Miklos,
Yes, I think that a divining rod would come in useful if you broke one output device - but then I suspect that if you broke one, you would rapidly find yourself with ALL of them defective.
That is why I put in the SOAR protection.
So, from a practical perspective, ummmmmm this the the (say it in a deep loud voice) "The Amplifier of 100 Output Devices". It is more about a bit of fun, and experiment, and something that turned out a lot better than you would expect.
The previous emails have triggered some deeper thinking on my part about the while front end piece.
- That is good, and I will see what I can learn from this. Learning is good, and new is good, and something different is good.
- I wont let that detract from the whole fun of the thing - 100 "effing transistors" - has to be good and has to be fun just because it is. If you don't get that, then you miss the whole "effing point". 🙂
So if one goes bad, then get the soldering iron out, and a large hot cup of coffee.
Yes, I think that a divining rod would come in useful if you broke one output device - but then I suspect that if you broke one, you would rapidly find yourself with ALL of them defective.
That is why I put in the SOAR protection.
So, from a practical perspective, ummmmmm this the the (say it in a deep loud voice) "The Amplifier of 100 Output Devices". It is more about a bit of fun, and experiment, and something that turned out a lot better than you would expect.
The previous emails have triggered some deeper thinking on my part about the while front end piece.
- That is good, and I will see what I can learn from this. Learning is good, and new is good, and something different is good.
- I wont let that detract from the whole fun of the thing - 100 "effing transistors" - has to be good and has to be fun just because it is. If you don't get that, then you miss the whole "effing point". 🙂
So if one goes bad, then get the soldering iron out, and a large hot cup of coffee.
Well - cascoded LTP for for keeping the input pair in constant power conditions, Wilson mirror - ok, all those comments are valid, but they are still circling aroung the blameless topology, one way or the other.
If you want some real inspiration from Samuel Groner for moving forward - see this article (originally published in Linear Audio in 2011, but "missed" by many designers by some reason):
A new audio amplifier topology with push-pull transimpedance stage - Part 2: Biasing, stability and AC performance | EE Times
Valery,
I need to do some playing, prototyping and thinking. Thanks for your thoughts and triggering yet another train of distraction for me. {not sure I will forgive you for this!}
I can assure you that any prototypes I build WILL NOT use 100 output devices! Too much soldering for what may prove a throwaway experiment.
Regards
Phil
I need to do some playing, prototyping and thinking. Thanks for your thoughts and triggering yet another train of distraction for me. {not sure I will forgive you for this!}
I can assure you that any prototypes I build WILL NOT use 100 output devices! Too much soldering for what may prove a throwaway experiment.
Regards
Phil
Infrared image of the 100 should reveal either a "hot head" or "cold head" out of place.
Make 100 transistors protrude from top of box and have a small fan blowing air by them. They are essentially individual "fins" of a large heatsink.
Vzaichenko, thanks for the link. Is this the topology behind the VHex front end? The article says PSRR is improved over conventional topology. Have you measured the PSRR of these amps and if so what is it? This is really an important advantage if indeed it makes the PSRR better.
Thanks,
X
This is why I suggested looking at the dominant pole before jumping into bigger things (no doubt that Valery's designs are very well thought out) but looking at two pole compensation and seeing how high you can move it up in frequency before stability concerns isn't so drastic! Just a resistor or two and a few caps.
Still, a very fun project. I've thought of similar ridiculousness for something like 1-2w computer speakers.
Still, a very fun project. I've thought of similar ridiculousness for something like 1-2w computer speakers.
Believe it or not - I came to it rather independently. Here is the first time I mentioned this topology in my thread:
http://www.diyaudio.com/forums/soli...e-old-ideas-1970s-ips-ops-17.html#post4515323
Later on somebody has sent me a Revox-B285 schematic (mid '80-s), utilising a current subtraction in controlled CCS, Edmond Stuart reminded me about his excellent Super TIS developments, I remembered about Accuphase-M60 input stage, I've seen long time ago, later on I've read that article by Samuel Groner.
All the pieces of the puzzle came together and took their places. I moved pretty much forward from that point now, adding the "constant power" principle and a few other things to the overall picture. The newer version of VHex+, called BiTurbo will be available in about a month, as well as some new designs (sorry for advertizing 😛).
It's not like something totally new is invented - it's just a matter of internal understanding of some key principles and designing some cool ways to implement them 😉
The Groner's scheme I refered in post #20 ( http://www.diyaudio.com/forums/soli...small-signal-devices-outputs.html#post4740056 ) is not of the "bootstrapped" cascode kind, and so does not keep the input power in constant power conditions. Schemes proposed for this purpose were proposed in France in 1987
L'tage d'entre de l'ampli -3- diffrentiels insolites (Hphastos)
"Blameless" is in the title of this thread and there are many good reasons why it is a firmly established topology.
