"Triodity" to cancel loudspeaker distortion

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
THD368Hznormalfas.jpg

THD368Hzfasvänd.jpg

IXTP32-sourcefollower.jpg

Based on Nelson Pass excellent articles about SITs and Triodes I made a little experiment.

With my small IXYS IXTP32P05T based source follower I tested the distortion canceling effects of using the positive going second harmonic to oppose the Bl - and suspension asymmetry of a loudspeaker driver. In this case a cheap 5 inch car coaxial driver in an open baffle.

It is a way to "work the complex load line" of the amp and loudspeaker together.

At 368 Hz the second harmonic dropped from 0,299% to 0,046% with this simple method. I can easily change the phase of the second harmonic with changing polarity of the speaker wire and at the same time flip the phase with my MiniDSP. Since the loudspeaker varies a lot more in its nonlinear behavior with frequency compared to the amp, the effect of this cancelation will also vary a lot with frequency - as seen in the distortion graphs above.

I must say that it is a most worthwhile experiment to do. REW is free and an Umik is cheap, so there is a lot of potential tuning of the harmonic profile of the combination of loudspeaker and amp to be had.

I would recommend everyone who is interested in this to download those two distortion graphs into a new folder so that you can switch fast back and forth between them. It is much easier to see the effects of this experiment that way.


Cheers,
Johannes
 
8Volt.jpg

10Volt.jpg

12Volt.jpg

15Volt.jpg

Here is the harmonic profile and distortion levels for different amounts of current through the IXTP32P05T and the total voltage from the power supply.

By varying the voltage from the power supply and the distortion profile of the amp it is easy to optimize the distortion canceling effect of the opposing asymmetry to whatever measures best or sounds best to you.

It is kind of a cheap mans second harmonic adjustment similar to the one of the First watt SIT1.
 
Thanks Nelson Pass! :)

One volt DC-offset across a speaker!!! :eek:



The four different distortion measurements above is starting from the top:

(volts across the amp and current through the mosfet)

8 volts and 0,35 ampere.

10 volts and 0,53 ampere.

12,2 volt and 0,74 ampere.

15,6 volts and 1,05 ampere .

All the measurements are with the positive going second harmonic reinforcing the asymmetric second harmonic of the loudspeaker.
 

PRR

Member
Joined 2003
Paid Member
> One volt DC-offset across a speaker!!!

1V in 3.2 Ohms is 0.312 Watts.

Car speakers are built to take most of 16 Watts (kids playing LOUD).

You are not even 2% of the way to disaster.

There are several sources of 2nd in a real speaker (an ideal moving-coil speaker would have none, though this is impractical). One big cause is a short coil (midrange) assembled off the center of the magnetic gap (lengthwise). The real fix is to un-do all the glue and assemble the cone/coil in the magnet properly, either shimming or trimming the spider. Obviously a stray Volt of DC is a lot faster, and you can play the speaker while adjusting the DC for maximum output at minimum THD.
 
Very interesting!

Apparently the phase of the second harmonic is everything.

There might be useful ideas here as well (by Eduardo de Lima):

AUDIOPAX - Technical Articles

But what are the chances that we can get this complete distortion cancellation, as it is normally called ? Very few, just like there are very few chances that the distortions will fully add so we would have 2.0%. If the difference between the phases of the 2nd harmonics is between 0° and ± 120° there will be an increase in the total system distortion compared to the distortion of one device alone up to a maximum of 2.0%. If the difference is between ± 120° and ± 180° there will be a reduction in the total system distortion down to a minimum of zero! Of course the audibility of different levels of distortion may not follow the same relation as these numbers suggest, but there is a great possibility that the reduction may be more noticeable than the increase.
 
Alas the distortion of a speaker is primarily related to the cone position (ie due to suspension forces and voicecoil position), not the voltage across the terminals, and the relationship between voltage and cone position varies substantially across resonance. This probably scuppers chances of naive cancellation.

Without a cone position/velocity/acceleration sensor, the next best way to cancel speaker distortion is perhaps sophisticated model and MPC (model predictive control) techniques. Very compute intensive.
 
Member
Joined 2007
Paid Member
It’s well understood that cone position (displacement) is directly related to the voltage across the speaker terminals throughout the passband, it’s also pretty well understood outside of the passband.

The complication lies in the fact that loudspeaker distortion varies with displacement, Klippel has a lot of great information on this, look up LSI:
http://www.klippel.de/fileadmin/_migrated/content_uploads/Klippel_Nonlinearity_Poster.pdf

The LSI parameters are typically shown for free-air operation, keep in mind they change when you put the loudspeaker in an enclosure. Ideally a loudspeaker in an optimal enclosure will have symmetrical curves. Conversely, you can start with a perfectly symmetrical speaker in free air and get worse symmetry once it’s placed in an enclosure.

