• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DIY 250W/4ohm amplifier based on "blameless" topology, and measurements

OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Burst power tests with 4ohm//2.2uF load

Tests were performed with 4ohm in parallel with 2.2uF load, with sine bursts (rectangular window) of 1kHz, 10kHz and 20kHz.

4ohm//2.2uF (measured) impedance plots:
4R+2.2uF_dummyload_impedance_REW.png


With amplifier approx. 35Vrms max. output and 4ohm//2.2uF load the output current at 20kHz looks like this:
4R+2.2uF_20kHz_current.png


This is a considerable stress to amplifier output stage requesting 13.4 Arms output current.

Below please see oscilloscope records of the 1kHz, 10kHz and 20kHz bursts. Bursts are generated by DAC at Fs=96kHz, thus the ringing. 35Vrms output voltage makes 306W/4ohm. At 20kHz, we get 430W/2.76ohm impedance magnitude and 1110W with respect to 1.07 ohm EPDR.

A250W4R_burst1kHz_4R+2.2uF.png



A250W4R_burst10kHz_4R+2.2uF.png



A250W4R_burst20kHz_4R+2.2uF.png


I will repeat this test with Hypex NC252MP soon.
 
Last edited:
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
A small, 0.45uH inductor wound on 6.8 ohm resistor (parallel) was added to the output to get ultimate stability with capacitive loads from 0pF to several uF.

For those experienced in electronic circuit analysis, below shown the loop gain analysis with 4ohm in parallel with 0pF, 100nF and 2uF. Loopgain analysis is crucial to understand amplifier stability.

A250W_loopgain_capload.png


Inductor placement
P1050362-2.jpg



Step response (the most important measurement to know amplifier stability) with 4R or 4R//47nF load (indistinguishable)

P1050367-1.JPG
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,834
Likes
16,496
Location
Monument, CO
@pma - Looks like negative phase margin around 130~140 kHz with the 2 uF load, but the step response doesn't really show anything... Maybe because (loop) gain is pretty low already by then? Or parasitics are dropping the gain lower than the sim?

I do not know how you defined Vout and VA. In general I use Middlebrook's loop analysis technique, to get a more complete loop response, though it has been refined and superseded by later art. The simple voltage-source scheme, or using a large LC to break the loop, did not work well for me -- I got bogus (wrong) answers that went both ways, with more or less margin than the actual circuit exhibited.

One question: It looks like the inductor is outside (after) the feedback loop? Can't really tell from the picture.

Did it already have a snubber for RFI?
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
I do not know how you defined Vout and VA. In general I use Middlebrook's loop analysis technique, to get a more complete loop response, though it has been refined and superseded by later art. The simple voltage-source scheme, or using a large LC to break the loop, did not work well for me -- I got bogus (wrong) answers that went both ways, with more or less margin than the actual circuit exhibited.
Yes, standard routine way of loop gain analysis, you open the loop and put the voltage generator into the loop. Input is shorted. This is a common way of amplifier loop gain simulation in MC and other simulation tools.
LC method is unreliable.

One question: It looks like the inductor is outside (after) the feedback loop? Can't really tell from the picture.
Sure, it would not make much sense to put it inside the loop - the stability would not be improved. Almost every class AB power amp AFAIK has the output inductor and of course outside FB, or there are 2 loops used. And it is not only the inductor, it is L//R.

Did it already have a snubber for RFI?
Sure. But the Zobel is not only for RFI, its function is more complex.

A250W_loopgain_capload2.png


We have 2 conditions - at the cross section of amplitude plot the phase must not fall below -180°. At the cross section of phase plot with -180° the amplitude plot is to be well below 0dB.
Please do not be confused by negative phase sign. It only depends on polarity of the injecting voltage source. Reverse + and - and you are in same but positive numbers.

Frankly - some of the questions are surprising to me.

From the attached paper:

1680900309230.png


The method used is that with the X4/GV source.
 

Attachments

  • Simulating Loop Gain - Spring 1997.pdf
    114.2 KB · Views: 120
Last edited:

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,834
Likes
16,496
Location
Monument, CO
Yes, standard routine way of loop gain analysis, you open the loop and put the voltage generator into the loop. Input is shorted. This is a common way of amplifier loop gain simulation in MC and other simulation tools.
LC method is unreliable.
Yes, as I mentioned I have found using a single voltage source is not always accurate (and most certainly not the LC method) but it is topology-dependent. Here is a link to an overview of Middlebrook's (and Rosenstark’s) loop analysis: https://www.edn.com/middlebrooks-and-rosenstarks-loop-gain-measurements/ (just a quick search; my Middlebook text is buried). The original derivation I have (directly from R. D. Middlebrook at CIT) uses voltage and current transfer functions in separate circuits then combines the responses to get the "true" open-loop response. There are some drawbacks even with Middlebrook, among them that it doubles the circuit complexity for the simulation. Not usually an issue for me since very high frequency circuits tend to be small, but for a full audio amp or op amp it can get tricky. Less the transistor count than getting initial d.c. convergence, at least in my experience.

