I have heard about "Memory Distortion" and have read this one
http://www.lavardin.com/lavardin-techE.html
http://peufeu.free.fr/audio/memory/ (overview)
http://peufeu.free.fr/audio/memory/memory-8-complete.html (complete circuit)
http://peufeu.free.fr/audio/memory/memory-6-test.html (advanced second stage - see Fig. 6.2)
and then I thought more accurately about the Pass Aleph versions
http://www.patentstorm.us/patents/5710522.html Pass Aleph active CS United States Patent 5710522
http://www.pat2pdf.org/ (put in 5710522 for download)
http://www.firstwatt.com/downloads/seclassa.pdf
and I note the follow advantages/disadvantages
1) Larvadin models:
- no memory distortion through cascoded stage and sziklai darlington by input diff gain stage and cascode current source
- no memory distortion through double cascoded stage by second voltage gain stage and cascode current source
- cascode buffer stage as predriver for drive the true complementary push pull output power class B follower
- unfortunately no single ended class A output buffer and no class A push pull buffer
- unfortunately current mirror to create single ended output for first stage (I don't like current mirror)
2) Aleph models:
- no Class AB distortion present through single ended Class A output power common source stage
- unfortunately no cascoded version in both stages
- output power stage (no output buffer stage, i.e. voltage and current gain) instead source follower output stage (advantage or disadvantage ? - I don't know really)
Now I think about various mixed circuit designs of both topologies.
The result must be actually the ideal power amplifier, free from all kinds of distortion. Are there already designs in this kind?
I appreciate about your comments and information.
Thank you very much
http://www.lavardin.com/lavardin-techE.html
http://peufeu.free.fr/audio/memory/ (overview)
http://peufeu.free.fr/audio/memory/memory-8-complete.html (complete circuit)
http://peufeu.free.fr/audio/memory/memory-6-test.html (advanced second stage - see Fig. 6.2)
and then I thought more accurately about the Pass Aleph versions
http://www.patentstorm.us/patents/5710522.html Pass Aleph active CS United States Patent 5710522
http://www.pat2pdf.org/ (put in 5710522 for download)
http://www.firstwatt.com/downloads/seclassa.pdf
and I note the follow advantages/disadvantages
1) Larvadin models:
- no memory distortion through cascoded stage and sziklai darlington by input diff gain stage and cascode current source
- no memory distortion through double cascoded stage by second voltage gain stage and cascode current source
- cascode buffer stage as predriver for drive the true complementary push pull output power class B follower
- unfortunately no single ended class A output buffer and no class A push pull buffer
- unfortunately current mirror to create single ended output for first stage (I don't like current mirror)
2) Aleph models:
- no Class AB distortion present through single ended Class A output power common source stage
- unfortunately no cascoded version in both stages
- output power stage (no output buffer stage, i.e. voltage and current gain) instead source follower output stage (advantage or disadvantage ? - I don't know really)
Now I think about various mixed circuit designs of both topologies.
The result must be actually the ideal power amplifier, free from all kinds of distortion. Are there already designs in this kind?
I appreciate about your comments and information.
Thank you very much
Hi
Having very similar thoughts - I want to try a single ended (half of an) F5 with cascoded first stage. Alas no time for a couple of months, but eager for any input from the experts.
Cheers
cv
Having very similar thoughts - I want to try a single ended (half of an) F5 with cascoded first stage. Alas no time for a couple of months, but eager for any input from the experts.
Cheers
cv
No idea what memory distortion might be, but if you like cascoded amps, look at Nelson's Threshold amps.
Have fun, Hannes
Have fun, Hannes
I've been thinking about how to reduce such things for a while. One easy way is to spend many times more $$$ on devices with very low temperature coeifficients like metal foil resistors (2-5ppm). Other ways sometimes might include using over rated parts
By that I mean using a 2W device where only 1/4 may be necessary. Also, designing the circuit so that under operation the signal range is a very small pecentage of the whole. I guess that's another
Basically build a circuit that swings 30V and 100mA but only use it for 3V and 5mA 😉 lots of Headroom 😀 😀 😀


