Winfield's 100W DC-10MHz 1000V/us amplifier

Rather than continue and extend a fragmented discussion on other threads, I'm starting a new one to discuss my amplifier design, the AMP-70. Fast slew rates are desirable for audiophile amplifiers, for better transient response. My 1000V/us amplifier, which can deliver 100W rms into 8 ohms, could be thought of as an awesome hi-fi amp, but it's really just a laboratory amplifier designed to deliver a full output swing up to 5MHz, even into capacitive loads. (I've included some reduced-size images in this post, the full versions are in the Dropbox folder linked below.)

An externally hosted image should be here but it was not working when we last tested it.


The design is based on the output amplifier in the Tektronix PG508 50MHz pulse generator. We examine this in detail as a Designs-by-the-Masters section in our upcoming book, The Art of Electronics, x-Chapters. I've included this 4-page section in the Dropbox link below. Here's the schematic.

An externally hosted image should be here but it was not working when we last tested it.


Basically there's an input stage creating a push-pull error current into a folded-cascode VAS stage, and emitter-follower outputs. The secret for fast slewing is to use a high 60mA VAS current with low Ccb capacitance transistors, for a fast slew rate S = i/C = 3000V/us for the PG508. The Tek pulse generator only had to deliver 0.2A to its load, whereas I have to deliver 5A, so I'm using larger, higher-capacitance transistors. Following the principle, F.O.M. = Pd / Ccb, to improve the slew rate, I selected low-capacitance types. They're fragile and I'm using 20, to limit their power dissipation.

An externally hosted image should be here but it was not working when we last tested it.


My amplifier design is somewhat more complicated, with various features to increase its capabilities, which we can discuss. Here's a dropbox link to my AMP-70A-2 schematic and related files for you to look over. https://www.dropbox.com/sh/dno89om1uezxs8a/AACoJsLyNazSQZvE9_cTcH4Ja?dl=0
 
Last edited:
And there was me thinking the 300V/us Stochino design I currently use was excessive! Although the BNCs were a bit of a giveaway! I didn't know you could get those rated for 5A.

I suppose the issue of class-AB biasing of the output MOSFETs has been discussed elsewhere in depth. Ordinary power MOSFETs are far worse than the lateral FETs we love. Looking at Fairchild's IRF640 and IRF9640 datasheets, they conveniently show Vgs to Id tempco curves down to 100mA. Assuming you bias at about 100mA (20 watts quiescent dissipation for two) the total Vgs is 7.8V at 25C and 6.4V at 125C. The 0R33 resistors only drop 0.07V, and don't look very helpful. That's -14mV/deg in this current region, and the two EF trannies increase that to about -18mV/C. The bias-setup network has four BJTs, or about -9mV/C, and this assumes it can track the MOSFET junction temps.

If we assume a +50C junction rise, the 9mV/C mismatch creates an unexpected -0.45V Vgs change. Going back to the curves, we see this can increase the quiescent current dramatically, perhaps to over 500mA. This would mean that playing extended loud passages could dramatically increase the quiescent dissipation over the carefully-set room-temp value. So I've always been nervous about using Ordinary power MOSFETs in this fashion.
 
I suppose the issue of class-AB biasing of the output MOSFETs has been discussed elsewhere in depth. Ordinary power MOSFETs are far worse than the lateral FETs we love. Looking at Fairchild's IRF640 and IRF9640 datasheets, they conveniently show Vgs to Id tempco curves down to 100mA. Assuming you bias at about 100mA (20 watts quiescent dissipation for two) the total Vgs is 7.8V at 25C and 6.4V at 125C.

Not wishing to stray too far from topic. I have the IR datasheets for those parts and they didn't go down that far so ty for pointing me to the fairchild one. Confession. I got two complete amps as freebies. The first (using the 640/9640) caught fire. the second, using IRFP240/9240 hasn't yet, but its horrendously underbiased and the protection trip is over sensitive. But free and works whilst I complete the amplifier I am building. Then a rebuild of both in a case where I can actually get at things. Given current rate I'll report back in 2020!

