They’re supposed to create a “loose” connection between signal ground, power ground and chassis earth to prevent hum. There are as many opinions on how to do this as there are audio nerds, but this is one way that I’ve seen used.
Honestly, it has no audible hum even if I short the resistors and use a single ground.
OK. I added some drivers. As you cautioned me, it didn't make much of a difference in the simulation. THD@1kHz/15W went down a smidge, but THD@10kHz/15W was cut in half. I doubt it's audible. I also tried some square waves and I couldn't really see a difference. Same thing with the frequency response. The corner frequency is still at about 100kHz. Again, you warned me that I wouldn't see much change in the simulation.
I guess next step is a listening test. I can probably patch it in fairly easily on an existing board. I'm excited to see if I can hear a difference. As I said, I love everything about the sound, except strings that sound a little metallic. Let's see if this fixes that problem. I'll see if I can find some better speakers than the Daytons that it's hooked up to now.
Here's what I have in SPICE right now. I wouldn't call it fully dialed in. It's pretty finicky on the VBE settings and the emitter resistor on the driver. I took the 680ohms from the design you shared a few posts back. The bias current sits at 58mA right now. Increasing it to 100mA makes a fairly small change to THD.
My humble project is evolving with you guys' help. Just as I was hoping! 👍
Last edited:
Looks good.
Here is a link to an interesting discussion about source restistors in single-pair hexfet OS.
Lots of reading, but very interesting stuff from reputable participants.
I can add that since that discussion in 2020 I'm not using source resistors in such amps, and they all work and sound fine.
Here is a link to an interesting discussion about source restistors in single-pair hexfet OS.
Lots of reading, but very interesting stuff from reputable participants.
I can add that since that discussion in 2020 I'm not using source resistors in such amps, and they all work and sound fine.
I'll gladly admit I have no idea what he's talking about in that math section. It looks like he's calculating the gate current needed to charge the gate capacitor(s). I think I=Q/t, which translates to Imax=Q*f for a sine wave. So far so good. But where the heck does he get the factor 100 from? He says he needs that to reproduce a 10kHz square wave. Then he somehow plugs in 50kHz and multiplies THAT by 100. So he basically says he needs headroom for 50kHz and 100 harmonics on top of it. That's a 5MHz bandwidth at full power! And those harmonics are definitely not full power. The number that pops out, 350mA per device seems really high! 15V/us gives a decent 10kHz square wave. At 50W of power you need to swing about 20V in each direction. That's 1.3us from 0-20V. Plugging that into the formula for current gives us I=70e-9/1.3e-6=54mA. SPICE actually shows a much lower number. I'm not entirely sure why.
Maybe someone can explain the math. because it makes no sense to me.
This is what 10kHz @ 15V/us looks like. Since this amplifier is supposed to be "good enough", I'd give this a passing grade.
As written earlier, the 33 µF input cap will be charged via the device that will be plugged in and that one powered on later resulting in a heavy plop (varies on the type of muting). When the amplifier is powered on and a device is connected its power on phenomenon will be spectacular if it does not have muting. Then both the DC offset voltage and the 33 µF cap charging will be consumed by your woofers for a non healthy amount of time. It is just 2 GND reference resistors that will make the effect way less. Since the amplifier does not have a delayed speaker relay/DC protection this effect will definitely show up in a few forms in various scenarios. Of course a lower value cap will also help in that aspect. Floating capacitor coupled inputs are a design no no.
Never have simulated anything but I used to build amplifiers in the past and this is what happens in real life. Stuff gets switched on and off.
Never have simulated anything but I used to build amplifiers in the past and this is what happens in real life. Stuff gets switched on and off.
Last edited:
I like the idea behind this thread. It just struck me how ironic that the more improvements are made, the less "good enough" the amp becomes.
I think it's part of the reason these projects sometimes don't get off the ground - nobody can agree on where to stop.
I looked at that thread a while back and it sort of lead to my current feelings on the state of things.
HoneyBadger - This is the amp I wanted to build originally, but the more I looked into it the more I found it to be quite complicated to get started on. While I understand you yourself did quite a bit of work updating the BOM, that was 8 years ago, and the build guide is 5 years older than that (and 0.4 board revisions behind). I think the project needs more attention to be viable as a 'straight forward' build.
Leach Amp - Looks like a good candidate - the work this member did was considerable. They haven't been active in 9 years, however.
SlewMaster - Much like the HoneyBadger, minus the boards in the store. A confusing array of versions, too.
AMB Lab β24 - AMB documentation is generally great and they seem to be very no-nonsense about design, but it's hard to know how hard the build is with this one as the documentation is available to board purchasers only (which is fair, I just can't say if it fits the bill).
Rod Elliot P3A - Well regarded and straight forward, but last I checked it relies on obsolete devices (themselves replacements for other obsolete devices).
