EEVblog #1104 - Omicron Labs Bode 100 Teardown

[sorin]

We have tested a different option and found that two huge chinese MnZn cores (search for "Cores Green 75mm x 39mm x 13mm") stacked on top of each other and afterwards wound in a similar fashion (not in the number of turns, but in terms of covering the core with windings, style of twisted wire and connection) to the Bode 100 transformer gives the attatched result. Seems like a lot of wideband transformer for about 10€ for the cores. The much increased volume of core material might also help with saturation?

Please can you post a photo of your final product

[Jay_Diddy_B]

We have tested a different option and found that two huge chinese MnZn cores (search for "Cores Green 75mm x 39mm x 13mm") stacked on top of each other and afterwards wound in a similar fashion (not in the number of turns, but in terms of covering the core with windings, style of twisted wire and connection) to the Bode 100 transformer gives the attatched result. Seems like a lot of wideband transformer for about 10€ for the cores. The much increased volume of core material might also help with saturation?

Hi,

Can you post a schematic showing how this transformer was measured?

If I calculate the inductance:

for one core:

Area = Thickness x (D - d)/2

Where
D= outside diameter = 75mm
 
d= inside diameter = 39mm

Thickness = 13mm

Area = 13 x (75-39) /2 = 234E-6 m²

Mean magnetic Path length

Pi x (D + d)/2 = PI x (75 +39) /2 = 180E-3

Assume that this core is a power grade of ferrite with a relative permeability of 2000.

AL= mu₀ x mu_r x Area/length

AL = 4E-7 x Pi x 2000 x 234E-6 / 180E-3

= 3.28uH per turn²

Assume 50 turns

Lmag = 50² x 3.3uH

= 8mH

and two cores stacked would be 16mH

50 Ohm test circuit



The LF -3dB for this configuration is calculated

R1 // R2 = R

LF(-3db) = 1/(2 x Pi x L/R)

= 250 Hz



It just seems too good …

Regards,

Jay_Diddy_B

[jmw]

I bought a VAC T60006-L2040-W424 and used a 23AWG pair from a plenum CAT6 cable. For the life of me, I could not get more than 29 turns on the core. Maybe the original uses 24AWG or something smaller?

I don't have a two-port VNA but I used the Bode function on my scope (with a 50 ohm feed-thru term on the sec. side). Flat and phase response below 5 deg until just below 1 MHz, good enough for my needs.

[Jay_Diddy_B]

Hi,

I suspect that you are not measuring the transformer correctly.

To test the transformer in a 50 environment you need to use the top circuit here:



I suspect that you are using the bottom one. In this configuration you should read -6dB in the flat portion.

If you use the bottom one you don't see the impact of the magnetizing inductance.


Regards,
Jay_Diddy_B

[jmw]

Hi Jay_Diddy_B, I repeated the test to your procedure:



So it seems by 3 dB points it is good from 40 Hz - 2 MHz, and by 5° phase limit from 500 Hz - 200 kHz. Not spectacular by others' work, but still adequate for evaluating the feedback look on many switchers. I'm still puzzled on the turns count and why I'm topping out at 29 if people got 40 using 23 AWG... but my expectation is that more turns would help at the low frequency and not so much the high end.

[Jay_Diddy_B]

Hi Jay_Diddy_B, I repeated the test to your procedure:

(Attachment Link)

So it seems by 3 dB points it is good from 40 Hz - 2 MHz, and by 5° phase limit from 500 Hz - 200 kHz. Not spectacular by others' work, but still adequate for evaluating the feedback look on many switchers. I'm still puzzled on the turns count and why I'm topping out at 29 if people got 40 using 23 AWG... but my expectation is that more turns would help at the low frequency and not so much the high end.

Hi,

40Hz means that the magnetizing inductance is about 100mH.

I am not sure that other people measured them correctly.

It is very useable.

Regards,
Jay_Diddy_B

[Roland_W]

Hi, All,

I read the all about the Injection Transformer.

I need one, and as Wolfgang I build one using the same core VAC T60006-L2030-W514 :
https://pl.mouser.com/ProductDetail/Vacuumschmelze/T60006-L2030-W514?qs=ePbE9GiMmvVKrWoTvqBaIQ==

I use a RJ45 cable twisted pair 10 m long.

I wind first layer and I measure using the Diligent Waveforms Analog Discovery 2 to measure frequency response.

I make first measurement when I done the first layer of the winding – orange one, second layer winding – blue one, and the third – yellow one.
More turns move lower the working frequency.

[MilkmanCDN]

I bought a VAC T60006-L2040-W424 and used a 23AWG pair from a plenum CAT6 cable. For the life of me, I could not get more than 29 turns on the core. Maybe the original uses 24AWG or something smaller?

Hey there.    Are you sure you bought the L2040?  There are a few posts on this thread stating that the L2030 is the transformer used in the BWIT-100.    I first purchased the L2030 and had the same problem as you (I could only get about 29-30 turns around the thing).   It was then that I realized that the L2040 is a larger core and in fact the one used in the BWIT-100, which will accommodate the required 40 turns.

See reply #216 in the thread for details.   I've since purchased a used Bode100 and verified performance of the transformer (which is identical to the original).

[jmw]

I bought it from Mouser, and part # from the order history checks out, but there is the small chance they did a bad pick. I'll make a note to check the dimensions on it tomorrow at the lab, and the datasheet for T60006-L2040-W424 says it is 40 mm OD / 25 mm ID.

Edit: dimensions match the L2040 datasheet...

[CChin254]

I did an IC Identification of this device. The back SOT-23 packages are probably all ON Semiconductor MDC3105LT1 Integrated Relay, Inductive Load Drivers.

[Wolfgang]

Hi, All,

I read the all about the Injection Transformer.

I need one, and as Wolfgang I build one using the same core VAC T60006-L2030-W514 :
https://pl.mouser.com/ProductDetail/Vacuumschmelze/T60006-L2030-W514?qs=ePbE9GiMmvVKrWoTvqBaIQ==

I use a RJ45 cable twisted pair 10 m long.

I wind first layer and I measure using the Diligent Waveforms Analog Discovery 2 to measure frequency response.

I make first measurement when I done the first layer of the winding – orange one, second layer winding – blue one, and the third – yellow one.
More turns move lower the working frequency.

Hi, looks interesting but for my taste the windings on top of each other create some wiggles at the top end due to interwinding capacitive coupling.
Anyway, its flat enough for many purposes.

[_Wim_]

I did an IC Identification of this device. The back SOT-23 packages are probably all ON Semiconductor MDC3105LT1 Integrated Relay, Inductive Load Drivers.

Nice job and interesting to see. Thanks!

[TimNJ]

For anyone who has DIY'd one of these injection transformers, what have you used to actually plot the gain/phase of the power supply under test? Some sort of low frequency network analyzer sounds good, but if you don't have one...what are some reasonable options? Are you better off doing it manually with a sine wave generator and oscilloscope?

From watching one of Keysight's demo videos , seems like he had to a do a fair bit of fiddling to get accurate measurements since an SMPS can be quite noisy, and you have to separate the uncorrelated noise from the measurement. Seems like the dedicated software probably has some decent DSP to extract only the real signal.

Measuring manually lets you make the decision on what's noise and what is not. But, also will probably take a while if you're going to cover 4 or 5 decades.

Any ideas?

[Wolfgang]

Hi,

I measured my stuff on a Bode100 and on a Keysight E5061C-3L5.
Bode100 is pricewise OK, E601C-3L5 is extremely expensive.

Whats different when measuring on a scope:

- dynamic range is a lot less due to scope resolution (cheap ones have 8 Bits only)
- no small bandwidth as on a VNA that filters out spurs
- FFTs a bit kludgy (except when scope is expensive)

What can you do:
- Get (if you can) your scope into some high-res mode
- Hook up your injection transformer to a signal generator
- Use a PC to step the signal generator and get scope traces per step into you PC.
- Make steps small enough so you dont miss a peak. Step logarithmically.
- do a manual FFT using Python in the PC. No strong harmonics of the injected signals should be there, otherwise its too strong.
- Map gain and phase for each step
- If you find suspicous peaks with gain close to one, those are the weak spots to work on.

Alternative:
- Connect a load pulser (fast enough for your PSU) and inject a step load current pulse
- Observe pulse response for ringing. Reverse transform gives transfer function.

Another idea:
- Use a flat noise source (could be an ARB)
- Amplify it
- Drive a current injector with it (watch for linearity !)
- Measure output voltage without and with noise on
- FFT and average the result. Voilá, here are your peaks.

[Jay_Diddy_B]

Hi,
 I use an old HP 3577A.

This is a 5 Hz to 200 MHz  VNA.

It has 10Hz filters it is very good for measuring SMPS control loops.

Regards,
Jay_Diddy_B

[TimNJ]

Thank you all.

