QuantAsylum QA400 Audio analyser

[abbey road d enfer]

The input impedance switching is unduly complex IMO.
In the 600 ohm position, the level is attenuated 3dB.

Excellent, thank you!  Would it be better to put resistors to ground, or just strap a resistor across + and -?  I thought I would try to use the 1200's common mode input, but I probably don't need to.

  With two resistors, the CMRR is dependant on their tolerance (maybe not so with the CM impedance bootstrap of the 1200 - I haven't give it much thought). One resistor across + and - is just simpler and does not affect CMR. But you should rally do the switching right at the input, not after the protection res.

 

[ruffrecords]

Just a couple of comments. Does it need to be that complex?  The fewer components in the signal path the better and the less they are stresses the better. So on the DUT to test gear side I would have a balanced switched attenuator feeding a de-balancing op-amp and straight out. That way, +26dBu or greater signals get reduced to around 0dBU before they hit the op-amp.You could also have a switched load in front of the attenuator.

Cheers

Ian

 

[abbey road d enfer]

Just a couple of comments. Does it need to be that complex?  The fewer components in the signal path the better and the less they are stresses the better. So on the DUT to test gear side I would have a balanced switched attenuator feeding a de-balancing op-amp and straight out. That way, +26dBu or greater signals get reduced to around 0dBU before they hit the op-amp.

Agreed. The attenuator needs special attention if one wants to retain the high CMRR performance, though.

 

[ruffrecords]

Just a couple of comments. Does it need to be that complex?  The fewer components in the signal path the better and the less they are stresses the better. So on the DUT to test gear side I would have a balanced switched attenuator feeding a de-balancing op-amp and straight out. That way, +26dBu or greater signals get reduced to around 0dBU before they hit the op-amp.

Agreed. The attenuator needs special attention if one wants to retain the high CMRR performance, though.

Agreed. The question is do you want to retain it? If CMRR is something you want to measure then clearly it needs to be retained. If it isn't then, unless the testing is taking place in an incredibly hostile environment, it does not matter too much.

For myself, I am a transformer in and out man so it is less of an issue for me - in fact I don't actually need a balanced input.

Cheers

Ian

 

[Andy Peters]

Having done a USB audio device with phantom power, I'm not quite sure what their issue is.

Are you allowed to share this with us?

Here's the gist of it. I apologize for the handwaving. Remember that the USB spec has both a current limit (based on what your device tells the host it requires during enumeration) and an inrush current spec. The latter is given basically by saying you can't hang more than 10 uF on the USB 5V. One way around the inrush limit is to power up things in stages. I'm discussing USB 2.0 and its 500 mA-per-port bus-power limit; USB 3.0's limit is higher but that has to be accounted for in the design.

So you have your USB 5V, which can really be as low as 4.4V and as high as 5.25V. I take that in and use a little switcher to give me 3.3V for my micro and the other digital things.

TI makes a series of small current-limited power switch parts, good for an amp at up to 5.5V. The bus 5V goes to the switch input. The switch output feeds two regulators.

One is a TI boost converter (actually in their catalog as an LED backlight driver) which takes the switched 5V and bumps it to 24V and a doubler on its output gives me 48V at enough current to supply two mics. The fun thing is that the app note for this converter has a 5V-to-48V circuit shown!

The other load on the switched 5V is a neat little regulated 5V-out charge pump. This guy takes in anything from 2.7V to 5.5V and gives you 5V +/- 4% at up to 150 mA, and it's decently quiet. This charge pump can be driven by an external clock, too, so I take the I2S BCLK signal and divide it down to a frequency the charge pump can use. Now my analog power regulator's switching is synchronous with the converter clocks.

So you plug in the device. The power switch and 48V and 5V regulators are disabled. The micro boots and responds to enumeration requests. At some point, enumeration finishes and the device is considered configured. The power switch is turned on. After a short delay (mainly to allow the regulator input caps to charge) the phantom power switcher is enabled. After another delay, finally the 5V charge pump is turned on (the I2S clocks start up before the charge pump is enabled, of course). I timed everything by watching the inputs and outputs of the switch and regulators on my 'scope. By staging the regulator enables you can limit the inrush current enough to basically keep the USB 5V from collapsing and resetting your device.

