The ringing you see on your scope could be from a scope probe ground lead that is too long. Are you using a probe with a spring ground?
Or even better, get an old used active probe and make a spring ground for it. You can get an old used active probe with low capacitance attenuators for maybe $100 - $150. Then you would be able to see what the waveforms really look like. Otherwise you might be adding caps to hide an imaginary problem caused by your measurement instrument.
OTOH, at some point you will need to filter the output of your digital audio device to get the RF out of the audio signal. However, you still need to know if you are fooling yourself with measurement problems.
Or even better, get an old used active probe and make a spring ground for it. You can get an old used active probe with low capacitance attenuators for maybe $100 - $150. Then you would be able to see what the waveforms really look like. Otherwise you might be adding caps to hide an imaginary problem caused by your measurement instrument.
OTOH, at some point you will need to filter the output of your digital audio device to get the RF out of the audio signal. However, you still need to know if you are fooling yourself with measurement problems.
Might be, but I'm using spring, to be honest osciloscope which I have using do not have an big diferencie between normal probe and one with spring, even connecting things by u-fll, not an big diference, instead I will do multitone to get idea if thing is realy improved. I will need to buy an much better osciloscope soon! What I have noticed is an good inprovement on mid tones, smooth like a buter! Let see what multitone say. Have no free time right now buty soon I will do measurement.
Its not just a better scope. Its active probes if you want to look at fast risetime square waves with low distortion. A good active probe costs more than many entry level 100MHz or 200MHz scopes. That's why I suggest hunting for a used one which includes attenuators. Look for an old P6201 which is good up to 900MHz. It needs analog +-15v power and a home made spring ground. The attenuators are needed because the active probe uses a low voltage FET. For lower capacitance and higher voltages, the attenuators are needed. For example: https://www.ebay.com/itm/3546728174...MIiZbjhoODhQMVvQ2tBh3bGglhEAQYAyABEgIvM_D_BwE
Yes, I have made diy AUX0025 FILTER, not used them in previous multitone measurement if I'm remember right, now I will use them
While worked on digital push pull regulator we have found that it is not possible to get strict 50:50 dutty cicle using something similar to our DDPD input complementary hi-lo, its imposible to get strict 50:50, but with help of one schmit trigered nand gate, two nor gates and one flip flop it is possible to make complementary gates for push pull with perfect 50:50 similarity on hi -lo transition which have apsolutly identic T on hi and lo, also it can work on fixed frequency with variable dutty cicle from 0 to 99%, perfect for push pull topology, for example I have wanted to make push pull isolated regulator with variable dutty cicle the same time variable voltage, definitelly posible! My bad I tried to use our ddpd complementary hi-lo circuity and found that it is not possible to get strict 50:50 for push pull : ) This one from attachment is good idea?
Attachments
Last edited:
And this is unsucesfull push pull regulator, hope in next revision it get much better mcu with much better pwm resolution and also our new complementary hi-lo. Idea is to use Voltage to Frequency converter -> AD7740 as a feedback, digital isolator, and one mcu which will count frequency and adjust pwm dutty cicle on a fixed 300khz freq and regulate output voltage, idea is crazy and seems nobody made something like that?? : )
Last edited:
This is results after 1nF at buffer position, not a good, I expected more but it didn't helped!.
Measurement setup:
Cosmos switch = 1.7V
Left channel input on Cosmos
CosmosTweak = stereo (slider to the left)
Aux0025 got a bit of modifications, I removed output attenuator so now total attenuation is ~2.81 times
The results:
ddpd 5V volume level = measured on ddpd out = 1.37Vrms
1.37 / 2.81 = 487mVrms goes to Cosmos input
multitone H1 result = -10.8db = ~OK! Close to what I measured by osciloscope.
ddpd 1.25Vvolume level = measured on ddpd out = 0.265Vrms
0.265 / 2.81 = 94mVrms
multitone H1 result = -25.1db = ~OK! Close to what I measured by osciloscope.
