Topping flexing their muscles - this is completely gratuitous and completely unnecessary, and I completely love it. I must have one .... one day....
It makes no rational sense with the L70 and A70Pro already in the range, and with noise figures that are just as good, but the extra headroom seals it.
Topping must have felt the need to make a preamp that could drive the B100 at low gain (into 4 ohms at least).
It would also be able to drive all those low-gain, low-input-impedance Hypex and Purifi amps directly, without a buffer.
The A900 generates 21.2 Vrms at 600Ω and holds it all the way down to 32Ω, and only drops 11% to 18.8V into 16Ω, which is even better load tolerance to the A70Pro (18%).
I wonder what the input impedance is? The Pre90 was anomalous at just 2kΩ , but the A70Pro was better at 20kΩ.
I agree about quoting Vpp figures instead of Vrms. Very annoying. Who does that?
The reason I'm so keen on headphone amplifiers like this is they also make good pre-amps. I think preamps are rather under-rated, and there aren't that many good ones. You can pay a lot of money for a preamp, if you want to.
They're particularly good for Topping, Benchmark, Hypex and Purifi amplifiers that have (optional) low input sensitivity and sometimes low input impedance (
1kΩ). They need a preamp or buffer stage with high output voltage, and preferably low output impedance and high current delivery. And adequate audio envelope and fidelity not to compromise the performance of the amp.
It's a tall order, but I suspect many buffer stages (even the expensive ones) drag the amplifier down slightly.
Amir has tested several amps at different sensitivity / gain settings, and they all perform best at low gain.
To get a handle on the requirements and achievements of actual buffers, I used data from Purifi tests to generate a system model.
The model has three stages - the source, the buffer, and the power amp.
I treated each stage as an amplifier, with an input voltage, gain, output voltage, and noise and distortion that it adds to the source and feeds to the power amp.
In the spreadsheets, the shaded cells are the input parameters, the orange ones are the variables, and the green one is the output.
To model the tests I supposed the source was an APx555 analyser with an imaginary clean input signal of 1V, plus 12.5dB gain and 1µV noise (DNR 132dB).
The power amp was a 1ET400 with no buffer, 12.8dB gain and 6.6µV noise (DNR 136dB).
By modelling the buffer with 7dB gain and 5.3µV noise, I got the end result to match Amir's amplifier tests at high gain.
Although every stage contributed to the end result, the buffer had a DNR of 125dB, and was clearly the bottleneck in the system.
Alternatively, you can see in the buffer the internal noise dominates the signal noise, yet in the amp the signal noise dominates it's internal noise.
The end result is a system DNR of 124dB (still really good) but limited by the buffer.
I then modelled the preamp as a Topping A70Pro with 7dB gain and (pessimistically) 1µV noise (DNR 139dB)
This time, the overall system DNR is increased from 124 to 130dB.
The signal noise dominates the internal noise in the preamp, and the amplifier output noise is halved compared with the buffer.
Every stage contributes to the end result, but now the system performance is predominantly constrained by the source.
I also performed the same analysis with THD+N, but the conclusions are simpler and clearer by just considering noise.
The expensive Sparkos, Sonic Imagery or Weiss discrete op-amps
measure no better than good integrated solutions, and cost hundreds more.
That money could be spent on a good headphone / preamp like the L30, L50 or A70, and get better performance.
I recently did a money-no-object model with a Topping D900 / A900 combo and 1ET9040 amp, and got an even better result:
