My choice is to isolate the USB or isolated the SPI and ever other data line I use.
It can add up quite quickly. I have range switching to do as well. if I do the range switching with relays it is already isolated. if I do all else with SPI expanders then it's just three control lines. If USB then the entire micro is isolated and no future ground loop problems.
I may want to use the internal ADC of the micro for detecting output level.
The SVO level is already available from the AGC ADC but is this a useful display.
The good new is the 20mHz osc for the micro seems to not show up in any way on the spectrum which mean the micro can be left running.
The frequency counter that I was going to include take it input from the quadrature pulse signal of the AGC which is synchronized to the zero crossing.
I think I will have to use a reciprocal counter if the fine tune is to be displayed. Other wise it would be fixed to a 1Hz resolution. This is a feature that would require calibration.
My biggest concern for having a windows interface is the issue with drivers. Microchip uses a generic driver for USB On The Go (OTG). The driver works with XP but will it work with Win7 & 8 and other future builds? I don't like being dependent on a third party for drivers. The app is written in .NET4 and that won't go away. Does any one know if an app compiled on a XP box will run on win7?
It can add up quite quickly. I have range switching to do as well. if I do the range switching with relays it is already isolated. if I do all else with SPI expanders then it's just three control lines. If USB then the entire micro is isolated and no future ground loop problems.
I may want to use the internal ADC of the micro for detecting output level.
The SVO level is already available from the AGC ADC but is this a useful display.
The good new is the 20mHz osc for the micro seems to not show up in any way on the spectrum which mean the micro can be left running.
The frequency counter that I was going to include take it input from the quadrature pulse signal of the AGC which is synchronized to the zero crossing.
I think I will have to use a reciprocal counter if the fine tune is to be displayed. Other wise it would be fixed to a 1Hz resolution. This is a feature that would require calibration.
My biggest concern for having a windows interface is the issue with drivers. Microchip uses a generic driver for USB On The Go (OTG). The driver works with XP but will it work with Win7 & 8 and other future builds? I don't like being dependent on a third party for drivers. The app is written in .NET4 and that won't go away. Does any one know if an app compiled on a XP box will run on win7?
Last edited:
You can isolate with something like this for the USB interface: ADUM3160 datasheet and product info | Full/Low Speed USB Digital Isolator | USB Isolators | Analog Devices You still need power on the isolated side. Its not fast enough for audio but fine for control. However you will have a micro sharing ground with the 140 dB + dynamic range signals from the oscillator. I would go over the top with a can and feedthroughs to couple the controls to the oscillator but that may be overkill.
Using a USB ADC or an internal sound card brings back the grounding issues anyway. My best results are from using an SPDIF connected external ADC for testing. Transformer isolated SPDIF does work well for this. Its a lot of hassle that's only necessary when going below -100 dB.
Using a USB ADC or an internal sound card brings back the grounding issues anyway. My best results are from using an SPDIF connected external ADC for testing. Transformer isolated SPDIF does work well for this. Its a lot of hassle that's only necessary when going below -100 dB.
plastic fiber optic has gotten much faster at low cost - 100 Mb optical ethernet for cheap is one example
chip isolators compared: http://www.analog.com/static/import...s/5406830898840Agilent_NVE_FMR_IsolatorsF.pdf
chip isolators compared: http://www.analog.com/static/import...s/5406830898840Agilent_NVE_FMR_IsolatorsF.pdf
Last edited:
I have not investigated this in sufficient detail for a competent answer (particularly WRT wirewound pots), but for ordinary mid-value (around 10 kOhm) cermet trimmers 1 mA is cleary very significant at the -120 dB distortion level (if directly in the signal path).
Samuel
Dick
I know you are just kidding, but there are two other important differences.
I used the resonator to power an AC line correction amplifier. If the AC line has a few cycles of drop out the resonator maintains an output and stays in sync.
The second feature is that matches phase with the locking signal, very important in a power supply adjuster!
Scott,
I do want to thank you for your confusion. Without it I would have had to dig out my very old stuff to show the measurements, instead you did it. So thanks.
