Some Interesting Drivers, a New 3-way Project

Why would one design an entire amplifier, if they can just be bought for cheap everywhere?
That sounds like reinventing the wheel to me for additional costs.

Or why would anyone purchase a dedicated test amplifier, if they can just be bought for cheap everywhere?
That sounds like reinventing the wheel to me for additional costs.

If doing it as cheep as possible is a priority then sure.

If reliable factory floor QA / QC or reliable DIY bench accuracy is the priority then cobbling things together is not worth the aggravation.

For me time and quality is the priority over low cost. If the next DIY person uses the same system then we can compare notes.

The new DATS may be a fine integrated speaker test system.
 
If doing it as cheep as possible is a priority then sure.
I never said that nor did I say anything about cobbling things together.
I said that there are plenty of good performing amplifiers that could be bought for cheap elsewhere.

That doesn't mean that the amplifier is cheap in sense of performance.
It means that those amplifiers have a very good performance price-quality ratio because they can be produced and manufactured in high numbers.

A DATS system will an can never reach those quantities, since it's a niche product = it has less use cases.
So you're basically trying to fight a game that you can never win in sense of amplifiers.

Not to mention that many people already have a dedicated test amplifier, or can be bought 2nd hand.

For large signal testing we also don't need amplifiers with insane distortion specifications.
Klippel actually has quite some documentation on this as well.
If they can integrate any 3rd party amplifier reliably, without "cobbling that together", there is no reason why others can't as well.

Your response to this was "Calibrated input => calibrated output".
I was saying that it's a non-issue since a total system calibration (with any amplifier) is very easy to do.
In fact, this is also supported by Klippel.

Or do you mean to say that Klippel's system isn't reliable for factory floor QA / QC or DIY bench accuracy?
Which I find hard to believe for a company that is basically being seen as the de-facto standard in the industry.

But since I am willing to listen, I would like to hear the ideas behind it.
Also, I might hope that it has thermal and DIS function (motor stability) as well?
Otherwise large parameters still don't mean much.
 
Last edited:
I built the plinth/base under the speaker cabinet, and assembled the crossover into the plinth. Eventually the plinth will be veneered and trimmed to match the rest of the cabinet, and the temporary feet will be replaced with a more decorative foot structure.

The wiring for the filter network looks a bit messy, but everything is clearly labeled. The two adjustable resistors are easy to access.
1726252093215.png

1726252155675.png


For validation/verification, or what some would call quality control, I compared the measured impedance to the simulation impedance. I don’t intend this to be a comprehensive set of measurements, but just a check to make sure I did not assemble the filters wrong.

Here is the impedance curve of the two speakers compared to simulation. The two measured speaker responses (dashed lines) lie on top of each other, which pleased me. They also match the simulation quite well, except for the known measurement error discussed earlier.

1726252263212.png


Another validation/verification comparison was a quick frequency response scan on the tweeter axis. This was done in my shop, and the longest reflection-free gate I can get is about 3.5 ms. In this case, I got even less, 3.3 ms. So I am showing the response from 300 Hz up since the measurement looses validity below that. This shows the measured response is matching simulation reasonably well.

1726252190190.png




I will do some longer gate measurements later on, as well as outdoor ground plane measurements, but for now I am doing some subjective listening evaluations.

j.
 
One thing that became very obvious right away was the tuning of the passive radiator. Back in post #422 I made some measurements of the woofer+PR with three levels of added mass: 30 g, 15 g, and 0 g. This resulted in a small measured response difference, and a shifting of the tuning frequency from 33 to 38 Hz.
https://www.diyaudio.com/community/...ivers-a-new-3-way-project.413182/post-7775930

The subjective difference was apparent after just a few minutes of listening. The higher tuning (0 g added mass) sounded muddy and congested, meaning a lack of low frequency detail. Switching to 30 g added mass made a noticeable improvement. Bass was tighter and more detailed. Even the upper bass and lower midrange was improved.
 
