Le of tweeter at 650k?

[sendler]

What is the approximate Le of an average mid woofer at 20k? 650k? For an average dome tweeter? Ribbon? I've been playing with some Class D amp modules from Sure Electronics.
http://www.sureelectronics.net/goods.php?id=918
They sound really excellent with some mods to the undersized components. They run on a single 20-32V supply and generate very little waste heat considering the fine sound. Some of the best amps I've heard after mods. The main roadblock to overcome in any Class D amp is the required output filter that is needed to remove the 650k carrier wave from the output. I'm interested in knowing if the Le of the driver will be significant in modeling the response of the output filter.
http://www.diyaudio.com/forums/showthread.php?p=1942646

[Pete Schumacher ®]

The inductance will definitely be high enough to suppress any current being built up, but you can't really rely on the driver inductance for filtering.

Set your filter knee with an inductor of around 15uH and a cap around .47uF. That puts the filter knee a decade below the switching frequency, putting the switch frequency 40dB down, minimum, above 20KHz.

[neildavis]

The inductance will definitely be high enough to suppress any current being built up, but you can't really rely on the driver inductance for filtering.

Actually, you can use the driver inductance for some digital amps, but not this particular one. Look up SLOA119A in Google: TI has an application note on the difference between class AD ("traditional") and class BD ("filterless") modulation. The TP2050 output stage uses AD modulation.

The Tripath amps that are being modded do not survive large currents during output inductor saturation, so Tripath recommends the use of distributed gap cores (powered iron/carbonyl) to avoid saturation. It would probably be hard to "saturate" the driver but it's something to consider...

[Pete Schumacher ®]

Actually, you can use the driver inductance for some digital amps, but not this particular one. Look up SLOA119A in Google: TI has an application note on the difference between class AD ("traditional") and class BD ("filterless") modulation. The TP2050 output stage uses AD modulation.

The Tripath amps that are being modded do not survive large currents during output inductor saturation, so Tripath recommends the use of distributed gap cores (powered iron/carbonyl) to avoid saturation. It would probably be hard to "saturate" the driver but it's something to consider...

I always used some kind of ferrite toroid for my power inductors in switch mode applications.

I'm looking at the PDF for BD modulation, but it still looks to me that there is a filter involved, even for "filterless" modulation. I'm not sure what I'm missing there. Also, AD sure looks superior in almost every way from a non linear distortion standpoint.

[neildavis]

I'm looking at the PDF for BD modulation, but it still looks to me that there is a filter involved, even for "filterless" modulation. I'm not sure what I'm missing there. Also, AD sure looks superior in almost every way from a non linear distortion standpoint.

Yeah, you're right in that the app note shows the inductor...but usually the class D filterless designs just use a ferrite bead to control EMI and the voice coil inductance is used as the output inductor. TI and Maxim have medium powered class BD amps, and there are a lot of low-powered class D amps from National, Linear, Maxim, TI...all using the "filterless" modulation approach.

You are right about the distortion. That big peak in the distotion vs frequency plot around 6K seems to be there on all of these smaller amps. I keep looking around for new high quality class D amps, but all of these new chips seem to be targeting markets where the cost and size of a coil is a concern.

I always used some kind of ferrite toroid for my power inductors in switch mode applications.

Ferrite can be OK for audio amps, but I don't understand why vendors sell "low distortion" ferrite inductors for audio applications. TI and Apogee use the Toko 13RHBP series (http://www.toko.co.jp/products/en/in...dex_amp_e.html). Sagami and Coilcraft (http://www.coilcraft.com/ga3416.cfm) also make "low distortion" ferrite coils. What makes these ferrite coils "low distortion"?

I tried several powdered iron cores but had way too much core loss...kinda disappointing building an amp where the hottest device on the board is the inductor. MPP and Sendust have comparable saturation properties to powdered iron, and the core losses are 10 times lower. Unfortunately, you have to wind your own, since it's hard to find value low-loss inductors as a standard part. I can't find a low-volume source for the Toko or Sagami products, so I've been winding my own Sendust cores until I can afford to get some made.

[Pete Schumacher ®]

Yeah, you're right in that the app note shows the inductor...but usually the class D filterless designs just use a ferrite bead to control EMI and the voice coil inductance is used as the output inductor. TI and Maxim have medium powered class BD amps, and there are a lot of low-powered class D amps from National, Linear, Maxim, TI...all using the "filterless" modulation approach.

You are right about the distortion. That big peak in the distotion vs frequency plot around 6K seems to be there on all of these smaller amps. I keep looking around for new high quality class D amps, but all of these new chips seem to be targeting markets where the cost and size of a coil is a concern.



Ferrite can be OK for audio amps, but I don't understand why vendors sell "low distortion" ferrite inductors for audio applications. TI and Apogee use the Toko 13RHBP series (http://www.toko.co.jp/products/en/in...dex_amp_e.html). Sagami and Coilcraft (http://www.coilcraft.com/ga3416.cfm) also make "low distortion" ferrite coils. What makes these ferrite coils "low distortion"?

I tried several powdered iron cores but had way too much core loss...kinda disappointing building an amp where the hottest device on the board is the inductor. MPP and Sendust have comparable saturation properties to powdered iron, and the core losses are 10 times lower. Unfortunately, you have to wind your own, since it's hard to find value low-loss inductors as a standard part. I can't find a low-volume source for the Toko or Sagami products, so I've been winding my own Sendust cores until I can afford to get some made.