Oh, one more thing: I use high-side switch MOSFETs to control the phantom power to each mic input. Those are of course off until enabled by the user. I discovered that evil things would happen if you had a load on the phantom power when you enabled it.

-a

 

[abbey road d enfer]

If CMRR is something you want to measure then clearly it needs to be retained.

I assumed dfutura was concerned about it, seeing how he's chosen the 1200 over the 1240; maybe I'm wrong...

 

[dfuruta]

Does this look somewhat more sensible?

Please forgive my ignorance;  I'm studying hard, but I'm still a neophyte amongst experts, here.

My QA400 is in the mail–once I get it in a few days, I'll be glad to answer any questions that I can.

EDIT:  R10 and R12 should be 620, sorry.  I had a long day at work...

 

[abbey road d enfer]

Does this look somewhat more sensible?

It should work.
Now you could dispense with the second 1200 and drive directly the 797 from the BNC.

 

[Kingston]

ruffrecords, it's QuantAsylum QA400 if you could correct the thread title. Increasing chances of successful searches.

 

[ruffrecords]

Does this look somewhat more sensible?
.

I think that is a definite improvement. I also think the stimulus chain could also be simplified. The source is unbalanced so do we really need the first THAT1200?. The 20dB gain is good for getting high drive levels. I would also like to see a balanced attenuator at the output so you can dial right down to -60dBu or even -70 or -80 for testing mic pres.

Cheers

Ian

 

[dfuruta]

What about something like this?  I'm not sure the output pad is the best way to go at it.

My QA400 is at the post office, apparently, so I'll pick it up tomorrow.

 

[abbey road d enfer]

As it is,the output impedance switches between 150r and 650r; not a big deal but I think the resistors should be 226r, for 600r.
Now, you must ask yourself, why do I need an output attenuator?
If you follow Ian's advice, you need 60 or more dB attenuation for testing high-gain preamps; you can't do that by simply turning the generator 60 or 70dB down because the electronics noise will become dominant, so you need an output attenuator. And also you need to mute the signal passively, at the attenuator's input in order to properly measure the preamp's EIN.
I reckon a single 20dB output attenuator is not very useful.

 

[ruffrecords]

I think we are getting there. This is much simpler and much more useful. I do think the range of the output attenuator needs to be extended - maybe 20dB steps from 0dB to 60dB. I am not sure about measuring EIN with the test set driving side providing the termination. I have some specially made XLRs with built in resistors or a short for measuring EIN. I would also like to see a wider range on the input attenuator from the DUT. 10dB steps from 0dB to 30dB would be good. I know 30dB sounds like a lot but many of my tube circuits can produce in excess of +30dBu.

Keep up the good work. I am really looking forward to when you get your QA400 and let us know what you think of it. They charge an extra $49 to send it to the UK so I am keen to hear what you think of it.

Cheers

Ian

 

[gemini86]

I think a few of us are waiting for a review, I know this little box is extremely tempting me.

 

[dfuruta]

I'm not working tomorrow, so I'll try to revise this again.  Thanks to you folks, it seems like this will be doable!

I'm not too experienced with this sort of gear--I'm comfortable with a scope, of course, but I haven't had the opportunity to get my hands on any real audio test equipment (thus my excitement about this relatively affordable unit).  So, is there anything specific that anyone would like me to try out with this new unit?  Until I can get this front-end sorted out and built (hopefully within a few weeks), it'll have to just be with the bnc, 3dBv-max inputs.  I should have some time tomorrow and later in the week to play with it.

 

[abbey road d enfer]

I am not sure about measuring EIN with the test set driving side providing the termination.

That's why I insisted on "you need to mute the signal passively". The 1646 line driver must be disconnected and replaced by a 50r for proper noise measurements.