Sure, it would not make much sense to put it inside the loop - the stability would not be improved. Almost every class AB power amp AFAIK has the output inductor and of course outside FB, or there are 2 loops used. And it is not only the inductor, it is L//R.
Yes, that is what I assumed, just a sanity check.

Sure. But the Zobel is not only for RFI, its function is more complex.
Again, I figured it was in there, was just curious. I had a vague memory of someone mentioning they had not added the snubber/Zobel but could not remember who said it (or when). I am aware of its other features for helping stability.

View attachment 277783

We have 2 conditions - at the cross section of amplitude plot the phase must not fall below -180°. At the cross section of phase plot with -180° the amplitude plot is to be well below 0dB.
Please do not be confused by negative phase sign. It only depends on polarity of the injecting voltage source. Reverse + and - and you are in same but positive numbers.
That was completely on me. I am used to plotting so the 180 degree phase crossing is at the same point on the y-axis as 0 dB loop gain just to make it easier to see when there is no margin. In my quick look I did not read the y-axis legend closely enough. Sorry about that.

Frankly - some of the questions are surprising to me.
Yeah, well, sometimes I just do Stupid Don Tricks. Keeps me humble in the presence of all you smarter guys.

From the attached paper:

View attachment 277785

The method used is that with the X4/GV source.
Oh, so you are already using Middlebrook's method! Some people (usually undergrad students or hobbyists) just stick a voltage source (X4) in the loop and measure across it. That is, or can be, badly inaccurate without using the current source and considering the total response. Again I was asking as a sanity check, not knowing exactly what you were doing and the same sort of basic questions I would ask in any review.

Back to lurking - Don
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Microcap 11 also has a direct tool called “Loopgain Probe”:

From the manual: Loop Gain Probe
The device is used to specify the point where a circuit loop is to be broken for the purpose of measuring gain and phase margins in Stability Analysis.
The direction of the probe does not matter if the Middlebrook method is used.


It is used with the function “Stability Analysis” directly. It plots loopgain amplitude and phase and also Nyquist diagram. It gives absolutely equal results as I have shown above with another method.

Below one can see the effect of the output inductor on stability with capacitive load.

Loopgain plots with the 0.45uH inductor - it effectively isolates the load capacitance
Loopgainprobe_L.png


Loopgain plots without the inductor
Loopgainprobe_noL.png

Capacitances list: 0, 100n, 200n, 500n, 1u, 2u

In a real amplifier, there are inductances of wires from PCB output to panel binding posts and the effect of the load capacitance is decreased. The critical area is usually between 47nF - 470nF, in the real life FB amplifier. With the additional inductor, there are usually no issues.

The special issue is a distributed capacitance of long speaker wires. Though it is usually no higher than some 2nF at audio frequencies, in MHz range it is up to uF values, at the points of complex impedance dips. This fact is usually omitted. But some members here like @KSTR are well aware of this issue.

------------------------

As we know, the NC252MP failed in the test with 4R7//2.2uF load and also in the test with 4R7//(2.2uF+1.6uH+0.3ohm) load. If we knew the loopgain plots, the same plots that I have shown here, we would know why it failed. But, such info is usually only available to the designer of the amplifier, as it clearly shows the limitations of the design. Nothing is perfect and we have to make a choice between various compromises.
 
Last edited:

Digital_Thor

Senior Member
Joined
Mar 24, 2018
Messages
381
Likes
330
Location
Denmark
Great work PMA :D

I'm so happy that I bought some good, strong AB modules with an extra driver stage, back when I started to build DIY. Running active as well, so my chances of running into problems are definitely lower.

My amplifiers have proved their worth several times, when used on passive speakers too. I clearly remember a set of Von Sweickert and B&W in particular, which really benefited from the amplifiers' ability to stay stable and resilient towards difficult loads.

The first time I truly saw the difference, was when I tried the Groundsound amplifiers on a LAT500 subwoofer. The membrane simply swung less, the bass was deeper, and I could play significantly louder. I quickly learned, that 200W from a Vincent SAV P-200 was something totally different than the 200W from the Groundsound modules.
http://groundsound.com/pa3cc.php
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Great work PMA :D

I'm so happy that I bought some good, strong AB modules with an extra driver stage, back when I started to build DIY. Running active as well, so my chances of running into problems are definitely lower.