I owned the Lavardin IT amp a few years back. It was unusal in that it had a midrange that was very natural and real...valve like. It was also very musical and had a good bottom end. It did lack top end extension and air despite numerous equipment changes.
But yes i do remember it well as one of the best solid state amps that i had lived with.
But yes i do remember it well as one of the best solid state amps that i had lived with.
flg said:Also, designing the circuit so that under operation the signal range is a very small pecentage of the whole. I guess that's anotherBasically build a circuit that swings 30V and 100mA but only use it for 3V and 5mA 😉 lots of Headroom 😀 😀 😀
I've been learning the same thing in my simulations. It makes for a very smooth sound...
audiojoy said:I owned the Lavardin IT amp a few years back. It was unusal in that it had a midrange that was very natural and real...valve like. It was also very musical and had a good bottom end. It did lack top end extension and air despite numerous equipment changes.
But yes i do remember it well as one of the best solid state amps that i had lived with.
"It did lack top end extension" ... typical effect of complex distortion (crossover distortion mainly create through only 20-50mA in the output stage. This distortion mostly audible above 2000 - 3000 Hz.
"Basically build a circuit that swings 30V and 100mA but only use it for 3V and 5mA lots of Headroom ......."
This is one of the main reason why in most cases tube amplifiers sound better than such with transistors (solid state amplifiers)
tiefbassuebertr said:This is one of the main reason why in most cases tube amplifiers sound better than such with transistors (solid state amplifiers)
This would be true if the tube amplifiers were operated that way, but
when you look at real power amplifier circuits, this causes such
inefficiency that power stages must be operated to the extremes
of voltage and current. They do not have the luxury of staying in
a small space on the load-line range.
😎
That particular mode I mentioned is ussually difficult to achieve at the output. You would be getting unbelievably low efficeincy and have more heat than sound, etc. etc... Although N.P. has a Pat. along those lines that improves the situation.
It's not to bad in a preamp (VAS) circuit though. I'm getting ready to play with another proto of such a device. Burns almost 50W a channel
But, the real difficulties are in the semiconductors. We don't really have all that much choice of die size or Pd and linear operating points vs desireable input characteristics.
I'm not ready to talk about my doo-dad but I'll tell you it's a VAS kinda circuit with an LU1014D gainstage on steroids 😀 Maybe a cross between a BOZ and a ZV8. I'm hopping to drive an F4 with it

It's not to bad in a preamp (VAS) circuit though. I'm getting ready to play with another proto of such a device. Burns almost 50W a channel

But, the real difficulties are in the semiconductors. We don't really have all that much choice of die size or Pd and linear operating points vs desireable input characteristics.
I'm not ready to talk about my doo-dad but I'll tell you it's a VAS kinda circuit with an LU1014D gainstage on steroids 😀 Maybe a cross between a BOZ and a ZV8. I'm hopping to drive an F4 with it



IMHO, about the tube designs dealing with clipping, and headroom, is the "softness" of the clipping and the Spectra of the furrier(You Know Who) you end up with (among other factors). Sand stuff ussually drives right up/down to saturation and turns your music into Jimmy Hendrix momentarily(almost square waves). You could have unmeasureable THD at 5W and 25% THD at 10W(Square waves are 33% THD). Music easily contains that dynamic of a range but finding a way to include all the dynamics of music into a "test", we still seem to be working on...




