also is a great starting point . It has extreme good linearity and Ft due to it being enclosed in it's own local feedback loop .
Where did this get me ? By the expedient of simple set up I vote for the MOS FET over feedback pair . I have never made a bias Vbe device I thought worked . Please accept you might not be so lucky and FET's cost money . My VAS was 2SD756 gain 400 .
The MOS FET's ( Exicon ) seem to have at least 10 times the bandwidth ( my guess more ) . The high input capacitance forms a useful pole in the amp . Thus nothing remarkable is seen . I suspect this still is a great asset to have this well controlled speed .One thing we are told not to do is measure the gate whilst doing it's job . I did and saw a beautiful low level sine wave at 5 MHz . If the scope induced it I can not say . Nothing found at the output . My series resistor 220R .Some say the inductance of cheap resistors is a bonus here . When a device seems to self oscillate at 5 MHz it might be stopping this going into the VAS . Typical Cg is 500 pF per device . VAS is usual 7 mA at 55 V . I dare say if it weren't for tweeters only needing 3 watts max this wouldn't be ideal . The measurement do not say my goodness 500 pF is killing the amp . Word of warning , listen if you use more VAS current . I use 5.6 mA . TID is noted by it's absence ( grittiness ) .
The output stage to VAS feedback path seems to clean up the AB residuals . I supect the simplicity of the VAS to FET's feedback helps a lot . By accident of guess I did what Baxendale suggests .
As a clean up for class G it looks worth a try . When I used it the AB residuals were gone . MOS FET's are quite dirty before that , this often is said to be of no consequence when FET . Not true .
The birds nest amp I showed to amuse you guys had this and worked fine . A friend of a friend took one of my amps to produce commercially . He has an Audio Precision measuring station . His version of my amp oscillates and is not nice ( why not copy the PCB I gave ? ) . The whole redesign was by computer . Funny how far I get just doing it logically . You guys saw the silly amp , it never missed a beat . It was built on RS tag strip just to be sure it couldn't work . To be honest it is the only one I built like that that didn't need slugging . I must be learning . Being nearly 60 it will be game over soon . Shame as I was just starting to get somewhere .
Jan republished this .
The Baxandall Papers: Transitional Miller compensation
Recently reading the Blomley amp I released exactly why I like MOSFET's . They allow a text book class B to work . If you feel rich you can cure the disadvantage compared with bipolar of Ron . If low biased even that is not a problem . 3 mA bias ( ie zero volts ) will allow descent 10 kHz performance . A full 100 mA to make the rest look nice . 20 mA bias might satisfying those who like the bipolar sound ( incisive and slightly metallic ) .
I built a complimentary feedback pair as a buffer . This was done Deadbug which I think was a master stroke . The measurements were better than FET and so was the sound . It was the silky smooth sound I like . So I thought lets put the MOS FET's in the VAS compensation loop . Now I had what measured the same and guess what , they sounded the same . To my astonishment most people fail doing this . I reverted to the feedback pair and tried the VAS to output compensation and to my surprise is wasn't improved or troubled , Deadbug helped I feel . I forget what I used , very fast bits usually for Rotel repairs .
Looking back to Blomely I realized the MOS FET to be a better candidate for applying negative feedback . The Hitachi graphs look identical to the Blomley ones . That is two amps divided by a 0 volts datum line , they don't quite touch the line . I supect why this matters is the 0 V is determined by the LTP and not a resistor divider ( there's a thought ) .
The feedback pair amp will start to conduct at 0.5 V yet have massive current gain . This also is a great starting point . It has extreme good linearity and Ft due to it being enclosed in it's own local feedback loop .
Where did this get me ? By the expedient of simple set up I vote for the MOS FET over feedback pair . I have never made a bias Vbe device I thought worked . Please accept you might not be so lucky and FET's cost money . My VAS was 2SD756 gain 400 .