I refered to figure #50 of Groner's article because it addresses the well known problem of the negative power supply rejection mentionned in post #11.
As an avid reader of Linear Audio since issue #0, I know the scheme proposed by Samuel Groner since five years. I am looking forward a first implementation of it in a power amp. Maybe I'll build it, I already built the amp of figure #56 of his comments.
Last edited:
Hi Forr, you're right - it's constant voltage (almost) there in fact.
Topologies from France you refer to - interesting stuff, as well as some Lavardin schematics, available on the net.
That's fine, I just don't see "blameless" as a "model" for further development for myself (at least for now), mostly because of the VAS drawbacks/limitations. Common base / current driven stages have got a lot of potential, in my view, deserving deeper study / thinking through.
See Ostripper's implementation here:
http://www.diyaudio.com/forums/solid-state/248105-slewmaster-cfa-vs-vfa-rumble-911.html#post4666755
Cheers,
Valery
OK, I bite.
I am intrigued by the simple elegance of the current output from the LTP using "current stealing" to drive a common base VAS. What is on the surface a very simple tweak, but fundamentally transforms the LTP / VAS interface.
A unique aspect of this is that it is really, really easy to make two amplifiers that use conventional or this altered interface.
I will lay out a PCB that as far as is practicable is identical to the current layout, but implement the single ended current output LTP / Common Base VAS to allow a true comparison of how it behaves. (Ref Post 14, single ended version)
I will also use the PCB that allows two pole miller comp, and make some of them too. Depending on enthusiasm.
It will be interesting to compare them. This will probably take the weekend, and if things get busy with family then next week too.
"The Amplifier of 100 Transistors" was always a bad idea...
I am intrigued by the simple elegance of the current output from the LTP using "current stealing" to drive a common base VAS. What is on the surface a very simple tweak, but fundamentally transforms the LTP / VAS interface.
A unique aspect of this is that it is really, really easy to make two amplifiers that use conventional or this altered interface.
I will lay out a PCB that as far as is practicable is identical to the current layout, but implement the single ended current output LTP / Common Base VAS to allow a true comparison of how it behaves. (Ref Post 14, single ended version)
I will also use the PCB that allows two pole miller comp, and make some of them too. Depending on enthusiasm.
It will be interesting to compare them. This will probably take the weekend, and if things get busy with family then next week too.
"The Amplifier of 100 Transistors" was always a bad idea...
Googlyone,
This is turning out to be a very interesting thread. Thanks for agreeing to make the new topology to compare.
This is turning out to be a very interesting thread. Thanks for agreeing to make the new topology to compare.
Member
Joined 2009
Paid Member
I was also thinking along these lines, common base VAS back on my TGM4, and I was certainly not the first 😉
But do remember there's more to explore than LTP.
But do remember there's more to explore than LTP.
The Amplifier of 100 Transistors - Current driven VAS
OK, here we go.
An utterly minimal mod to the amplifier to implement the LTP -> VAS mod to a current driven common base VAS.
I will throw these up here on the off chance that someone looks at them before I pop out to the man-cave and etch some PCB up.
This may not be the perfect layout, but it is as close to the blameless version that I could keep things.
Once the wife works out what I am up to I imagine she will have me committed to an asylum!
View attachment Many_Tiny_Sch_Current_FB_1.pdf
View attachment Many_Tiny_Sch_Current_FB_2.pdf
View attachment PCB.pdf
OK, here we go.
An utterly minimal mod to the amplifier to implement the LTP -> VAS mod to a current driven common base VAS.
I will throw these up here on the off chance that someone looks at them before I pop out to the man-cave and etch some PCB up.
This may not be the perfect layout, but it is as close to the blameless version that I could keep things.
Once the wife works out what I am up to I imagine she will have me committed to an asylum!
View attachment Many_Tiny_Sch_Current_FB_1.pdf
View attachment Many_Tiny_Sch_Current_FB_2.pdf
View attachment PCB.pdf
Last edited:
Looks good. Just be careful with SOA - in your configuration N22 will always see rail to rail and dissipate roughly 2x more than each of P7 and N3.
In my design I use +/-15V rails for LTP, making my life a bit easier (well, makes sense, having the main rails at, say +/-50...75V). In your case the rails are lower in general, but still...
In my design I use +/-15V rails for LTP, making my life a bit easier (well, makes sense, having the main rails at, say +/-50...75V). In your case the rails are lower in general, but still...
