John Curl's Blowtorch preamplifier

Status
Not open for further replies.
Jump to Latest
#15,944
john curl said:
Back in the old days, like 40 years ago, we used to think about transistor design in terms of beta, almost exclusively. Transistors were CURRENT magnifying devices, not really voltage magnifying devices, back then, in the minds of most design engineers.
Click to expand...

The beta early voltage product might be a useful process metric, but the basic logarithmic relationship between Vbe and Ic (which holds over MANY decades) is the basis of IC design. Beta is in fact mostly considered a nuisance. The Gummel plot for any decent process is pretty flat (straight) and there are several decades of “flat” beta.
 
#15,945
Yes, I was there when Dr. Pederson mentioned the Gummel-Poon model. Still, the devices operate in the real world, and beta non-linearity is real, and I tend to think in terms of the hybrid-pi model, that I was raised with. Of course, I don't have access to IC devices, except the devices I got in a kit in 1975 or so. Would they be a useful comparison?
I am trying to remove 7th harmonic here, not play games.
 
#15,946
Edmond Stuart said:
Not yet. Are the spec comparable to a Lynx Two for example?
Thanks for the hint Scott.
Click to expand...

Haven't done a numbers comparison. I picked up and old M Audio USB Duo for $100, the drivers sort of suck but the input preamps are OK and it works fine to ring out basic performance stuff. One of the biggest pains is the fact that these devices are not instruments and getting a decent level cal is almost impossible, not to mention the lack of DC performance.

I built a transformer coupled battery powered calibrated broadband noise source (380nV/rt-Hz @25C) that's handy to check against.
 
#15,947
Soundcards are not well suited for direct connection to amps for measurement. And are not really right for preamps. The inputs are fixed for the most part and they can easily be fried from the output of an amp.

Using an audio analyzer in front extends the dynamic range (before damage) a lot.

I think the best deals in usb soundcards are the emu 0202 and 0404. They both have sampling up to 192 KHz and are pretty cheap. The Lynx L2 and L22 are very good but very expensive and still need the input conditioning. They have a distortion floor a little lower (5-10 dB if you go by published specs) but they are all constrained by the adc chips available.

I have a Boonton 1120 analyzer in front of my fft. The auto everything just makes measurements easier and the FFT doesn't need to be exceptional if the analyzer removed the first 100 dB of signal. The elusive 5th, 7th and 9th harmonics are much easier to see. However the looking at the residual in real time synchronized to the source really helps see the source of the problem.
 
#15,948
Someone should do a comparison of the passive prenull/instrumentation amplifier technique vs the massive pile of instruments technique. Knowing the ultimate limit of it would be very useful to the DIY community considering the former could be done by anyone for very little money.
 
#15,949
scott wurcer said:

One of the biggest pains is the fact that these devices are not instruments and getting a decent level cal is almost impossible
Click to expand...

It's actually worse; resulting spectra depend significantly on the input level, and it's not about crappy opamps.

And for all caps freaks, look at all those cheapo electrolytics on the sound card boards 🙂

And the SNR degrades quickly over 20KHz. M-Audio 192 has under 80dB (see my measurements here: http://www.diyaudio.com/forums/showthread.php?postid=1734692#post1734692 ) , my EMU1212 has 94dB at 60KHz and a Lynx Two I was testing had around 100dB. So much about measuring THD20 better than 10ppm without an analyzer front end.

Actually my Ambers have 1V normalized output for both the input signal and for the input signal with the notched fundamental, this is pretty convenient to feed into a sound card. This would make the sound card pretty good for up to 3-5KHz (fundamental) spectra measurements.

Oh, and sound cards don't have any provisions for floating inputs. Cable converting their balanced input to single ended significantly degrades the performance.
 
#15,950
The new Asus soundcard might be one to look at, though it is an internal card, not usb (which has pros and cons).

Some info here:

http://en.expreview.com/2008/10/20/asus-xonar-essence-stx-nice-sound-card-with-snr-up-to-124db.html


Quoting:

"Designed in Essence
Perfectly sealed by improved EMI shield, Xonar Essence STX’s analog output path generates the purest audio ever from a sound card -124dB signal-to-noise ratio and 0.0003% distortion.

Fine-selected Components
With premium-quality digital-to-audio converter and audio-specific capacitors selected, Xonar Essence STX provides <10~90kHZ frequency response and 124dB dynamic range for your unlimited pleasure."


Segran has noted good performance of this card in his measurements: post 8 in this thread:

http://www.diyaudio.com/forums/showthread.php?s=&postid=1742809&highlight=asus#post1742809


Steve.
 
#15,951
If you read through the info on the Praxis system Praxis you will see the hoops Bill Waslo jumped through to make it usable. He created an external USB interface and calibration device to get it to work. His focus was a little different and its not perfect for electronic systems.

The easy software for testing is the RMAA package (and its free). An external box like Scott is implying with RMAA could give a good picture of an amp pretty easily and quickly. I don't know anything about the pro version yet RMAA .

None of this is a substitute for a distortion analyzer and a scope when adjusting bias or looking for a problem.
 
#15,952
syn08 said:

Actually my Ambers have 1V normalized output for both the input signal and for the input signal with the notched fundamental, this is pretty convenient to feed into a sound card. This would make the sound card pretty good for up to 3-5KHz (fundamental) spectra measurements.
Click to expand...