AKSA 55 - I see Hugh's stuff mentioned a lot in DIY circles, but never seem to figure out where the DIY stuff is. I expect to be able to find a store page or a gerber or a group buy, but I come away empty-handed. Maybe I'm just not looking in the right places or I'm misunderstanding what the deal is.
Circlophone© by Elvee - I'm confused as to whether this is actually Class AB or not. It looks cool.
And so on and so forth. They all seem to be cool designs, but most seem to rely on the builder going off a circuit diagram and a general understanding of what they're doing before they begin.
Surprisingly, this amplifier has no pop on power up whatsoever. It does, however, have a bit of a “fart” when turned off and the filter caps are discharging.
But I agree that 33uF is too much. The reason for that value was simply that I had a lot of high quality 33uF non polarized caps lying around when I built it. It will be changed to 4.7uF.
Then connect a source to the freshly powered on power amplifier and switch the source on/off.
Replace an interlink with the power amplifier being energized.
Replace an interlink with the power amplifier being energized.
Last edited:
He meant apply an external transient (like what happens with Idiot Users swapping things hot), observe the result, and see if you still want an oversized input coupling cap.
If you are following all the “rules” for proper cap sizes and pole staggering the input coupling cap should be the one with the shortest time constant / dominant pole.
If you are following all the “rules” for proper cap sizes and pole staggering the input coupling cap should be the one with the shortest time constant / dominant pole.
Yup. There’s a significant pop when I reboot the Volumio player that’s connected to it that I think might be caused by this. As I said, I realize a cap that large isn’t ideal. I just happened to have a bag of good Panasonic non-polarized 33uF caps that I used. I will update the schematic with a smaller cap. I’m thinking 4.7uF.
I agree 100%. If one want's to improve on this issue, I see two things that need to happen with a new design:
- Provide full, public documentation.
- Keep this documentation up to date for a long time.
The more I read this article, the more confused I get. This for example:
| Fix: | Drive the gates with as much current as possible. This may include adding a class AB driver stage. | |
| Why: | HEXFETs have a nonlinear transfer curve up to about an amp or two, depending on the device(s) used. In a class AB amplifier, this characteristic is the cause for a majority of the THD. When driven with enough current, the device will follow the 'new' linear curve, since it is balancing out the nonlinear gate capacitance. The lower impedance of the driver stage the better. |
This makes no sense at all to me. When he's talking about the transfer curve, he must be talking about the Vgs-to-Ids curve. The "amp or two" is the Igs, not the gate current. To put the transistor into the more linear region, you need more quiescent Ids, not more gate current. A gate driver won't do anything to mitigate this.
In an audio application, can't force current into the gate. It draws what it draws and it's a function of the dV/dt of the signal. Thus, the ONLY thing a gate driver will do is to help with slew rate/transient response. A MOSFET has a very high input impedance, so once it's settled, it draws almost no current. The only time is draws significant current is when changing the charge of the gate capacitance. It's easy to show that max(Igs)=Q*f. So to faithfully reproduce 100kHz at full power (which never happens), you'd need 70e-9*100000=7mA of gate current. Again, I have no idea where the factor 100 comes from in his math.
I'm sure you are all correct that gate drivers are needed, but the reasoning in Rod's article doesn't make sense to me and. Or is it just my inexperience that shows its ugly head? It wouldn't be the first time!
The gate capacitance is highly nonlinear from zero to several amps of current, then it becomes more linear. Drive it from a low impedance source and it’s nonlinearity contributes less to distortion.
Hexfets are far worse about this compared to laterals. And the modeling of that gate charge characteristic is simplistic at best, so spice models rarely correspond to reality here. In switching applications where the only thing that matters is total gate charge, and how it accumulates isn’t important, the simple model works well enough.
Hexfets are far worse about this compared to laterals. And the modeling of that gate charge characteristic is simplistic at best, so spice models rarely correspond to reality here. In switching applications where the only thing that matters is total gate charge, and how it accumulates isn’t important, the simple model works well enough.
Several amps of what current? Gate current? No matter what I drive it from, it’s limited by dV/dt of the signal, isn’t it?
I'm not a fan of using body diodes for much of anything either. Some MOS devices now have detailed specs for the body diode and those devices could be an exception. But I don't think the IRFP240/9240 are among them.
I'm curious about the decision to use PNPs for the VAS, though. I would have thought that you'd get better performance with NPNs. After all, electron mobility in silicon is ~4x the hole mobility. I know you're not aiming for the ultimate performance, but why not get the best out of the topology?
I'm not a super fan of running the bias current in the input stage through a trimpot. How bad is the DC offset? Does it need correcting?
I'd make sure to take a close look at the slewing behaviour of the amp, especially with 330 Ω gate stoppers on the output.
I don't think you need the reverse biased diodes on the rails. Are you trying to protect against reverse polarity during the assembly process?
You need to use a minimum of one (1) red LED in the CCSes. Otherwise nobody will believe the amp is working. 😉
Nice work, though. I like the down-to-basics approach.
Tom