The MOS FET's ( Exicon ) seem to have at least 10 times the bandwidth ( my guess more ) . The high input capacitance forms a useful pole in the amp . Thus nothing remarkable is seen . I suspect this still is a great asset to have this well controlled speed .One thing we are told not to do is measure the gate whilst doing it's job . I did and saw a beautiful low level sine wave at 5 MHz . If the scope induced it I can not say . Nothing found at the output . My series resistor 220R .Some say the inductance of cheap resistors is a bonus here . When a device seems to self oscillate at 5 MHz it might be stopping this going into the VAS . Typical Cg is 500 pF per device . VAS is usual 7 mA at 55 V . I dare say if it weren't for tweeters only needing 3 watts max this wouldn't be ideal . The measurement do not say my goodness 500 pF is killing the amp . Word of warning , listen if you use more VAS current . I use 5.6 mA . TID is noted by it's absence ( grittiness ) .
The output stage to VAS feedback path seems to clean up the AB residuals . I supect the simplicity of the VAS to FET's feedback helps a lot . By accident of guess I did what Baxendale suggests .
As a clean up for class G it looks worth a try . When I used it the AB residuals were gone . MOS FET's are quite dirty before that , this often is said to be of no consequence when FET . Not true .
The birds nest amp I showed to amuse you guys had this and worked fine . A friend of a friend took one of my amps to produce commercially . He has an Audio Precision measuring station . His version of my amp oscillates and is not nice ( why not copy the PCB I gave ? ) . The whole redesign was by computer . Funny how far I get just doing it logically . You guys saw the silly amp , it never missed a beat . It was built on RS tag strip just to be sure it couldn't work . To be honest it is the only one I built like that that didn't need slugging . I must be learning . Being nearly 60 it will be game over soon . Shame as I was just starting to get somewhere .
Jan republished this .
The Baxandall Papers: Transitional Miller compensation
Last edited:
So its trial and play .....? Can you put up a schematic as to your approach,
Last edited:
Pretty much yeah - I'll design around some toroidal cores I've decided to use most probably. I rather like Sendust for this application as it has about the highest inductance (the -125 type that is) meaning lowest copper losses. But simulating doesn't get very far coz frequency dependent losses are a pain in Spice.
PS - while ferrite its great for classD inductors, its the pits for power supply filtering. Use powdered iron. The schematic just has caps to ground and inductors from cap to cap. Can your imagination stretch that far?
PS - while ferrite its great for classD inductors, its the pits for power supply filtering. Use powdered iron. The schematic just has caps to ground and inductors from cap to cap. Can your imagination stretch that far?
Last edited:
Ben Duncan suggests there are 3 mechanisms that might be mistaken .
TID by hitting slew rate limit .
Crossover distortion .
Crossover distortion unmasked by clipping ( slew limiting induced ) .
He also suggests back EMF enters the amp under the latter condition .Others say back EMF of speakers has no significance and is absorbed by diodes etc .
Another effect I read of is Ricochet effect . In sealed rooms where power is high the feedback is room induced via time delay trough the speaker ( a car ) . Connecting amps one to another to simulate the condition was suggested to simulate the condition . I was too young to take it in at the time . The suggestion was sensible amps are not troubled and the fixes are easy . Put speakers face to face seems an idea . Digital delay added and phase .
TID by hitting slew rate limit .
Crossover distortion .
Crossover distortion unmasked by clipping ( slew limiting induced ) .
He also suggests back EMF enters the amp under the latter condition .Others say back EMF of speakers has no significance and is absorbed by diodes etc .
Another effect I read of is Ricochet effect . In sealed rooms where power is high the feedback is room induced via time delay trough the speaker ( a car ) . Connecting amps one to another to simulate the condition was suggested to simulate the condition . I was too young to take it in at the time . The suggestion was sensible amps are not troubled and the fixes are easy . Put speakers face to face seems an idea . Digital delay added and phase .
Pretty much yeah - I'll design around some toroidal cores I've decided to use most probably. I rather like Sendust for this application as it has about the highest inductance (the -125 type that is) meaning lowest copper losses. But simulating doesn't get very far coz frequency dependent losses are a pain in Spice.