this is particularly pronounced in all OTL tube amps - the Graaf GM-20 is such a device, that I know from a friend of me - go toNelson Pass said:
This would be true if the tube amplifiers were operated that way, but
when you look at real power amplifier circuits, this causes such
inefficiency that power stages must be operated to the extremes
of voltage and current. They do not have the luxury of staying in
a small space on the load-line range.
😎
http://www.diyaudio.com/forums/showthread.php?threadid=142528
for the schematic, that I have created, and some comments.
In the original condition the quiescent current is 300mA through each triode 6C33C (150VDC from anode to cathode), i.e 4x45W idle dissipation and only 2x 10 to max 15 watts undistorted output power at 8 ohms.
However, by normal hifi sonic volume a normal Class AB transistor amplifier (even Mark Levinson or Bryston) sounds by high frequencies like scratch mashine resp. circular saw in action by AB comparison. Only solid state power amplifiers with "Single Ended" output stage sounds like a GM20. Class AB are ideal for the lower frequencies below arround 400-500 Hz.
Cause the bad realibility I cannot recommend to buy the GM20 (despite the very good sound quality), because it is nessecary to redesign the complete circuit to enhance the realibility (e. g. only max. 90 VDC instead 150VDC and so on)
Concerning the efficiency is the very heavy weight model TVA1 (TVA-1) from the old company "Michaelson and Austin" (later "Papworth") exactly the contrary from Graaf GM20 - i. e. high power output (approximately 2x100W / 8 ohms, and comparatively low idle power (4x12,5 W, 4x25mA). It also sounds better than all Class AB transistor amplifier, that I know, especially by upper frequency range, but not quite as pure and clean as transistor power amplifiers with "single-ended"output stages. This tube power amplifier outperform clearly the No 23.5 (both amps I have for service at the same time - some years ago.
Schematic of me - made a few years ago - is here as a PDF
Attachments
Mmmmmm
Well, it's easy enough to alleviate memory distortion with proper
design and money, both of which are routinely lacking.
😎
Well, it's easy enough to alleviate memory distortion with proper
design and money, both of which are routinely lacking.
😎
I now doubt also the existence of memory distortion. But nevertheless the Lavardin model "IS Reference" is one of the best sounded amp with used low idle current through the output (i. e. Class AB with 30mA quiescent current).Mmmmmm
Well, it's easy enough to alleviate memory distortion with proper
design and money, both of which are routinely lacking.
😎
This I have experienced while a sound check after replace the main caps.
What about the circuit topology by VAS in fig. 8-2 about
Memory Distortion Philosophies - Part 8 : More tests
Perhaps the fusion of that topology with that one from Aleph isn't the badest idea - especially in cases, where an extremly sweet sounded amplifier for high frequency and high efficiency horns is wanted.
Attachments
Last edited:
just a note... there was a discussion in the Hafler DH200(long thread) about memory distortion (Fab, I think) with some details about implimentation on a DIY level...
Regards All, Elwood
Regards All, Elwood
I think, you mean post #22 aboutjust a note... there was a discussion in the Hafler DH200(long thread) about memory distortion (Fab, I think) with some details about implimentation on a DIY level...
Regards All, Elwood
http://www.diyaudio.com/forums/solid-state/31131-hafler-dh-200-220-mods-3.html
There this phenomenon was descriptive only regarded to the differential amplifier at the input and not regarded the VAS.
If the open loop gain factor is high, then the voltage swing at the input stage is very low and thus the advantage of this modification rather of academic interest and less for audible sonic quality enhancing.
Last edited:
in this case this thread is of interest, because the approach there goes in the direction to my consideration:
http://www.diyaudio.com/forums/solid-state/204019-two-stage-amplifier-mosfet-class.html
http://www.diyaudio.com/forums/solid-state/204019-two-stage-amplifier-mosfet-class.html
The really absurd part is when Lavardin bases a hypothesis upon
his 300B having no memory. Are we forgetting Iron? Hysteresis?
his 300B having no memory. Are we forgetting Iron? Hysteresis?
Last edited:
You want to defeat thermal memory in semiconductors? Then use
a quadrature feedback from constant temperature current sensors
to control and shape bias in real time. Actively shape for whatever
class you desire, and thermal memory ceases to be a problem.
Constant temp current sensors need then be compared to constant
temp references. Here too, it often goes wrong...
Its class AB circuits that abuse hot variable output stage emitters
across a fixed or slow tracking voltage reference as the only form
of quadrature (common mode current) feedback that suffer worst.
Stop trusting variable temperature devices to shape crossings!
The emitter drop in Aleph is local example how to do this right.
The MOSFET doing all the hot work is not in control. The BJT at
constant current and voltage is making those critical decisions.
No thermal memory here, problem solved...
a quadrature feedback from constant temperature current sensors
to control and shape bias in real time. Actively shape for whatever
class you desire, and thermal memory ceases to be a problem.
Constant temp current sensors need then be compared to constant
temp references. Here too, it often goes wrong...
Its class AB circuits that abuse hot variable output stage emitters
across a fixed or slow tracking voltage reference as the only form
of quadrature (common mode current) feedback that suffer worst.
Stop trusting variable temperature devices to shape crossings!
The emitter drop in Aleph is local example how to do this right.
The MOSFET doing all the hot work is not in control. The BJT at
constant current and voltage is making those critical decisions.
No thermal memory here, problem solved...
Last edited:
maybe this will work?
Hi
let me rewarm the discussion about memdist.
Please look at attached simulations, I think this idea is worth building (K170s should be used instead of BF245B):
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 2.099e+01 1.000e+00 179.92° 0.00°
2 2.000e+03 7.876e-05 3.752e-06 121.23° -58.70°
3 3.000e+03 6.382e-06 3.040e-07 -139.59° -319.52°
4 4.000e+03 1.457e-06 6.941e-08 -0.77° -180.69°
5 5.000e+03 2.902e-07 1.382e-08 62.42° -117.50°
6 6.000e+03 1.213e-07 5.780e-09 95.36° -84.56°
7 7.000e+03 8.862e-07 4.221e-08 -66.55° -246.47°
8 8.000e+03 2.464e-07 1.174e-08 29.63° -150.29°
9 9.000e+03 1.022e-06 4.868e-08 -54.24° -234.16°
Total Harmonic Distortion: 0.000377%
...
What about the circuit topology by VAS in fig. 8-2 about
Memory Distortion Philosophies - Part 8 : More tests...
Hi
let me rewarm the discussion about memdist.
Please look at attached simulations, I think this idea is worth building (K170s should be used instead of BF245B):
Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 2.099e+01 1.000e+00 179.92° 0.00°
2 2.000e+03 7.876e-05 3.752e-06 121.23° -58.70°
3 3.000e+03 6.382e-06 3.040e-07 -139.59° -319.52°
4 4.000e+03 1.457e-06 6.941e-08 -0.77° -180.69°
5 5.000e+03 2.902e-07 1.382e-08 62.42° -117.50°
6 6.000e+03 1.213e-07 5.780e-09 95.36° -84.56°
7 7.000e+03 8.862e-07 4.221e-08 -66.55° -246.47°
8 8.000e+03 2.464e-07 1.174e-08 29.63° -150.29°
9 9.000e+03 1.022e-06 4.868e-08 -54.24° -234.16°
Total Harmonic Distortion: 0.000377%
Attachments
You want to defeat thermal memory in semiconductors?
You could use semiconductors with low or zero temp coefficient. A number of
Lateral Mosfets and (for example) Sony Vfets fit this description easily at
reasonable current values.
😎
- Status
- Not open for further replies.
- Home
- Amplifiers
- Pass Labs
- Aleph (Single Ended) + Larvadin (no Memory Distortion) = Ultimate Aleph?