My amplifiers have proved their worth several times, when used on passive speakers too. I clearly remember a set of Von Sweickert and B&W in particular, which really benefited from the amplifiers' ability to stay stable and resilient towards difficult loads.

The first time I truly saw the difference, was when I tried the Groundsound amplifiers on a LAT500 subwoofer. The membrane simply swung less, the bass was deeper, and I could play significantly louder. I quickly learned, that 200W from a Vincent SAV P-200 was something totally different than the 200W from the Groundsound modules.
http://groundsound.com/pa3cc.php
Looks good with 3 pairs of MJL3281/1302 devices. They are fast enough, with good linearity at high currents, and good enough SOA. The distortion is a bit higher, but still at inaudible level. The key is stability, unconditional stability into complex load.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
A250W4R THD+N vs. output voltage at 1kHz into 4.7ohm resistor (green) and 4.7ohm//(2.2uF+1.6uH) complex load (orange), BW22kHz. The results are almost identical.

A250W4R_THDN_4R7_vs_complex.png
 

777

Senior Member
Joined
Jan 1, 2019
Messages
486
Likes
359
I don't think that between 100hz and 1Mhz the total equivalent capacity can increase a thousand times. However, the dielectric of the speaker cable insulation is quite stable, with relatively constant parameters with respect to frequency. I shall measure the capacity of a speaker cable between 100hz and 100khz.
 

Attachments

  • 20230408_202212.jpg
    20230408_202212.jpg
    90.2 KB · Views: 40
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
I don't think that between 100hz and 1Mhz the total equivalent capacity can increase a thousand times. However, the dielectric of the speaker cable insulation is quite stable, with relatively constant parameters with respect to frequency. I shall measure the capacity of a speaker cable between 100hz and 100khz.
Where did I say that between 100Hz and 1MHz the total equivalent capacitance would rise 1000x ??

Electrical engineering is about facts, not about feelings, the cable impedance is to be considered in its distributed parameters, the impedance then is shown below, for a 7m cable. If the dip at 4.5 MHz is in coincidence with specific amplifier loopgain plot, then you are in troubles.

7mcable_imp.png


Below is a real life oscillation story with a class AB amplifier and 10m of output cable, fixed later by the output inductor.
oscill_10m_TaskerC121_sm.jpg




In case that cable length is Lambda/4 (use corrected velocity of propagation for frequency calculation), the input impedance of the ideal cable with open end is zero, short. The cable terminated by loudspeaker is close to open at frequencies above 1MHz, as its impedance is inductive there. Long speaker cable thus may bring unexpected problems that many people are unable to realize and to explain. Simplistic view of the cable as a capacitance and inductance, that is valid in audio band, is unacceptable at high frequencies. Terminating of the cable at the speaker end by resistor equal to cable characteristic impedance prevents such problems. Or use Zobel R-C to avoid power dissipation in the resistor.
 
Last edited:
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Measuring and simulating amplifier step responses with capacitive load up to 2.2uF

1. Measured step responses

A250W4R_stepresp_4R7@47nF.png


A250W4R_stepresp_4R7@2.2uF.png



2. Simulated step responses

A250W4R_stepresp_simulated.png


We can see that circuit theory and physics both work quite well.
 

777

Senior Member
Joined
Jan 1, 2019
Messages
486
Likes
359
Where did I say that between 100Hz and 1MHz the total equivalent capacitance would rise 1000x ??

Electrical engineering is about facts, not about feelings, the cable impedance is to be considered in its distributed parameters, the impedance then is shown below, for a 7m cable. If the dip at 4.5 MHz is in coincidence with specific amplifier loopgain plot, then you are in troubles.

View attachment 278063

Below is a real life oscillation story with a class AB amplifier and 10m of output cable, fixed later by the output inductor.
View attachment 278065



In case that cable length is Lambda/4 (use corrected velocity of propagation for frequency calculation), the input impedance of the ideal cable with open end is zero, short. The cable terminated by loudspeaker is close to open at frequencies above 1MHz, as its impedance is inductive there. Long speaker cable thus may bring unexpected problems that many people are unable to realize and to explain. Simplistic view of the cable as a capacitance and inductance, that is valid in audio band, is unacceptable at high frequencies. Terminating of the cable at the speaker end by resistor equal to cable characteristic impedance prevents such problems. Or use Zobel R-C to avoid power dissipation in the resistor.
Thank you, now I understand you.
 

Damnati

Member
Joined
Mar 12, 2024
Messages
5
Likes
1
Hi there, where can I find all the PCB plans and parts lists for this amp? I really want to build it. It would be nice if you would share it! This project looks fun!

Thanks!
 
Top Bottom