Result is on picture 1 and 2 ! The next thing, seccond attempt, I removed all the gates from H bridge so that H bridge do not have "16 averaging" gates per 1/2 half bridge so the only one gate per 1/2 HB, the result is similar, something is improved but something not, nothing particularly different, I will not post more result its very similar. My conclusion to all this is our small pcb have common ground, wrong idea. All sections must be fully isolated, but this is not possible in our litle ddpd. The problem here is most probably the ground noise on schmit triggered input gates and on evrything else, I give up on this little concept and soon I will switch to the version with gan mosfets, everything will be digitally isolated and each section will have its own isolated simple bonbon push pull predregulator, this is something specialy designed and very simple two transistor @100kHz self oscilating, 1.2mH prim&sec, TN10/6/4-3F3, called bonbon pushpull, works very well, and with their post regulator, ground plane also all independent, that way this concept should be much better!
One more important thing here is ground noise + ground loop. Ground loop is caused by: PC usb1 goes to Amanero, than from Amanero goes to DDPD, than from DDPD goes to AUX filer with dummy load, than from dummy load it goes to Cosmos ADC, from Cosmos ADC it goes to PC usb2, this is definitelly bad idea without digital isolators, so in general I need to prepare better for the version with gan mosfets if I want better results.
Measurement setup:
Cosmos switch = 1.7V
Left channel input on Cosmos
CosmosTweak = stereo (slider to the left)
Aux0025 got a bit of modifications, I removed output attenuator so now total attenuation is ~2.81 times
The results:
ddpd 5V volume level = measured on ddpd out = 1.37Vrms
1.37 / 2.81 = 487mVrms goes to Cosmos input
multitone H1 result = -10.8db = ~OK! Close to what I measured by osciloscope.
ddpd 1.25Vvolume level = measured on ddpd out = 0.265Vrms
0.265 / 2.81 = 94mVrms
multitone H1 result = -25.1db = ~OK! Close to what I measured by osciloscope.
Result is on picture 1 and 2 ! The next thing, seccond attempt, I removed all the gates from H bridge so that H bridge do not have "16 averaging" gates per 1/2 half bridge so the only one gate per 1/2 HB, the result is similar, something is improved but something not, nothing particularly different, I will not post more result its very similar. My conclusion to all this is our small pcb have common ground, wrong idea. All sections must be fully isolated, but this is not possible in our litle ddpd. The problem here is most probably the ground noise on schmit triggered input gates and on evrything else, I give up on this little concept and soon I will switch to the version with gan mosfets, everything will be digitally isolated and each section will have its own isolated simple bonbon push pull predregulator, this is something specialy designed and very simple two transistor @100kHz self oscilating, 1.2mH prim&sec, TN10/6/4-3F3, called bonbon pushpull, works very well, and with their post regulator, ground plane also all independent, that way this concept should be much better!
One more important thing here is ground noise + ground loop. Ground loop is caused by: PC usb1 goes to Amanero, than from Amanero goes to DDPD, than from DDPD goes to AUX filer with dummy load, than from dummy load it goes to Cosmos ADC, from Cosmos ADC it goes to PC usb2, this is definitelly bad idea without digital isolators, so in general I need to prepare better for the version with gan mosfets if I want better results.
Attachments
Last edited:
Sometimes Schmitt triggering is worse than not having it. Also, how you manage bypass currents have a significant effect in a dac (layout can be critical).
This currents in most case cause our little pcb with a common gnd to produce our bad Multitone result, I'm believing! Lmg5200 half bridge in full brdge configuration, with their internal gate driver and protections will be enought for next project with fully independent and isolated things, it must work well. Bonbon iso push pull will supply the most of things including isolated side of the lmg5200.
Last edited:
This is volume control for next ddpd. Very well power supply, with negligible noise, and cheap! : )
https://www.aliexpress.com/item/1005005747865961.html
https://www.aliexpress.com/item/1005005747865961.html
lately i've started practicing running gnd lines to star gnd (master gnd), i ignored it on pcbs until recently, also I'm practicing now cannaled gnd planes, current must go trought capacitor first and not trought their own way, I learned a little regarding gnd : ) For example like this one with very good result https://www.diyaudio.com/community/threads/super-regulator.247281/post-7600121
Last edited: