Bob Cordell Interview: Error Correction

Jump to Latest
#2,602
Re: Halcro

Edmond Stuart said:


Hi Bob,

I'm afraid that the findings of JA are indeed correct (read: a much higher distortion), the more so as my simulations also reveal that it is almost impossible to beat the 1ppm barrier without additional highly sophisticated tricks (unless one falls back to class-heat, which is unacceptable from an engineering point of view)
So I'm more than inclined to agree with you on this matter.

Cheers, Edmond.
Click to expand...


Yep, I'm not surprized, as I had looked very carefully at many different measurements JA did over many different Halcro's, and compared those to numerous other amplifiers he tested, such as the Boulder.

Cheers,
Bob
 
#2,605
PMA said:
There was a follow-up (dm-88). JA had oxidized banana plug binding posts /on his test load/ and measured pseudo-distortion.
Click to expand...


Hi Pavel,

Actually, I think the follow-up was on the DM-38 or DM58, and it was only a couple months ago. I think it is on the Stereophile site. Some distortion measurements did indeed improve, but not by enough to confirm Halcro's claimed distortion numbers.

Cheers,
Bob
 
#2,608
Edmond Stuart said:

Do you have a link?
Click to expand...


That story was about the Halcro DM38. Originally JA measured a relatively high THD numbers, then he published an errata (claiming some issues with the test load linearity, at Bruce Candy's suggestion) here:

http://www.stereophile.com/amplificationreviews/1004halcro/index5.html

Honestly, I am having a hard time buying this explanation. On another occasion when JA measured performances away from the spec, BC claimed that some part was out of tolerance (does it sound familiar 🙂 and sent JA another unit.

Personally, I had the chance to do some measurements on a DM-58 and I can tell that BC's spec is overrated by a factor of 5-8.

All the above does not mean that the Halcro can't reach it's spec, but only that most likely BC's Halcro has manufacturing issues. Which is easy to understood given the nature of the required EC adjusting/trimming.
 
#2,609
syn08 said:



TOn another occasion when JA measured performances away from the spec, BC claimed that some part was out of tolerance (does it sound familiar 🙂 and sent JA another unit.
Click to expand...

But, as I and you have already linked, JA measured the same high distortion on 2nd sample. Then he found oxidized connection on his load and got far better results.

I would tend to believe it, as I know what makes iron metal screw or bad contact in the test load, especially for such low level of distortion.
 
#2,610
PMA said:

I would tend to believe it, as I know what makes iron metal screw or bad contact in the test load, especially for such low level of distortion.
Click to expand...

I'm not debating this effect. What worries me is that a highly regarded measuring shop was not aware of such potential issues before the Halcro came in. At these THD levels, you'd better know what you are doing and be prepared for all kind of strange effects. If I would be JA, I wouldn't publish anything until I would be sure that the results are consistent and reproducible.

Here's one from my experience:

I was measuring the output spectra in our amp and noticed some spurious components that were acting like noise (but with a limited bandwidth). These artifacts were affecting the floor noise while measuring our amp. We already knew from back to back measurements that the differential noise floor was at -160dB, so why was the noise floor only some -128dB when measuring the amp?

The culprit was a defective cooling fan. For whatever reason, that fan was radiating a KHz range noise, even if it was fed from a separate 12V power supply. Replaced the fan and everything went back to normal.

We are still experiencing some very strange 60Hz odd harmonics that extend well into the KHz range (mentioned on the web site). I am currently assuming these are due some DC imbalance effects in the power toroid, but we won't really know until I'll replace the toroid with another model, sometime this weekend.
 
#2,612
Hi syn08,
Most technicians are not that aware of test and measurement issues and only learn once they have been stung at least once. For measuring distortion and noise down at these levels, there can be no mistakes, An intermittent load (even slightly) would show up big time as the currents came up. Now he knows.

Just wait until he gets into dissimilar metal contacts and thermocouple effects! Most technicians would have no concept of these issues. Mind you, there is no need for them to worry about it either on average.

You, on the other hand, saw something not expected and it didn't look right. What is great is that you took the time to trace the fault and correct it. It takes time to make good, valid measurements. Most people will not take that time from my observations. That's what it takes to get it right.

How many of use have a book like "Calibration: Philosophy in Practice" published by Fluke to refer to? How many took a course in instrumentation in post secondary school? When you start getting down to the nitty gritty of small measurements, you need to know this stuff, or at least have something to refer to for the answers.

-Chris
 
#2,614
syn08 said:


I'm not debating this effect. What worries me is that a highly regarded measuring shop was not aware of such potential issues before the Halcro came in. At these THD levels, you'd better know what you are doing and be prepared for all kind of strange effects. If I would be JA, I wouldn't publish anything until I would be sure that the results are consistent and reproducible.

Here's one from my experience:

I was measuring the output spectra in our amp and noticed some spurious components that were acting like noise (but with a limited bandwidth). These artifacts were affecting the floor noise while measuring our amp. We already knew from back to back measurements that the differential noise floor was at -160dB, so why was the noise floor only some -128dB when measuring the amp?

The culprit was a defective cooling fan. For whatever reason, that fan was radiating a KHz range noise, even if it was fed from a separate 12V power supply. Replaced the fan and everything went back to normal.