</offtopic>
 
Rather than continue and extend a fragmented discussion on other threads, I'm starting a new one to discuss my amplifier design, the AMP-70. Fast slew rates are desirable for audiophile amplifiers, for better transient response. My 1000V/us amplifier, which can deliver 100W rms into 8 ohms, could be thought of as an awesome hi-fi amp, but it's really just a laboratory amplifier designed to deliver a full output swing up to 5MHz, even into capacitive loads. [snip]

Following the principle, F.O.M. = Pd / Ccb, to improve the slew rate, I selected low-capacitance (power-transistor) types. They're fragile and I'm using 20, to limit their power dissipation.

Here's something else to consider. In the quest for suitable power transistors, especially ones with a high slew-rate FOM, I've come up with parts that you may not find familiar. In this power-BJT candidate spreadsheet, you may recognize a few parts, but they may have rather poor FOM ratings. As usual there's the issue of discontinued parts and availability, but many are stocked by mainline distributors. Others are available at Rochester, etc., at low per-item prices, but their line-item- and order-minimums lead to rather large lifetime buys! BTW, if anybody wants to experiment with my lifetime-buy transistors, just let me know.
 
Last edited:
Not wishing to stray too far from topic. I have the IR datasheets for those parts and they didn't go down that far so ty for pointing me to the fairchild one. Confession. I got two complete amps as freebies. The first (using the 640/9640) caught fire. the second, using IRFP240/9240 hasn't yet, but its horrendously underbiased and the protection trip is over sensitive.

Ouch! Here's a link to the Fairchild datasheets. The IRFP240/9240 appear to have a larger die, and hence a higher Pd rating. But their Vgs vs Id curves are similar. If you bias the amplifier for reasonable currents under high load, it may be badly under-biased when limping along with no load.
 
Rather than continue and extend a fragmented discussion on other threads, I'm starting a new one to discuss my amplifier design, the AMP-70. Fast slew rates are desirable for audiophile amplifiers, for better transient response. My 1000V/us amplifier, which can deliver 100W rms into 8 ohms, could be thought of as an awesome hi-fi amp, but it's really just a laboratory amplifier designed to deliver a full output swing up to 5MHz, even into capacitive loads. (I've included some reduced-size images in this post, the full versions are in the Dropbox folder linked below.)

An externally hosted image should be here but it was not working when we last tested it.


The design is based on the output amplifier in the Tektronix PG508 50MHz pulse generator. We examine this in detail as a Designs-by-the-Masters section in our upcoming book, The Art of Electronics, x-Chapters. I've included this 4-page section in the Dropbox link below. Here's the schematic.

An externally hosted image should be here but it was not working when we last tested it.


Basically there's an input stage creating a push-pull error current into a folded-cascode VAS stage, and emitter-follower outputs. The secret for fast slewing is to use a high 60mA VAS current with low Ccb capacitance transistors, for a fast slew rate S = i/C = 3000V/us for the PG508. The Tek pulse generator only had to deliver 0.2A to its load, whereas I have to deliver 5A, so I'm using larger, higher-capacitance transistors. Following the principle, F.O.M. = Pd / Ccb, to improve the slew rate, I selected low-capacitance types. They're fragile and I'm using 20, to limit their power dissipation.

An externally hosted image should be here but it was not working when we last tested it.


My amplifier design is somewhat more complicated, with various features to increase its capabilities, which we can discuss. Here's a dropbox link to my AMP-70A-2 schematic and related files for you to look over. https://www.dropbox.com/sh/dno89om1uezxs8a/AACoJsLyNazSQZvE9_cTcH4Ja?dl=0
interisting.
 
hi,winhill2,could you send the image to my email?