Yes, very nice work Googlyone. Your wife probably already thinks you are nuts for designing and making your own amps when you can just buy them. So no worries there. 🙂
If your new current drive topology is confirmed to work with a build - I will make it but will need Gerbers. Would it be too much to ask if the output stages can be made on say identical 10cm X 10cm max size boards with 26 pairs each? That way we can order rather cost effective PCB's by the tens and use same one with headers to connect. The input and driver stage would be a separate board.
Thank you.
Btw, would using a higher power TO126 driver improve things? I am not opposed to KSC3503 and KSA1381's for this purpose. Could use a little local heat sink even.
If your new current drive topology is confirmed to work with a build - I will make it but will need Gerbers. Would it be too much to ask if the output stages can be made on say identical 10cm X 10cm max size boards with 26 pairs each? That way we can order rather cost effective PCB's by the tens and use same one with headers to connect. The input and driver stage would be a separate board.
Thank you.
Btw, would using a higher power TO126 driver improve things? I am not opposed to KSC3503 and KSA1381's for this purpose. Could use a little local heat sink even.
Last edited:
I think, initial idea is to use small signal devices throughout the whole amplifier - if they are used even in OPS, it makes sense to use them in the gain stages as well. In the worst case it's possible to parallel some number of them 😉
I will happily crank out some single sided PCBs to undertake the test tomorrow morning. Before Anne gets out of bed, so I will avoid questions about the sanity of my actions until the boards are etched and drilled.
An interesting consequence of a current drive into a common base VAS is that the input stage can operate from pretty much any rail voltage you want. And drive pretty much any output stage you want.
Yes, the dissipation of the LTP --> VAS current mirror/multiplier/translator (help -what do you properly call this device? A current translation / reflection device?) Does dissipate a bit.
In this implementation it will be about 240mW. A good low Cbe device here would be good. One should be OK for tests, more might be necessary for long term reliability.
Just for now I am sidetracked onto the question of "Does this VAS conversion make a difference?". It is just bloody mindedness and the original premise of this thread that makes me keep using BC549. Else I would be rolling out a good VAS like device🙂
I will know the answer to the big question in the next couple of days, busy family stuff on the the weekend allowing🙂
An interesting consequence of a current drive into a common base VAS is that the input stage can operate from pretty much any rail voltage you want. And drive pretty much any output stage you want.
Yes, the dissipation of the LTP --> VAS current mirror/multiplier/translator (help -what do you properly call this device? A current translation / reflection device?) Does dissipate a bit.
In this implementation it will be about 240mW. A good low Cbe device here would be good. One should be OK for tests, more might be necessary for long term reliability.
Just for now I am sidetracked onto the question of "Does this VAS conversion make a difference?". It is just bloody mindedness and the original premise of this thread that makes me keep using BC549. Else I would be rolling out a good VAS like device🙂
I will know the answer to the big question in the next couple of days, busy family stuff on the the weekend allowing🙂
A thought. That needs to be considered.
In clipping and error conditions, the input differential pair will at times drive (on post 34) N22 (translation / reflection device) very hard. Indeed the full differential pair current will go into the base of this device and turn it on into saturation - or in this case until the voltage across R4 equals the voltage the differential pair current can generate across R3.
Whoa, that is a messy sentence. Talk English Phil.
If N22 has infinite HFE, the current through it will be in about (R4/R3)*Idiffpair. We need to ensure that either:
- N22 can handle the dissipation this requires - 40ma at 30V in this case. Ouch.
- Or a current limiter be included.
The current limiter can actually be simple, a single transistor:
- Base to top of R4
- Emitter to -ve rail
- Collector to base of N22.
Size R4 appropriately, and you can limit the current in N22.
I will load a VAS transistor into N22's location for my tests tomorrow.
In clipping and error conditions, the input differential pair will at times drive (on post 34) N22 (translation / reflection device) very hard. Indeed the full differential pair current will go into the base of this device and turn it on into saturation - or in this case until the voltage across R4 equals the voltage the differential pair current can generate across R3.
Whoa, that is a messy sentence. Talk English Phil.
If N22 has infinite HFE, the current through it will be in about (R4/R3)*Idiffpair. We need to ensure that either:
- N22 can handle the dissipation this requires - 40ma at 30V in this case. Ouch.
- Or a current limiter be included.
The current limiter can actually be simple, a single transistor:
- Base to top of R4
- Emitter to -ve rail
- Collector to base of N22.
Size R4 appropriately, and you can limit the current in N22.