Other than EMI containment, why not just use air core? The inductance value is small enough that even 18ga would likely be adequate. Of course, if size is a concern, then some type of ferrite core is required.

[neildavis]

Other than EMI containment, why not just use air core? The inductance value is small enough that even 18ga would likely be adequate. Of course, if size is a concern, then some type of ferrite core is required.

You got it--It's both size and EMI containment. With that much current, EMI can be a big problem. I'm using Sendust cores that only cost about 80 cents, and they require 19 turns. They are the bigger ones in the picture, near the mounting plate. Not too hard to do, really, but I'm going to get a quote for a small quantity.

This amp uses the same "output technology" as the 100W amp referenced in the original post. ST was the Fab house for both Tripath and Apogee, and ST got rights to use the output design from Tripath in their licensing arrangement. The amp under the heatsink is the STA328, which was originally designed by Apogee and now sourced by ST. http://www.st.com/stonline/products/...803/sta328.pdf

[sendler]

Here's one of the filters I have modeled for the Sure 2X100 amp which is bridged. I actually like the sound of the one that I built using 4uH air cores much better that the filter with 13uH toroids wound on MicroMetals T106-2 cores. The air cores sound great but pollute AM reception up to 20 feet away so shouldn't be used in an on going basis. I will wind some smaller toroids to 6uH to see if that will get me closer to the sound I am getting with the forbidden air cores.
.
.
.


.
.
.

Here's the response into 8ohms.
.
.
.


.
.
.

It has a slight peak into 8 ohms but I like to keep the coils as small as possible and it is pretty flat into 4 ohms. Now look at the simulated response when .2mH is added in series to the 8ohm load. This would be a low amount of inductance even for a tweeter at 2k, let alone a midwoofer at 20k but is a fair demonstration just to get started.
.
.
.


.
.
.

Down 12 volts/6db at 30k. Shouldn't we correct our filters to account for driver Le?

[Pete Schumacher ®]

Here's one of the filters I have modeled for the Sure 2X100 amp which is bridged. I actually like the sound of the one that I built using 4uH air cores much better that the filter with 13uH toroids wound on MicroMetals T106-2 cores. The air cores sound great but pollute AM reception up to 20 feet away so shouldn't be used in an on going basis. I will wind some smaller toroids to 6uH to see if that will get me closer to the sound I am getting with the forbidden air cores.
.
.
.


.
.
.

Here's the response into 8ohms.
.
.
.


.
.
.

It has a slight peak into 8 ohms but I like to keep the coils as small as possible and it is pretty flat into 4 ohms. Now look at the simulated response when .2mH is added in series to the 8ohm load. This would be a low amount of inductance even for a tweeter at 2k, let alone a midwoofer at 20k but is a fair demonstration just to get started.
.
.
.


.
.
.

Down 12 volts/6db at 30k. Shouldn't we correct our filters to account for driver Le?

All that really matters is the final impedance of the system, including the XO.

So if you feel the tweeter inductance must be accounted for, there's nothing wrong with adjust the "zobel" at the output of the amp to provide for a minimum impedance at a desired frequency. In other circles, they're called "snubbers."

[Pete Schumacher ®]

You got it--It's both size and EMI containment. With that much current, EMI can be a big problem. I'm using Sendust cores that only cost about 80 cents, and they require 19 turns. They are the bigger ones in the picture, near the mounting plate. Not too hard to do, really, but I'm going to get a quote for a small quantity.

This amp uses the same "output technology" as the 100W amp referenced in the original post. ST was the Fab house for both Tripath and Apogee, and ST got rights to use the output design from Tripath in their licensing arrangement. The amp under the heatsink is the STA328, which was originally designed by Apogee and now sourced by ST. http://www.st.com/stonline/products/...803/sta328.pdf

On another note, have you ever thought about doing the PWM for the class D output directly from the digital signal? It would seem fairly simple to translate the digital word into an appropriate pulse width, but then, I'm no machine language programmer . . .

But having the pulse width controlled digitally, in a 1-1 relationship to the 16bit word, would theoretically enable very low distortion drive signals instead of relying on a sawtooth comparator.

[neildavis]

On another note, have you ever thought about doing the PWM for the class D output directly from the digital signal? It would seem fairly simple to translate the digital word into an appropriate pulse width, but then, I'm no machine language programmer . . .

That's how these ST amps work. Digital input, followed by a general-purpose DSP block, followed by the PWM logic. No analog at all and no feedback. The larger Tripath amps use an analog controller (TK2001); these use the same output devices but have a digital PWM controller.

The DSP block is potent enough to do active crossovers and EQ. These ST amps are similar to the TI "Purepath" series, which are also nice. Brian Powers is playing with one of these amps, but I'm still working the control code for the on-board micro to make it more useful. Getting there, though. Also, he's helping me rethink the Windows app that lets you build the crossovers.

The only real down side to the pure digital approach is power supply rejection. The electrical model for this type of amp is a DAC, and variations in the supply will appear on the output. So it's best to use well regulated supplies, preferrably with a sense line. But that's not too hard.

[sendler]

I see what I did wrong with the model. I inserted the tweeter inductance on the wrong side of my measuring point. The new sim doesn't have the sagging response but does show that the zobel is very important to keep the filter from peaking.