Rather than continue and extend a fragmented discussion on other threads, I'm starting a new one to discuss my amplifier design, the AMP-70. Fast slew rates are desirable for audiophile amplifiers, for better transient response. My 1000V/us amplifier, which can deliver 100W rms into 8 ohms, could be thought of as an awesome hi-fi amp, but it's really just a laboratory amplifier designed to deliver a full output swing up to 5MHz, even into capacitive loads. (I've included some reduced-size images in this post, the full versions are in the Dropbox folder linked below.)

An externally hosted image should be here but it was not working when we last tested it.


The design is based on the output amplifier in the Tektronix PG508 50MHz pulse generator. We examine this in detail as a Designs-by-the-Masters section in our upcoming book, The Art of Electronics, x-Chapters. I've included this 4-page section in the Dropbox link below. Here's the schematic.

An externally hosted image should be here but it was not working when we last tested it.


Basically there's an input stage creating a push-pull error current into a folded-cascode VAS stage, and emitter-follower outputs. The secret for fast slewing is to use a high 60mA VAS current with low Ccb capacitance transistors, for a fast slew rate S = i/C = 3000V/us for the PG508. The Tek pulse generator only had to deliver 0.2A to its load, whereas I have to deliver 5A, so I'm using larger, higher-capacitance transistors. Following the principle, F.O.M. = Pd / Ccb, to improve the slew rate, I selected low-capacitance types. They're fragile and I'm using 20, to limit their power dissipation.

An externally hosted image should be here but it was not working when we last tested it.


My amplifier design is somewhat more complicated, with various features to increase its capabilities, which we can discuss. Here's a dropbox link to my AMP-70A-2 schematic and related files for you to look over. https://www.dropbox.com/sh/dno89om1uezxs8a/AACoJsLyNazSQZvE9_cTcH4Ja?dl=0


I,m glad to read the very interisting design. It's excellent. But because I'm from China, the Dropbox has been blocked by GFW. So I need your help.Could you send these image and material about your amp-70a and amp-70a-2 to my email? my email adress is 892461095@qq.com, thks!!!
 
Thanks for the comments. As you can see from the schematic, the amplifier can actually slew to 1330V/us. That gives us a full output swing to 5MHz. It's response begins to drop off at 10MHz. It can be made to go faster at the expense of higher power dissipation in Q15, 16, 29, 30, 35 and 36.

On the drawing I say, don't use the 1-ohm resistance output connection. But of course that's what we are using. 🙂 Furthermore, we have two amplifiers each driving a 1:4 step-up transformer, to get two 240 Vpp signals, 90-degrees out of phase, over the range of 1 to 5MHz, to create a smoothly-rotating electric field across four electrodes, for electro-rotation experiments.

I have added the Gerber file to the DropBox link, and if someone is really, really interested, I could provide a blank PCB.
 
Last edited:
I have added a BOM and the Gerber file. If you are really, really interested in experimenting with this, despite it's painful collection of 20 output transistors on a pair of nested machined heat-spreading plates**, etc., I could send a blank PCB.

** I searched just now but couldn't locate my drawings with the detailed plate-machining dimensions.
 
Sorry for the confusion. The board was only designed for five complementary output pairs. Yes, I said it takes 20 transistors on the heat sinks, because the VAS, driver and thermal-biasing transistors also need to be on the heatsink. So you see 10 pairs of power transistors and screw holes.
 
"My amplifier design is somewhat more complicated, with various features to increase its capabilities, which we can discuss."

One simple thing I've done that I haven't seen others do, is to place a 5-volt dc-dc converter supply on top of each of the ±48V power rails. These are used to power the VAS stage. By driving above the rails, we insure the full capability of the final output-transistor emitter-followers is realized. You can use 1- or 2-watt six-pin SIP dc-dc modules, which are small and cost about $5.

I used 2-watt modules that can deliver 400mA, because I'm running the VAS stage's quiescent bias current at 80mA, with 160mA peak current. If the DC amplifier is driven off-scale beyond the rails, the VAS will sit at 160mA. A 200mA max capability would be too close.