Tried LT 1357, 25MHz, 600V/us op-amp.
Also works like a champ, with the same rise times 620ns (under load).
TLE 2071 also works fine, 720ns rise times.
AD711 also works fine, 540ns rise times.
TL 071 and NE 5534 and LT 1363 don't work at all.
That's all DIP-8 op-amps I have in stock.
Now I go back to LT1223, and to try to make it work.
Also works like a champ, with the same rise times 620ns (under load).
TLE 2071 also works fine, 720ns rise times.
AD711 also works fine, 540ns rise times.
TL 071 and NE 5534 and LT 1363 don't work at all.
That's all DIP-8 op-amps I have in stock.
Now I go back to LT1223, and to try to make it work.
Last edited:
It turns out, that simply removing C4 (2pF, feedback cap) fixed everything when using LT1223 op-amp.
And that the 2nd thing I tried 🙂
First I tried to add 2pF of capacitance to C4 (because it's easier then removing it), but this
was a move in a wrong direction.
So next I tried to lower the capacitance...
I guess the traces on the PCB provide enough capacitance, and together with 2pF cap it was too much....
I modified my test bench by using vise, to make whole thing more accessible and stable while soldering..
I guess it doesn't look very professional, but it worked.
Sinus and square waves show up perfectly, with or without load.
All screenshots taken with 8 Ohm load.
Rise times are similar like with the other op-amps, so it seems that using LT1223 doesn't have any advantage over using dirt cheap LF356,
at least as far rise times of square waves is concerned.
Even 100kHz square waves with load (or without), look perfect.
I'm going to build 2nd channel, use proper linear PSU, cooling, and do music tests.
I'll try different op-amps again.
And that the 2nd thing I tried 🙂
First I tried to add 2pF of capacitance to C4 (because it's easier then removing it), but this
was a move in a wrong direction.
So next I tried to lower the capacitance...
I guess the traces on the PCB provide enough capacitance, and together with 2pF cap it was too much....
I modified my test bench by using vise, to make whole thing more accessible and stable while soldering..
I guess it doesn't look very professional, but it worked.
Sinus and square waves show up perfectly, with or without load.
All screenshots taken with 8 Ohm load.
Rise times are similar like with the other op-amps, so it seems that using LT1223 doesn't have any advantage over using dirt cheap LF356,
at least as far rise times of square waves is concerned.
Even 100kHz square waves with load (or without), look perfect.
I'm going to build 2nd channel, use proper linear PSU, cooling, and do music tests.
I'll try different op-amps again.
Last edited:
2e-12 farad made the difference? Yikes! That's in the range of the opamp's Cin.
But the effective value of Cin is a bit nebulous in terms of the Andersen topology, since the opamp's supply rails are provided by active devices. They have their own frequency-dependent characteristics as well. I'm basically saying that the effective Cin could depend on the opamp's supply rail's AC response. One interesting thing to try would be to add some capacitance between the opamp's Vcc and Vee pins, to prevent some types of shenanigans. Since current variations in the opamp's Vcc and Vee supplies will track, there shouldn't be any bandwidth penalty in doing that.
But the effective value of Cin is a bit nebulous in terms of the Andersen topology, since the opamp's supply rails are provided by active devices. They have their own frequency-dependent characteristics as well. I'm basically saying that the effective Cin could depend on the opamp's supply rail's AC response. One interesting thing to try would be to add some capacitance between the opamp's Vcc and Vee pins, to prevent some types of shenanigans. Since current variations in the opamp's Vcc and Vee supplies will track, there shouldn't be any bandwidth penalty in doing that.
Have a 2nd channel up and running. Everything the same as the 1st channel.
As previously reported, C4 should be omitted when using LT1223 op-amp,
BUT for the other op-amps I tried yesterday, C4 (2pF) should NOT be skipped.
The following op-amps have been confirmed to work correctly (besides LT1223):
LT1357, TLE2071, TLE2081, AD711, OPA134, LF356.
These other op-amps were not simulated in LTSpice, just confirmed that they seem to work OK in the real amp.
The schematic and sim may not be optimal for these op-amps, so there might be some space for improvement
if using non-LT1223 op-amp.
As previously reported, C4 should be omitted when using LT1223 op-amp,
BUT for the other op-amps I tried yesterday, C4 (2pF) should NOT be skipped.
The following op-amps have been confirmed to work correctly (besides LT1223):
LT1357, TLE2071, TLE2081, AD711, OPA134, LF356.
These other op-amps were not simulated in LTSpice, just confirmed that they seem to work OK in the real amp.