Now as you are aware the late great Jim W. was building his super oscillator to test analog to digital converters. The circuit you have can be tweaked to be better than 24 bits. Is there really any use for this in actual testing of analog to digital converters?
And of course I should welcome you to the quest for 160 db!
ES
Your computer is dumping a lot of noise back into the AC ground. The scope ground creates a loop to push it back in. Laptop on batteries is the best fix, except for the other radiated noises.
The USB for the control has nothing to do with the audio side of things so the AD isolator would work well for that. The micro has no visible effect with the EMU0204 but it might with other sound cards. The power to the micro could be isolated using a DC to DC convertor but convertors operate at a relatively low frequency of 100kHz or so at best and this could be problematic. It's no problem to isolate the data lines and is cheap to do. The other option for the micro power isolation is to run it off it's own transformer winding and let it float.
I need a rail to rail op amp that can operate from a single supply. The negative supply
must be connected to ground and positive +15. It must have at least 16 bit settling, no latch up or sticking under these condition. It must be capable of being driven hard into saturation on the negative rail and recover gracefully from this. It must be internally compensated and require no FB capacitor. It must be wide band at a gain of -4 with no phase shift up to 200kHz. It must be capable of driving capacitive loads up to 100pF and must be low noise, < 5nV/root Hz.
Any suggestions.
must be connected to ground and positive +15. It must have at least 16 bit settling, no latch up or sticking under these condition. It must be capable of being driven hard into saturation on the negative rail and recover gracefully from this. It must be internally compensated and require no FB capacitor. It must be wide band at a gain of -4 with no phase shift up to 200kHz. It must be capable of driving capacitive loads up to 100pF and must be low noise, < 5nV/root Hz.
Any suggestions.
Yes I know the comp is driving noise into the AC ground.
I don't usually have the scope connected while doing low disto measurements.
Even floating it drives the noise up a bit.
At these level things are so sensitive I can see the cat farting.
The LT1468 is what I have in there now but it can't be operated rail to rail and it doesn't like single supply operation with it's non inverting input tied to ground.
It also requires a FB cap to balance input capacitance otherwise it will oscillate at 7 to 10MHz.
This op amp is the signal conditioner for the input of the ADC in the AGC.
Although the ADC I'm using will withstand a +/- 25Vpp input without damage, driving the ADC into the negative increased the SVO's distortion 3 orders of magnitude. Yuk.
Therefore the op amp must be clamped with a FB diode to confine the swing to positive only. The combined capacitance of the schotkey clamping diode and FB capacitor introduces enough phase shift to cause a level difference from the input signal to output signal of the conditioning amplifier. This causes a 100mVrms or more level error of the SVO output over the frequency span of the highest range and than 10mVrms over the frequency span of the lower ranges.
I new about this before designing the input conditioner but I thought I could operate the ADC full swing and with minimal FB capacitance around the 1468 it would work.
The input conditioner needs to be redesigned. Without a clamping diode and no FB capacitor this problem is solved. I new this before as well. A problem with the rail to rail op amp I tested was sticking from being driven to hard saturation into the negative supply which was ground. Otherwise it worked without amplitude error. With the negative supply connected to ground the output is limited to ground. No clamping diode and no FB cap was needed.
I need an op amp that will perform well under these conditions and has sufficient settling time for a 16 bit conversion. If the amplifier doesn't settle fast enough then we're back to amplitude errors.
Last edited:
How much of the sine wave are you digitizing? Would a clamping amp like used for settling time measurements work?
Anything not rail to rail won't work.
What works okay is an opa1641. But this is an audio op amp and doesn't give information on setting time. It also sticks a bit coming out of saturation.
Thanks Dimitri.
That will do.
I'm out of stock. Have to order more.
There is some filtering on the unused input pins. I wonder if the filter is causing attenuation. This is an LTC1609. Take a look at the data sheet under additional input filtering section. If the higher frequencies are attenuated the AGC response is to raise the level.