Here’s a article/paper on how DC offset relates to distortion:
DC displacement – dynamic offset the voice coil
This is technique typically used to determine if the voicecoil is offset, typically in prototype speakers. You adjust the DC offset to find the minimum distortion and from that you can tell how far your rest position is offset.

-Olen

Alas the distortion of a speaker is primarily related to the cone position (ie due to suspension forces and voicecoil position), not the voltage across the terminals, and the relationship between voltage and cone position varies substantially across resonance. This probably scuppers chances of naive cancellation.

Without a cone position/velocity/acceleration sensor, the next best way to cancel speaker distortion is perhaps sophisticated model and MPC (model predictive control) techniques. Very compute intensive.
 
The simple fact that you can use the superficially perceived "flaw" of a simple single stage, single end amplifier to lower the distortion through cancellation instead of relaying on dumb brute force methods like many cascaded gain stages and copious amounts of negative feedback is truly an ingenious use of synergistic effects.

Thank you Mr Pass!!! Your articles about Triodes, SITs and amplifier design in general have really tough me a lot!!! :)
 
TRiodityIRFP7430.JPG

Amp schematic.

irfp7430-THDfas.jpg

IRFP7430 normal phase.

irfp7430-THDinvtfas.jpg

IRFP7430 inverted phase.

IXYS-THDinvtfas.jpg

IXYS as comparison.

Today I tested my IRFP7430 mosfet in the same source follower configuration as the IXYS IXTP32P05T.

As the IXYS is a normal "pentody" device it will not cancel second harmonic within the amplifier as well. No working the loadline withing the amp - hence it will give a pronounced second harmonic asymmetry that can be used to work the whole loadline including the speaker and thus cancel the asymmetry of the loudspeaker to some degree.

The IRFP7430 has pretty radical "Triody" transfer curves and will cancel the second harmonic within the amp much better.

The difference is staggering. The THD does not change as much with different amounts of current through the device and there is not much difference when changing the phase of the loudspeaker cables to the amp.

This is of course as expected, but it is very fun to see in real life.

At 15,6 volts over the amp and 1,54 ampere of current through the device (3 parallel 22 ohm TO247 resistors) it does sound more powerful then the comparably weak p-channel IXYS device. Not really better, but just more powerful and a lot more playful. It does add "Oohmpfh!" to everything, even if there is non to be had from the record.
 
And yes! At 6,68 kHz we start to see the effects of the tremendous gate capacitance of the IRFP7430 being driven by the quite high output impedance of my MiniDSP2X4HD through a few meters of low grade cheap coaxial cable.

15,4% THD @ 6,68 kHz does not sound very nice while REW is sweeping through the treble when measuring.

It does measure a lot better when driven though a small jfet buffer, but this is only some casual fooling around with what I have on my work bench at the moment.
 
THDirfp7430lowVolt.jpg

The IRFP7430 is a strange little beast. This measurement is with only 5,21 Volts from the powersupply and only 220 mA through the IRFP7430.

The Ixys version clips horrendously with this little Voltage over the amp and current through the device, while the IRFP7430 has a sharp edge to the treble but is quite okey to listen to.

I would not call the IRFP7430 "linear" in any sense of the word, but it does display some strange occasional unforeseen linearity when all other mosfets fail miserably. It is quite fun little beast, but a very colorful acquaintance.

irfp7430-kurvor.JPG
 
You need to keep the junction pretty cool to see that, though

I have Vds = Vgs for a total of 3.8 volts and 1.5 ampere and 5.8 watts of dissipation.

The IRFP7430 seems perfect for Cascode or "Bastode" operation then... High transconductance at low voltages. Nice Triody curves. The Cascode usage lowers the gate capacitance substantially and keeps the dissipation low.

Thanks for the tips of keeping the device cool to keep the Triodity of the transfer curves! :happy1:
 
Last edited:
It’s well understood that cone position (displacement) is directly related to the voltage across the speaker terminals throughout the passband, it’s also pretty well understood outside of the passband.


Speaker distortion is greatest for the bass drivers as the displacement is largest. Bass drivers go through resonance (with ports through two resonances) in their "passband", so the voltage is not related to displacement in any simple way at exactly the frequencies where displacement is large.


Reducing speaker distortion is really the job of the speaker designer, not the amplifier designer, except when positional feedback is used. How could it be otherwise?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.