The schematic and sim may not be optimal for these op-amps, so there might be some space for improvement
if using non-LT1223 op-amp.
I feel really silly, but I discovered that my cheap square wave generator produces waves with 500-700ns rise time.
That's why all these results for different op-amps were similar.
I decided to pull out from the closet my Siglent SDG810 generator, which I've never really used before,
as it's more bulky and too complex to configure and control.
Don't like all these buttons, would prefer two simple potentiometers to control amplitude and frequency.
But I guess now it's time to learn...
Anyway, as we can see from the screenshots, we are getting 120ns rise time, which I believe corresponds to the actual speed of the amp, as
the generator alone produces waves with rise time around 60ns.
I didn't repeat the tests with other op-amps, only with LT1223.
I burnt Zobel resistor during these experiments, and had to replace it with a new one...
That's why all these results for different op-amps were similar.
I decided to pull out from the closet my Siglent SDG810 generator, which I've never really used before,
as it's more bulky and too complex to configure and control.
Don't like all these buttons, would prefer two simple potentiometers to control amplitude and frequency.
But I guess now it's time to learn...
Anyway, as we can see from the screenshots, we are getting 120ns rise time, which I believe corresponds to the actual speed of the amp, as
the generator alone produces waves with rise time around 60ns.
I didn't repeat the tests with other op-amps, only with LT1223.
I burnt Zobel resistor during these experiments, and had to replace it with a new one...
Last edited:
Final amp assembled, and playing music. As usual, it's a 'naked' chassis powered by separated PSU (box below with VU meters).
USB powered very quiet fan blowing gently from below the heatsink.
Amp behaves perfectly and sound is great.
USB powered very quiet fan blowing gently from below the heatsink.
Amp behaves perfectly and sound is great.
Hovercraft Amp!!
Love it Minek, fantastic job!
Please give an update after 'latest build is greatest' syndrome wears off and let us know if she's a keeper 😉.
Love it Minek, fantastic job!
Please give an update after 'latest build is greatest' syndrome wears off and let us know if she's a keeper 😉.
Judging from the sound of Vivaldi's "Winter" playing right now - yes!
Hi Minek,
Nice work indeed.
Did you get a higher measured max SR with the LT1223 and faster input? Or did I miss it.
Nice work indeed.
Did you get a higher measured max SR with the LT1223 and faster input? Or did I miss it.
Yes, see post #46.
120 ns rise time, and very stable.
LT1357 also gave very good results, just slight;y slower (and cheaper).
I tried today HexFet version of this amp - schematic posted in post #1.
New PCB arrived very quickly this time, and there was no extra tariffs/duties, so it's still very reasonable priced.
Basically, HexFet version of the amp performs pretty much the same as LatFet version. Very stable, rise time on square
waves in the the same range (160ns - fastest measured).
The only difference is that C4 (2pF) had to be present. If it's omitted, there was a little of ringing present on square waves.
With C4 included, everything worked fine.
From simulations looks like it will work fine with IRFP240/IRFP9240 outputs as well.
Q8 was connected to the PCB with wires, and mounted on the heatsink close the output devices (temperature sensor).
New PCB arrived very quickly this time, and there was no extra tariffs/duties, so it's still very reasonable priced.
Basically, HexFet version of the amp performs pretty much the same as LatFet version. Very stable, rise time on square
waves in the the same range (160ns - fastest measured).
The only difference is that C4 (2pF) had to be present. If it's omitted, there was a little of ringing present on square waves.
With C4 included, everything worked fine.
From simulations looks like it will work fine with IRFP240/IRFP9240 outputs as well.
Q8 was connected to the PCB with wires, and mounted on the heatsink close the output devices (temperature sensor).
Last edited:
There is a side story here - for this amp I tried to use MosFets purchased from Utsource.
I bought from the previously (not mosfets though), and I considered them to be a reliable source.
But this time, either FQA28N15 or FQA36P15 (or both) turned out to be fake.
The amp simply refused to work with them, oscillating like a madman.
I wasted few hours trying to make it work, but when tried to use 'real' Mosfets purchased from Mouser
few years ago, everything worked perfectly.
So basically Utsource can't be trusted.
I bought from the previously (not mosfets though), and I considered them to be a reliable source.
But this time, either FQA28N15 or FQA36P15 (or both) turned out to be fake.
The amp simply refused to work with them, oscillating like a madman.
I wasted few hours trying to make it work, but when tried to use 'real' Mosfets purchased from Mouser
few years ago, everything worked perfectly.
So basically Utsource can't be trusted.