Douk Audio H7 Amplifier Review

[solderdude]

If the output impedance is essentially zero at 3 kHz how can the 0.2dBr difference between the 4 Ohm and 8 Ohm measurement be explained?

Can you calculate once for me how big the frequency response deviation is, for example at 3kHz, if the 4 Ohm and 8 Ohm measurement shows 0.2dBr difference and the speaker has 60 Ohm impedance at 3kHz?

At 3kHz I arrive at (around) 0.2ohm output R (higher at higher frequencies of course) so DF of 20 for 4ohm and DF of 40 for 8ohm at 3kHz.
A bit higher DF below 1kHz.
Note: the slightly lower output voltage may come from the used output inductor (DC resistance), feedback topology/point, PCB traces or wiring to the output so the above calculations may not be accurate but will probably in the ballpark for this amp at the given frequencies.

 

[pma]

If the output impedance is essentially zero at 3 kHz how can the 0.2dBr difference between the 4 Ohm and 8 Ohm measurement be explained?

The output impedance at 3kHz is quite high and the impedance is inductive above 1kHz up to 50kHz, then it jumps to capacitive.
(BTL)

 

[hansik]

Sorry to hear that, but as @tonycollinet said, that's likely a problem with the specific modules, power supply or control circuitry you were using, not anything inherent to the TP3255.

There are many poor implementations around, some even designed with caps and inductors that assume a lower voltage power supply, or without sufficient heatsinking. Even at 90% efficiency, 100wpc is 20w of heat to dissipate. The 3e Audio ones were good but no longer available (https://www.aliexpress.com/item/10000090083368.html)

Once those basics are taken care of a good amp design with the 3255 should have a strongly regulated power supply, ideally with soft start as most modules don't have that built in. The bricks are always a bit suspect to me, as 48V/10A stuffed in a small plastic box like that can't be great. Something like https://connexelectronic.com/product/smps600rs/ would be ideal, but it costs almost as much as this whole amp!

And then there should also be circuitry to hold the chip disabled until the power supply settles, and disable it again before the power is removed, to deal with on/off pops.

I learned these lessons the hard way: https://audiosciencereview.com/foru...-power-amp-with-microprocessor-control.25851/

Yeah, it was years ago. At that time nothing came close to those little TPA3255 Aliexpress amps, soundwise. I was blown away that time by its sound quality, and so cheap. At the end it wasn´t so cheap. I used a Mean Well 48v power supply which was more expensive than the amps )).
Later on I bought the then new Behringer A800 (x3) which costed 154 euros, very powerful and stable. Its price has doubled since.

 

[pma]

The output impedance is essentially zero at 3 kHz. That means the voltage loss relative to 60 ohm with be just the same. So no impact at all.

No, the proof is in the post #122.

 

[thewas]

No, the proof is in the post #122.

So the frequency response of a typical loudspeaker with large impedance peaks in the audio band can have some significant and audible frequency response deviations, correct?
Why doesn't someone just connect such an amp to such an loudspeaker to measure it exemplary?

 

[pma]

So the frequency response of a typical loudspeaker with large impedance peaks in the audio band can have some significant and audible frequency response deviations, correct?
Why doesn't someone just connect such an amp to such an loudspeaker to measure it exemplary?

Yes, it HAS significant deviations. I have shown it here dozen times, so please do not ask me to copy and paste, and as this is a review thread, it would not be welcome, though it could help for understanding the issue in general public.

Let me restrict to one example:

Can I see the response with the complex load from measurements with 4ohm and 8ohm resistors? I do not think so.

 

[thewas]

Yes, it HAS significant deviations. I have shown it here dozen times, so please do not ask me to copy and paste, and as this is a review thread, it would not be welcome, though it could help for understanding the issue in general public.

Let me restrict to one example:

View attachment 278545

View attachment 278546

View attachment 278547

Thank you, truth to be said the deviations below 6 kHz seem to be under approximately 0.2 dB so not really something I would worry about considering audiblity (except the expected peak above 10 Khz which possibly is audible but not significant for most of us >40 YO old audio fan here )

 

[amirm]

Yes, it HAS significant deviations.