One thing that became very obvious right away was the tuning of the passive radiator. Back in post #422 I made some measurements of the woofer+PR with three levels of added mass: 30 g, 15 g, and 0 g. This resulted in a small measured response difference, and a shifting of the tuning frequency from 33 to 38 Hz.
https://www.diyaudio.com/community/...ivers-a-new-3-way-project.413182/post-7775930

The subjective difference was apparent after just a few minutes of listening. The higher tuning (0 g added mass) sounded muddy and congested, meaning a lack of low frequency detail. Switching to 30 g added mass made a noticeable improvement. Bass was tighter and more detailed. Even the upper bass and lower midrange was improved.

@hifijim

Thanks for the subjective listening impressions.

It is likely that the outputs of the Woofer and the Passive Radiator are more than a few degrees out of phase and perhaps smear the listening experience but less so with the lower PR tuning frequency. There is a mountain of stored energy in compressed air of the enclosure and "spring" mounted weights attached to the PR.

I am thinking about microphones placed near field at the woofer and at the passive radiator to compare phase.

Also thinking of Bass through lower mid-range HD and IMD measurements. Enclosure resonance with the addition of PR resonance could produce some interesting summing and canceling wave forms. Think two-tone FFT plots (frequency domain) and o'scope time domain plots.

Thanks DT
 
Fascinating. I wonder whether that relates to a better match for a smoother in-room response… or something else.
My guess is that, yes, it relates to the perceived tonal balance... in other words, the frequency response. But at this point that is a guess.

I am thinking about microphones placed near field at the woofer and at the passive radiator to compare phase.
I wonder if this audible differences has something to do with differences in groupdelay. For me your comment sounds like a timing difference.

Timing/phase issues are a possible culprit... I have some data which could shed some light on this line of thought. I will try to post it later today.

Xos look like LR2 acoustic, which are my facourite. Jim, how do you like piano, ac. guitar etc. sound?

Yes, crossover is at 400 Hz LR2 and 2400 LR2. I was hoping for a 3k upper crossover, but in order to manage the mid driver response, I had to move the crossover frequency down to 2.4k, and this slightly sacrificed the directivity performance. Piano sounds very nice, although my current main system is a Purifi mid driver + wave guide beryllium tweeter, and that does an outstanding job on piano... so I have to keep that in mind when making comparisons...

j.
 
  • Like
Reactions: Juhazi and stv
Lower reflex tuning close to 30Hz makes GD much better ar say 42hz. It must sound better!
I think that should be, "It could sound better".

Based on the literature, the audibility of group delay in real-world music and practical listening environments is still a topic of debate.
Most claims I've come across are anecdotal and (often) not (always) tested in an objective manner.

So, we can't really draw any conclusions.
 
Last edited:
  • Like
Reactions: A4eaudio
It is likely that the outputs of the Woofer and the Passive Radiator are more than a few degrees out of phase and perhaps smear the listening experience but less so with the lower PR tuning frequency. There is a mountain of stored energy in compressed air of the enclosure and "spring" mounted weights attached to the PR.

Here is some data to chew on... To me, it does not seem like timing/phase issues would explain the muddy/congested quality of the 0-g added mass tuning, but I would like to hear opinions on this.

First is a comparison of the merged woofer+PR response including the effects of the woofer low pass filter.

1726344414789.png

1726344439464.png


Next we look at the woofer and PR responses separately. In this case, these are the near field responses without the low pass filter

1726344464871.png

1726344511842.png


1726344541492.png
 
  • Like
Reactions: DualTriode
To me, it does not seem like timing/phase issues would explain the muddy/congested quality of the 0-g added mass tuning, but I would like to hear opinions on this.
How are things with the level of nonlinear distortion at the woofer output with different masses on the passive radiator? It is quite possible that the resonance frequency of the passive radiator is shifted down in frequency, the woofer is unloaded at this frequency and the woofer distortion is reduced.
 
  • Like
Reactions: DualTriode
Hello All,

To avoid mixing things up Phase (delay) addresses a single frequency and Group Delay addresses possibly many frequencies as a complex waveform, two different things.



@hifijim,

First blush We* always think of Frequency Response as most responsible for listener preference. The muddy impression may be because of the change in FR. Perhaps add a little DSP equalization to determine if that makes the difference.

Looking at your Woofer and PR Near Field (2pi) plot We see the largest Phase (delay) centering near 50Hz at 30 + degrees of delay. The PR with the added 30g mass has much lower Phase (delay). Combining the driver output and the PR with 0 added weight PR We can picture a time smeared output.