I will load a VAS transistor into N22's location for my tests tomorrow.
Interesting results building this:
View attachment Many_Tiny_Sch_Current_FB_1_asbuilt.pdf
Along with a bagload of output devices. The only omission was C47, left unloaded.
The aim as noted previously was to assess the impact of using a current drive from the input differential pair into a common base VAS.
The initial results were surprising:
- A nice slew of 2nd, 3rd, 4th and 5th harmonics when the output is at -3dB clipping.
- With 2nd at about -60dBc, and the remainder falling into the noise floor.
- Distortion that was extremely constant as a function of frequency
- Noise floor higher than I would have expected (refer following)
Notes and observations:
- The amplifier did just start up -and did not need tweaking to make stable etc.
- I have not tested things like slew rate, as the distortion has me floored. I was expecting better.
- I remain unconvinced about the stability of this in a more subtle / high frequency manner. There was no tendency to burst into transistor smoking oscillation, but I did note what appeared to be intermittent transients on the output of the order of 20mV or so. I could not get a decent trigger on these to take a look, but my gut says there is something going on.
- Getting the input differential pair balanced in current is non trivial. The absence of a current mirror is a significant setback.
- Getting things adjusted required tweaking of a resistor in the differential pair, and adding a pot in the current source (not on schematics).
- The balance seemed to drift with both temperature and time. This points to me to an issue that should be addressed in an inherent aspect of the design, not a potentiometer.
If you don't get the balance on the input differential pair just right, you can get some spectacular output offsets... And this input stage, which uses the differential pair to inject current into N22/R4 as a translator has the potential to get way out of whack - you cant expect "component selection" to get this right for you.
So where does this leave me?
- I am convinced that the high level of distortion is real, I dropped one of the "blameless" units into the test jig and got expected low distortion.
- I am convinced the distortion should be lower - something is wrong. I am hoping finger trouble with a component selection!
- I do want to modify the differential pair to include a current mirror, and can see a simple way to whack in an extra current mirror / multiplier (N22).
- I am pretty sure this will allow a real balance of the current in the input differential pair, whilst amplifying and mirroring the current "up" to the VAS
Like this - or something....
View attachment Many_Tiny_CFB_ALT.pdf
I am interested in your thoughts on:
- WTF might cause the original high level of distortion
- Thoughts on the alternative current mirror
View attachment Many_Tiny_Sch_Current_FB_1_asbuilt.pdf
Along with a bagload of output devices. The only omission was C47, left unloaded.
The aim as noted previously was to assess the impact of using a current drive from the input differential pair into a common base VAS.
The initial results were surprising:
- A nice slew of 2nd, 3rd, 4th and 5th harmonics when the output is at -3dB clipping.
- With 2nd at about -60dBc, and the remainder falling into the noise floor.
- Distortion that was extremely constant as a function of frequency
- Noise floor higher than I would have expected (refer following)
Notes and observations:
- The amplifier did just start up -and did not need tweaking to make stable etc.
- I have not tested things like slew rate, as the distortion has me floored. I was expecting better.
- I remain unconvinced about the stability of this in a more subtle / high frequency manner. There was no tendency to burst into transistor smoking oscillation, but I did note what appeared to be intermittent transients on the output of the order of 20mV or so. I could not get a decent trigger on these to take a look, but my gut says there is something going on.
- Getting the input differential pair balanced in current is non trivial. The absence of a current mirror is a significant setback.
- Getting things adjusted required tweaking of a resistor in the differential pair, and adding a pot in the current source (not on schematics).
- The balance seemed to drift with both temperature and time. This points to me to an issue that should be addressed in an inherent aspect of the design, not a potentiometer.
If you don't get the balance on the input differential pair just right, you can get some spectacular output offsets... And this input stage, which uses the differential pair to inject current into N22/R4 as a translator has the potential to get way out of whack - you cant expect "component selection" to get this right for you.
So where does this leave me?
- I am convinced that the high level of distortion is real, I dropped one of the "blameless" units into the test jig and got expected low distortion.
- I am convinced the distortion should be lower - something is wrong. I am hoping finger trouble with a component selection!
- I do want to modify the differential pair to include a current mirror, and can see a simple way to whack in an extra current mirror / multiplier (N22).
- I am pretty sure this will allow a real balance of the current in the input differential pair, whilst amplifying and mirroring the current "up" to the VAS
Like this - or something....
View attachment Many_Tiny_CFB_ALT.pdf
I am interested in your thoughts on:
- WTF might cause the original high level of distortion
- Thoughts on the alternative current mirror
Last edited:
- Status
- Not open for further replies.