It doesn't have hardly any changes at 3 kHz which is what was asked. You can't even see the effect in your measurements. I suggest following the conversation instead of just being insulting.

 

[solderdude]

Damping factor for the woofer is pretty much O.K.
There can be variations in FR above a several kHz depending on the actual load.
This is true for most of these devices using a similar architecture.
How audible this is/becomes depends on the speaker and age and experience of the listener.
Probably sounds fine for most people.

 

[ctrl]

So the frequency response of a typical loudspeaker with large impedance peaks in the audio band can have some significant and audible frequency response deviations, correct?
Why doesn't someone just connect such an amp to such an loudspeaker to measure it exemplary?

There are examples of this, as @dominikz has shown here. Under 4 and 8 Ohm resistive load the amplifier looks pretty good, but already a speaker with "soft" to average dynamic load (he used the Revel M16 as load), the amplifier is actually "no longer usable" and shows between 100-10000Hz more than 1dB ripple:

Source and here @dominikz shows the deviation to a 4 Ohm resistive load.

Naive as I am (no idea of amplifier technology), I thought that one can simply create a virtual loudspeaker load, which is so demanding that 90% of the speakers are covered with it and this load can then somehow simulate the virtual amplifier frequency response based on the 4 and 8 ohm resistive load measurements - which does not work, as I now know.
My virtual demanding speaker load:


So as a layman I am completely confused, because @amirm says a "soft" load dependency has no effect, on the other hand there are measurements that show the exact opposite.

 

[amirm]

So as a layman I am completely confused, because @amirm says a "soft" load dependency has no effect, on the other hand there are measurements that show the exact opposite.

??? I have not said that. You definitely want to get a load independent amplifier if you can afford it. That kind of amp with this level of power simply doesn't exist in this price tier. My explanation was meant to say that the impact is lower than it seems, not that it doesn't exist.

 

[pma]

There are examples of this, as @dominikz has shown here. Under 4 and 8 Ohm resistive load the amplifier looks pretty good, but already a speaker with "soft" to average dynamic load (he used the Revel M16 as load), the amplifier is actually "no longer usable" and shows between 100-10000Hz more than 1dB ripple:

But in his example, the amplifier has about 0.5 ohm output resistive impedance that has an effect even at low frequencies and then there is an interaction in the low and mid band.

 

[solderdude]

I thought that one can simply create a virtual loudspeaker load

The problem with such a load is that it will vary from speaker to speaker.
The load can vary sharply at several points in the frequency response. The load variance can be anything between inductive to resistive to capacitive.
Also there is back EMF that differs per speaker.
So there is no single load that is 'worst case' as some amps are not designed for 2 ohm loads yet other amps may have no issues with that.
It would not be fair to load an amp with a load it will never see. Fun for the one testing perhaps to see if it can but not needed.
Also that load should be able to handle up to 1kW if the goal is to test a wide variety of amps.

 

[ctrl]

??? I have not said that. You definitely want to get a load independent amplifier if you can afford it. That kind of amp with this level of power simply doesn't exist in this price tier. My explanation was meant to say that the impact is lower than it seems, not that it doesn't exist.

But in his example, the amplifier has about 0.5 ohm output resistive impedance that has an effect even at low frequencies and then there is an interaction in the low and mid band.

Thank you both for the explanation.

But does that mean for me as a layman that if I want to be sure that the amplifier should not show more than 1dB ripple even at demanding speaker load, that these cheap but load depending amplifiers are then basically out of the question?

What does "impact is lower than it seems" mean? Is a ripple below 1 dB almost certain with demanding speaker load?

 

[pma]

Also there is back EMF that differs per speaker.

The same "back EMF" you get from the RLC circuit that simulates woofer resonance. Remember the mechanical-electrical analogy. The only difference is the non-linearity of real-life speaker impedance and dummy impedance.

Also that load should be able to handle up to 1kW if the goal is to test a wide variety of amps.