We are still experiencing some very strange 60Hz odd harmonics that extend well into the KHz range (mentioned on the web site). I am currently assuming these are due some DC imbalance effects in the power toroid, but we won't really know until I'll replace the toroid with another model, sometime this weekend.
Click to expand...


I agree with what you have said here. I have spent many years with instrumentation for measuring very low distortion levels (some of it specialized, like the Distortion Magnifier), and I know how very difficult it can be to do. However, overall I think that JA does a consistently good job. Although we all get caught with a dirty connector now and then, the overall evidence from numerous Halcro reviews and in comparison with other reviews is that what JA is seeing is largely real. I believe in many ways the Boulder monoblock showed lower distortion than the Halcro, and yet both were well above Halcro's spec.

Cheers,
Bob
 
#2,615
Bob Cordell wrote:
Hi Edmond,

We'll see. It sounds like we have a challenge here 🙂.

Indeed, I have not ruled out TMC, but I am not yet certain I need it to get there.

First of all, it depends a little on how we set the goal. If I set the goal as 100W into 8 ohms being 1 ppm THD-20 in a 80 kHz BW, I can almost certainly make it, I think. However, you would, perhaps rightly, say that was cheating, since your result is into 4 ohms.

My existing 6 ppm into 8 ohms, to the extent that it is dominated by the output stage, will drop down to 2 ppm or less just by adding the two additional output pairs to the output stage and the consequent increase in bias to 450 mA that you have.

I do agree that I have no plans at all to increase the closed-loop bandwidth to beyond 2 MHz.

I would further add that a perhaps better goal would be to have all distortion components of the CCIF twin tone 19+20 kHz test be below -120 dB at near full power into 4 ohms.

Your fine design has indeed inspired me to try a little harder 🙂.

Cheers,
Bob
Click to expand...
How the goals are set is key to this competition. As far as I can see, E&O's THD20 achievement has been at the expense of complexity and stability. It is no wonder because this thread has been focussed on the primary goal of THD.
I asked the question about whether stability matters to sound quality a few pages back. There hasn't been any interest in this question so far. Does that indicate that stability is not very important; does it indicate that it is important but the importance is not appreciated?

When Edmond bullishly claims that Bob's circuit is inadequate compared to his he may or may not be right but so far he isn't making a fair comparison.
Certain parameters need to be fixed, like:
Output transistor type
Total OS bias current
Test conditions: load, frequency, avg power
THD/IMD
Stability margin
Output Z

We know that both circuits use nested negative feedback loops to "control away" distortion. We also know that the reduction in distortion will be limited by the linearity of the correction circuits and will be increased as stability declines...until the circuits oscillate. We know that the stability and distortion depend on the loading.

Bob & Edmond, how about trying to enumerate a more comprehensive comparison between your two circuits?
 
#2,616
Edmond wrote:
Don't worry. We ordered not only the trannies from Toshiba but also one ounce of audio grade complementary electrons. They feel quite happy while pushing the slew rate to a maximum of 500V/us.
Click to expand...
Ah. You didn't say that your design was born at CERN. That makes all the difference. 🙂
BTW, why should I be impressed by 500V/us? Why does this matter?

Your prey, Halcro, uses a single LTP input. Why do you think they chose to do this?
 
#2,619
traderbam said:
Bob Cordell wrote:
How the goals are set is key to this competition. As far as I can see, E&O's THD20 achievement has been at the expense of complexity and stability. It is no wonder because this thread has been focussed on the primary goal of THD.
I asked the question about whether stability matters to sound quality a few pages back. There hasn't been any interest in this question so far. Does that indicate that stability is not very important; does it indicate that it is important but the importance is not appreciated?

When Edmond bullishly claims that Bob's circuit is inadequate compared to his he may or may not be right but so far he isn't making a fair comparison.
Certain parameters need to be fixed, like:
Output transistor type
Total OS bias current
Test conditions: load, frequency, avg power
THD/IMD
Stability margin
Output Z

We know that both circuits use nested negative feedback loops to "control away" distortion. We also know that the reduction in distortion will be limited by the linearity of the correction circuits and will be increased as stability declines...until the circuits oscillate. We know that the stability and distortion depend on the loading.

Bob & Edmond, how about trying to enumerate a more comprehensive comparison between your two circuits?
Click to expand...

Brian,

You put a large amount of questions, concerns, suggestions in the same post. Each of these are worth a separate discussion and I would suggest doing so, either by posting new threads or new posts under this or another appropriate topic.

I am personally having troubles in either understanding or agreeing with most of your statements. It would certainly be helpful explaining and substantiating your statements, perhaps with some theoretical background or measured data.

Thanks again for your input.
 
#2,620
Ovidiu wrote:
Brian,

You put a large amount of questions, concerns, suggestions in the same post. Each of these are worth a separate discussion and I would suggest doing so, either by posting new threads or new posts under this or another appropriate topic.

I am personally having troubles in either understanding or agreeing with most of your statements. It would certainly be helpful explaining and substantiating your statements, perhaps with some theoretical background or measured data.

Thanks again for your input.
Click to expand...
I don't think I've introduced anything that hasn't already come up in this thread, not all by me, and I believe those parameters are all relevant and important when comparing EC schemes.
Can you be more specific about what is troubling you?
 
You must log in or register to reply here.
Top Bottom