From: bill.sloman@ieee.org
On 16/11/2025 7:01 pm, Liz Tuddenham wrote:
> Bill Sloman wrote:
>
>> On 16/11/2025 6:04 am, Liz Tuddenham wrote:
>>> Bill Sloman wrote:
>>>
>>>> On 15/11/2025 8:23 pm, Liz Tuddenham wrote:
>>>>> Bill Sloman wrote:
>>>>>
>>>>>> On 15/11/2025 4:46 am, Liz Tuddenham wrote:
>>>>>
>>>>>
>>>>> [...]
>>>>>>>
>>>>>>> The Johnson noise of a 50-ohm resistor at 20 C with 5 Kc/s bandwidth is
>>>>>>> -130 dBm. If we take the rail-to-rail output voltage of an op-amp to
be
>>>>>>> about +20 dBm, that gives 150 dB difference. Claims that a device is
>>>>>>> capable of achieving this sort of figure in any practical situation is
>>>>>>> stretching the limits of credibility.
>>>>>>>
>>>>>>> If the signal is lower than +20 dBm, you either have to start cooling
>>>>>>> the device or reducing the measurement bandwidth, otherwise 150 dB S/N
>>>>>>> ratio is impossible to achieve.
>>>>>>
>>>>>> Johnson noise is broadband. The harmonic of interest in evaluating a low
>>>>>> distortion sine wave oscillator are rather narrow band.
>>>>>
>>>>> When you get down to signals as low as -150 dBm there will be Johnson
>>>>> noise and intermodulation products contributed by the components within
>>>>> the oscillator. Selecting individual harmonics with a narrow-band
>>>>> filter will give an unrealistic number that doesn't represent the total
>>>>> unwanted output.
>>>>>
>>>>>> You are comparing apples and pears as your "5 Kc.s bandwidth" makes
>>>>>> clear. if you have been cribbing from a more modern source it would have
>>>>>> been a 5kHz bandwidth.
>>>>>
>>>>> You have no idea what I did, so stop posting offensive nonsense.
>>> [...]
>>>>> I have written a calculator based on a spreadsheet which gives the RMS
>>>>> noise as a voltage and as dBm from user inputs of resistance, bandwidth
>>>>> and temperature. I chose 5 Kc/s as a tolerably low figure for audio
>>>>> work; it would be a suitable measurement bandwidth for a 1 Kc/s signal
>>>>> if you wanted to include the third harmonic. I chose 50 ohms as it is
>>>>> the lowest output impedance in common use - for audio work 600 ohms
>>>>> might have been more appropriate but the figures would have been even
>>>>> worse.
>>>>>
>>>> I'm pretty confident that you don't have much of an idea what you did
>>>> either.
>>>
>>> You will notice that I have moved your comment by one paragraph so that
>>> it now comes after my explanation of exactly what I did. Your post
>>> appears to have been done purely to be offensive and obfscate the facts.
>>
>> Your explanation is an exact explanation of what you thought you did.
>> The fact that you fixed on a specific bandwidth is clear evidence that
>> you didn't actually understand the problem you were tackling.
>
> You appear to have been interested in selected harmonic distortion
> whereas I was interested in total distortion (which includes noise).
> The former is more academic, the latter is more practical.
There is very little you can do about noise levels, apart from keeping
the impedances low. Harmonic generation is more susceptible to careful
design, and working out how much of each of the harmonics is actually
present does help the design process.
If you have a controllable wide band frequency source, ideally producing
separate in-phase and quadrature components, you can scan the frequency
across the signal of interest and use two wideband four quadrant
multipliers to multiply the in-phase and qudadrature components from the
frequency generator with the signal and plot the products as the
amplitude of each frequency component in the signal being looked at.
This is what a classic low frequency spectrum analyser does.
Specifying a 5kHz bandwidth for the noise you will condescend to report
is idiosyncratic.
--
Bill Sloman, Sydney
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