Still thinking of bass through lower mid-range HD and IMD measurements. Enclosure resonance with the addition of PR resonance could produce some interesting summing and canceling wave forms. Think two-tone FFT plots (frequency domain) and o'scope time domain plots. Thinking of distortion testing.

Thanks DT

*
Now that there is the opportunity to change our pronouns:
I am now We; Who I was was, Who I am and who I will be, always he.
 
Last edited:
  • Like
Reactions: hifijim
Next we look at the woofer and PR responses separately. In this case, these are the near field responses without the low pass filter
How is there such a frequency response difference between the 2pi and 4Pi without/with the low pass filter? In the 2pi graph the difference between 40 and 80Hz seems higher, in the 4Pi it looks less.

In the 4pi graph it hard to imagine either would be boomy, in the 2Pi graph the response of the 0g with a rise below 80Hz boom seems plausible.
 
  • Like
Reactions: hifijim
How is there such a frequency response difference between the 2pi and 4Pi without/with the low pass filter? In the 2pi graph the difference between 40 and 80Hz seems higher, in the 4Pi it looks less.
Because some dingbat was in a hurry to post plots. The aforementioned dingbat scaled the passive radiator response by 20*log(area ratio) rather than 10*log(area ratio). This means the PR response was incorrectly scaled by 4 dB, rather than the correct 2 dB.

1726359620000.png


@DualTriode and @uriy-ch - - I will run some distortion measurements of the baseline tuning (30 g added mass), but I will not be removing the PR to change the mass again. So I will not be able to make a distortion comparison between the two tunings. I do not want to run the risk of damaging the screw threads in the plywood, so I only plan to remove the drivers one more time for veneering and finishing. Thank you both for your thoughts.
 
About GD hearability, I ask you to listen and think of eg. upright bass 1/8 notes picked one after another or bass drum kicks. Looking at waveform with eg. Audacity you can see that fundamentals are just 1-3 cycles. Just leaves me thinking about how it sounds when fundamenal (coming out of port/PR) is full cycle late compared to harmonics...
 
Last edited:
  • Like
Reactions: hifijim
40Hz would be 40 cycles a second, so if one cycle was 1/8th note tempo of the song is 20 beats per second, or 1200 beats per minute. Not even the fastest fusion jazz is this fast!🙂 If it's average 120bpm song, 40Hz makes ten full cycles within an 1/8th note. One beat is 60/120 so 500ms, so 1/8th note would be about 250ms long. Still, ear can detect very short delays, it's just hard to differentiate how much it is in numbers. Blink of an eye is about 100ms, feels fast as you do it, but is actually quite slow. Try and do overdubbing with +5ms delay in the digital recording chain and it's difficult, audible error, although it's mere ~50cm of delay, as if your quitar was in your knees instead of in your lap. So delays you heqr might be true, but likely much much faster than 1/8th note, perhaps related to all microstuff that haopens within an eight note. Weird but fun stuff to think about.
 
  • Like
Reactions: hifijim
Just leaves me thinking about how it sounds when fundamenal (coming out of port/PR) is full cycle late compared to harmonics...

It is not Group Delay you are talking about.

Group Delay is the entire complex waveform being delayed with the waveform intact.

Phase (delay) is each frequency delayed a different amount depending on wave length. The complex waveform made up of multiple frequencies is mangled, some frequencies delayed more than others.

Do not get GD confused in you brain.

Think of something that you can see like a impulse response plot of the near field bass driver and a separate impulse response plot near field at the Passive Radiator. The IR at the PR will be delayed and a different shape.
 
  • Like
Reactions: hifijim and Juhazi
I made a scan on the tweeter axis, with a time window of 4.3 ms. In the plot below I show the solid line is measured response, the dashed line is simulated (but with different resistor values).
View attachment 1352806

I am curious as to why the crossover region between the midrange and the tweeter is so wide. This seems to come about as a result of the relatively slow roll-off rates in the relevant low-pass and high-pass filters. Did you have some particular performance characteristics in mind? Your components seem to have the ability to create at least 2nd-order behaviors, but the implementation seems to follow a 1st-order behavior.