PS - while ferrite its great for classD inductors, its the pits for power supply filtering. Use powdered iron. The schematic just has caps to ground and inductors from cap to cap. Can your imagination stretch that far?
Where can I get Sendust in E and I ? I need some . Ideally something better than M6 that is OK on price . I have written to an Indian company hoping for this . I am only using the bottom of the BH curve so might be able to use a wacky solution ( by only I mean usually ) . Saturation isn't an issue . This is work related and a million miles from what we write about here . I can't say more as I have a confidentiality clause I try not to brake . M6 works but isn't small enough . 20 VA typical . I wrote to Edcore , they were stumped . Who builds for me is effectively the old Partridge company . Graham is stumped also . They are mostly MOD these days . I am the customer they dread . We spend a lot of money so I am tolerated .
Never seen Sendust in anything other than toroids. It probably won't have enough inductance anyway if you've been using steel unless your air gaps are relatively huge. Its a fair bit more expensive than steel too. If you need smaller size then look at nano but again the price is going to skyrocket.
Plenty examples, though none US manufacture.
Even in the amp kit scene, CLC power supplies have been available for decades.
Oldy example from the empire : www.albs.de/Spec/Netzteile1_Info.pdf
I have to say that one of the very worst system that I have heard was Spectral cartridge - the TT was a known good one - the power amps were big Jadis Monos driving big Sound Labs electrostatics. It was the Spectral stuff which sucked as the system worked averagely well before swapping out the Pre and Cartridge for Spectral.
The all time worst was again Spectral based! This time all Spectral, MIT cables, driving those massive multi box Wilson Wamms.
One of the best I have heard was the really big Sound Labs (A3?) driven by all Yamamura tube amps (845 PP) tube pre (Two box mono tube stuff). One of the best jazz outfits ever. Putting the owners highly respectable and expensive amps back in the system lost all of the magic.
This was all some 20 years ago.
The all time worst was again Spectral based! This time all Spectral, MIT cables, driving those massive multi box Wilson Wamms.
One of the best I have heard was the really big Sound Labs (A3?) driven by all Yamamura tube amps (845 PP) tube pre (Two box mono tube stuff). One of the best jazz outfits ever. Putting the owners highly respectable and expensive amps back in the system lost all of the magic.
This was all some 20 years ago.
It sounds like you need either pure nickle or even mumetal core materials. Sendust doesn't come in laminations. Another real option would be to get amorphous steel: Amorphous metal - Wikipedia, the free encyclopedia. Its used primarily in power distribution transformers to save money, even though its very expensive. Its also been used for tape heads.
Depending on the frequencies in question you may also need really thin laminations and make sure they alternate on every lam. These are how to make a $20 transformer into a $200 transformer. Needless to say this would be a bobbinless transformer with every layering/winding trick in the book. Do your vendors have the simulation software?
Never had any such experience when listening to Spectral, but i never heard them on ESL., but i have heard Jadis , my experience with them mirrors your Spectral experience ...
Last edited:
Abraxalito, there are other ways to achieve that.
Looking from the PSU board with the rectifier diodes on it,towards the amplifier, I use:
10,000 // 10,000 // 4,700 uF symmetrical for + and - of course)
cut to power amp board, where we find:
100 uF // 3.3 uF // 220 nF // 1 Ohm + 220nF.
The above is true when using Fisher & Tausche caps. No birdies, no RF, no nothing, clean as a whistle.
Looking from the PSU board with the rectifier diodes on it,towards the amplifier, I use:
10,000 // 10,000 // 4,700 uF symmetrical for + and - of course)
cut to power amp board, where we find:
100 uF // 3.3 uF // 220 nF // 1 Ohm + 220nF.
The above is true when using Fisher & Tausche caps. No birdies, no RF, no nothing, clean as a whistle.
- Status
- Not open for further replies.