This is what I was talking about except I was refering to the broadband null technique like we used to test capacitors. Low noise in-amps exist, I just don't know how good you can get the common mode linearity to be. Transformers again? 🙂

For some basic THD tests I'm willing to bet you can equal an AP with <$20 DIY.
 
#15,953
scott wurcer said:
Haven't done a numbers comparison. I picked up and old M Audio USB Duo for $100, the drivers sort of suck but the input preamps are OK and it works fine to ring out basic performance stuff. One of the biggest pains is the fact that these devices are not instruments and getting a decent level cal is almost impossible, not to mention the lack of DC performance.

I built a transformer coupled battery powered calibrated broadband noise source (380nV/rt-Hz @25C) that's handy to check against.
Click to expand...

Hi Scott,

I have about the same problems with my (heavily modified) Waveterminal 192X (PCI audio card). The outputs drive a bank of band filters in order to remove the last traces of harmonics and provide galvanic isolation is as well. At 20kHz however, the outputs fail sort. Perhaps the latest generation of op-amps will cure this problem (is there a kind of AD4899 for +/-12V?). Calibration is a PITA and subject to drift.

Regards,
Edmond.
 
#15,954
Scott, back to the beta peak question. I found a good IC model in the Raytheon 'Linear FET Array Design and Applications Manual' from 1991. PP 2-47, 2-48. The NPNM Medium NPN model looks about right to me, except that peak current limitations might get you to use more than one device block in the output. That was your design choice. The peak beta in this example is about 4 ma. This is a typical transistor. Is this in your ballpark? Also, the Gummel-Poon model has been superseded by another model that models beta more accurately.
 
#15,957
scott wurcer said:


Keep checking the catalog we have some new 36V parts coming out of CA. Audio folks are in luck it's no surprise the instrument guys are chasing the same numbers.
Click to expand...

It's already there, ADA4898-1. I managed to pull (tooth style 🙂 ) 2 pcs. from AD, apparently there's a huge log for samples. They are very, very good, competing with the new SiGe devices from TI.
 
#15,958
john curl said:
Is this in your ballpark? Also, the Gummel-Poon model has been superseded by another model that models beta more accurately.
Click to expand...

The output array is sized for min beta and onset of quasi-sat at max output current at 125C. The change in beta from a few ua to a few mA is so slight I would hardly call it a "peak". I think MEXTRAM uses an extended Gummel-Poon model for bipolars. Quasi-sat is far more important than beta since it destroys beta and Ft.
 
#15,959
anatech said:
Hi Bob,
I often look at the residuals from my distortion analyzer. The HP 339A has switchable filters at 30 KHz and 80 KHz. Left out, the bandwidth is very wide. After clobbering the fundamental, a spectrum analyzer's range is extended down somewhat. The spec for the 339A is 0.0018 % residual. Not the best, but not too shabby either.

I use an HP 3585A and I have a (broken) 3580A. The 3580A is nice because it has the adaptive sweep. Cool feature that allows you to ignore the noise floor (by setting a higher limit) and so sweeps take far less time to do. Some day I might get lucky and find a 3562A, like syn08 has.

Some audio sound cards have very good response and a high sampling rate. You may be able to get up to 96 KHz with a 24 bit card. That may work better than a 3580A once you have the interface worked out. The Sound Blaster (or Creative) X-Fi Music shows promise as it isn't bandwidth limited. The EMU 1212 should also be affordable and also has balanced in and out. I almost had mine worked out, but a Microsoft update cured me of that effort. One day I may be inspired to try again.

I hope there is an idea in there for you.

-Chris
Click to expand...


Hi Chris,

Thanks. Yes, I love my old 3580A, and it's adaptive sweep is nice.

About 18 months ago I got a Juli@ soundcard, and it has very good performance. I haven't used it a lot lately, but I think I recall it can sample at 192/24. I've fooled around with a few different free software programs for it, like Visual Analyzer, but a lot of these programs could benefit from better user documentation.

Cheers,
Bob
 
#15,960
john curl said:
Back in the old days, like 40 years ago, we used to think about transistor design in terms of beta, almost exclusively. Transistors were CURRENT magnifying devices, not really voltage magnifying devices, back then, in the minds of most design engineers.
The little amp that I showed you Scott, operated at the BETA PEAK of the output devices. Now I know that you saw the amp, because you and I are in a picture with the amp.
Of course, Ft is also important, but it usually is close to the beta peak, so if you know one, you usually know the other. Usually beta peak is an easier number to find than Ft peak, so I asked for it. Of course VOLTAGE drive reduces the beta nonlinearities' effect on the total distortion, but since you are current driving the output stage, MAYBE, JUST MAYBE, peak beta could be important.
For example, I once VOLTAGE DROVE the HA911 IC op amp and REDUCED THE output stage distortion at least 20 dB. I'll have to tell you how I did it, sometime.
Click to expand...


Hi John,

In many cases, output stages in power amps and op amps are essentially voltage driven over most of their bandwidth, depending on the type of compensation used. For example, a power amplifier using conventional Miller compensation and an output Triple is essentially voltage driven due to the low relative output impedance of the VAS caused by the shunt feedback of the Miller compensation. Such an amplifier is, of course, largely current driven at very low frequencies.

Cheers,
Bob
 
Status
Not open for further replies.
Top Bottom