  I have some specially made XLRs with built in resistors or a short for measuring EIN.

That's what I used to use before having an AP at work. When you've grown accustomed to making noise measurements in continuity with other tests without disconnecting cables, there's no turning back.

 

[ruffrecords]

I am not sure about measuring EIN with the test set driving side providing the termination.

That's why I insisted on "you need to mute the signal passively". The 1646 line driver must be disconnected and replaced by a 50r for proper noise measurements.

  I have some specially made XLRs with built in resistors or a short for measuring EIN.

That's what I used to use before having an AP at work. When you've grown accustomed to making noise measurements in continuity with other tests without disconnecting cables, there's no turning back.

Yes, I can see the advantages if you are making a lot of measurements. I have never been 100% comfortable about measuring EIN with a termination resistor since what you measure is the noise of that resistor plus that of the amplifier. I know in some semiconductor designs the noise of the amplifier itself depends on the source resistance. This is not the case with tubes so I generally measure with a short circuit.

Cheers

Ian

 

[abbey road d enfer]

I have never been 100% comfortable about measuring EIN with a termination resistor since what you measure is the noise of that resistor plus that of the amplifier. I know in some semiconductor designs the noise of the amplifier itself depends on the source resistance.

So much so that measuring noise of a solid-state piece without quoting the source impedance is meaningless. Sometimes noise is measured with typical 150-200 r load and a short-circuit; I don't believe it's for particular enlightenment purpose, more for the sake of publishing better figures, though.

This is not the case with tubes so I generally measure with a short circuit.

Indeed thermionic noise is independant of load, except for very faulty tubes or dust contamination. When a xfmr is used at the input, the magnetic pick-up noise is impedance-dependant, though. Visualization of noise (or harmonic analysis) allows discrimination of noise and hum components.
FET's are supposed to be similar, although noise-current exists, but is generally insignificant for loads < 1Megohm. However some bias mechanisms introduce parasitic noise that is impedance-dependant.

 

[ricardo]

I have never been 100% comfortable about measuring EIN with a termination resistor since what you measure is the noise of that resistor plus that of the amplifier. I know in some semiconductor designs the noise of the amplifier itself depends on the source resistance. This is not the case with tubes so I generally measure with a short circuit.

These are 2 different measurements and are BOTH important.

Short circuit noise measures the Input Voltage Noise Eni of your LN amp.  It is probably the most important figure when using a modern capacitor mike.

Measuring with a Source Resistor Rs gives you the Noise Factor, NF.  This is how much the LN amp degrades the noise of the resistor.  It includes not only the Voltage Noise of the LN amp but also its current noise Ini * Rs.  If the amplifier doubles the noise power of the resistor, you have NF = 3dB.  This would be a good mike amp and in fact appears as a requirement in BBC & IBA specs for broadcast desks.  This is the important spec. if you are using ribbon or other low output dynamic.

eg.  A 150R resistor (or mike) has 223.4nV noise over 20kHz or -133dBV.  A perfect 60dB amp would show -133+60 = -73dBV on its output.  If it showed -71dBV, the noise is 2dB above "perfection" so the amp has a NF of 2dB.  BTW, this would be an excellent LN amp and VERY hard to better in practice.  The best I've done is NF = 1.7dB with a large Lundahl and 5534 in my Calrec days.

A valve or FET has very small current noise but it exists.  A LN amp gives best performance when Eni & Ini * Rs contribute about equally to the final NF.  A transformer changes Rs so the LN amp sees approximately this ideal resistance, Ropt.

The sensible way to quote a LN spec is to state both SC noise as Ein and also NF for 150R or 200R source.  Quoting NF is better than quoting "Ein for 200R" cos you don't have to remember how to calculate resistor noise.  The NF tells you immediately how close to 'perfection' it is.

 

[JohnRoberts]

I've been a advocate of NF specs for years... but the advertising types prefer to use ein noise voltages that are easier to game and sound so much more impressive to the unwashed.

I like to think of how close to perfect I got....

JR