I use the load assembled from 5W - 20W resistors, inductors, MKT/MKP capacitors and electrolytic capacitors successfully for sweeps up to 200W/4R equivalent (= 28Vrms). The load resistors get occasionally hot, but they survive. It is definitely good to use it instead of numerous excuses why not to do it.

To verify FR dependence on a complex load, you may stick with the FR test at say 10W and then there is really no space for further excuses.

 

[thewas]

There are examples of this, as @dominikz has shown here. Under 4 and 8 Ohm resistive load the amplifier looks pretty good, but already a speaker with "soft" to average dynamic load (he used the Revel M16 as load), the amplifier is actually "no longer usable" and shows between 100-10000Hz more than 1dB ripple:

Source and here @dominikz shows the deviation to a 4 Ohm resistive load.

That is truly way too much!

Naive as I am (no idea of amplifier technology), I thought that one can simply create a virtual loudspeaker load, which is so demanding that 90% of the speakers are covered with it and this load can then somehow simulate the virtual amplifier frequency response based on the 4 and 8 ohm resistive load measurements - which does not work, as I now know.

That was also my thought exactly one week ago:

NC252MP (class D) vs. A250W4R (classAB) burst measurements into 4ohm//2.2uF load

Nope. It is you that is not understanding the topic or that measurement. First, show the link to the reviews on stereophile and not just the graphs. Here is the link: https://www.stereophile.com/content/primaluna-dialogue-seven-power-amplifier-measurements Right there, JA explains what is...

www.audiosciencereview.com

 

[ctrl]

So there is no single load that is 'worst case' as some amps are not designed for 2 ohm loads yet other amps may have no issues with that.
It would not be fair to load an amp with a load it will never see. Fun for the one testing perhaps to see if it can but not needed.

My naive idea was, if speaker loads could be simulated based on Amir's measurements, then I would have created a few typical speaker loads as templates and calculated the corresponding amplifier frequency responses with an Octave script to estimate how big the influence of the dynamic load on the amplifier is.

 

[Sokel]

My naive idea was, if speaker loads could be simulated based on Amir's measurements, then I would have created a few typical speaker loads as templates and calculated the corresponding amplifier frequency responses with an Octave script to estimate how big the influence of the dynamic load on the amplifier is.

That is a really good idea if it can be done,I proposed the same in another thread like a wish!
I mean we have punished amps because of their load dependency with bitter comments in threads like that,it would only be fair to deserve it if causes audible effects.
Not specifically in that amp (we're talking beer money here as much as I want to stay far from price points) but as a general rule of thumb.

 

[solderdude]

My naive idea was, if speaker loads could be simulated based on Amir's measurements, then I would have created a few typical speaker loads as templates and calculated the corresponding amplifier frequency responses with an Octave script to estimate how big the influence of the dynamic load on the amplifier is.

You would also need to have a no load (or say 100ohm load) frequency response plot which Amir does not do and may need to use inductive and/or capacitive loads too.
Not as simple as it would seem to be.
With headphone amps one can safely use a 600ohm load as 'no load' plot but for speaker amps these plots aren't made with a 1k or 100ohm load.

 

[solderdude]

The load resistors get occasionally hot, but they survive. It is definitely good to use it instead of numerous excuses why not to do it.

This is the reason Amir does not want to use such resistors as with a different temperature the resistor wire inside can get even hotter and change/modulate the drawn current.

For a quick test it'll work of course or when just evaluating FR deviations quickly with a short sweep.

One would also need a no load or 1k load FR test as a benchmark if one wants to calculate Rout.

Measurements are far from a complete suite. In any case, the measurements that are already there show a load dependency that is substantial enough not to ignored but do not paint the complete picture. Also this is just a cheap amp and for the money it can provide a lot of power.
I would not consider this amp myself as a serious amp but decent enough for the price and form factor. I would not recommend it to anyone that is serious in audio but might for those looking for cheap power. Maybe I am just a bit too neurotic for these kind of devices and would prefer a more serious amp.