For most of the projects I've worked on, checking stability with a dynamic electric load has worked well. With current mode control, and some other newer control methods (for example, TI's "hybrid hysteretic control"), reaching stability seems pretty trivial these days. I am still curious roughly the gain margin of these converters, from an educational perspective.

The semi-automated PC method seems reasonable, but do not have a PC compatible function generator. (Could purchase the AWG package for my scope and use it remote.)

I wonder how Analog Discovery 2 would fare in this application. It has a 14-bit DAC, so I suppose that's a plus for measuring low amplitude signals.

https://reference.digilentinc.com/reference/instrumentation/guides/waveforms-network-analyzer

I downloaded the Waveforms application, and checked out the network analyzer function. It looks okay, but am not sure what the measurement RBW is for the network analyzer. Or, in general, I don't know how it would handle being blasted by switching noise/ripple.

[jmw]

The Keysight 1000-X has a frequency response analysis program, and I think it's an option on other scopes. The dynamic range is pretty terrible, if one channel is more than 30 dB than the other, then the plot gets really noisy. It's still good for looking at the crossover region.

This is Bode chart I made for a flyback converter, using my homebrew injection transformer. Prior to the test, I picked the compensation components to get a crossover at 12 kHz with a 60° phase margin based on an analytical model of the feedback loop.

[TimNJ]

The Keysight 1000-X has a frequency response analysis program, and I think it's an option on other scopes. The dynamic range is pretty terrible, if one channel is more than 30 dB than the other, then the plot gets really noisy. It's still good for looking at the crossover region.

This is Bode chart I made for a flyback converter, using my homebrew injection transformer. Prior to the test, I picked the compensation components to get a crossover at 12 kHz with a 60° phase margin based on an analytical model of the feedback loop.

(Attachment Link)

Interesting. Thanks. I was not particularly impressed by the amount of human effort required to get a reasonable measurement, in the Keysight video. I do not yet have a DIY injection transformer...I will try JDB's EPCOS common-mode choke transformer suggestion + an oscilloscope in hi-res or averaging mode. I wonder how long it would take to "manually" record 5 decades of information with 10 samples per decade. Probably about 45 minutes if you're dedicated? Not great, but maybe okay if you only are doing it occaisionally.

[Alextsu]

Hi, Tim,
You're looking in the right direction.
Yes, AD 2 will do the job. The amount of noise filtering depends on the integration settings.
I had used a black legacy AD to measure frequency responses of several AC-DC PSU.
I had also compared it's performance against other test instruments.

The only issue You may run into with AD is a poor USB PC connection noise immunity.

I wonder how Analog Discovery 2 would fare in this application. It has a 14-bit DAC, so I suppose that's a plus for measuring low amplitude signals.

I downloaded the Waveforms application, and checked out the network analyzer function. It looks okay, but am not sure what the measurement RBW is for the network analyzer. Or, in general, I don't know how it would handle being blasted by switching noise/ripple.

[TimNJ]

Hi, Tim,
You're looking in the right direction.
Yes, AD 2 will do the job. The amount of noise filtering depends on the integration settings.
I had used a black legacy AD to measure frequency responses of several AC-DC PSU.
I had also compared it's performance against other test instruments.

The only issue You may run into with AD is a poor USB PC connection noise immunity.

I wonder how Analog Discovery 2 would fare in this application. It has a 14-bit DAC, so I suppose that's a plus for measuring low amplitude signals.

I downloaded the Waveforms application, and checked out the network analyzer function. It looks okay, but am not sure what the measurement RBW is for the network analyzer. Or, in general, I don't know how it would handle being blasted by switching noise/ripple.

Thanks very much. Do you think a USB isolator + separate USB battery pack would be a useful approach? As I am not familiar with the Waveforms software, where are the integration settings located? I could not find much adjustment in the GUI menus.

[Alextsu]

Thanks very much. Do you think a USB isolator + separate USB battery pack would be a useful approach? As I am not familiar with the Waveforms software, where are the integration settings located? I could not find much adjustment in the GUI menus.

Hi, Tim!
To modify the settings, go to Wavegen window presets menu. Located at right top in the network analyzer workspace.
Options are 'min periods' and 'averaging'.
Also, You may need to choose the 'Table' mode in the Wavegen settings to get a variable voltage of the test signal across the frequency span of interest. At low frequencies, the test signal amplitudes must be higher to compensate for the injection transformer signal attenuation.
I can't tell if the USB isolator would work with AD.
I had tried a cheap Chinese usb isolator once with other usb instruments just to find out that it didn't work with them.
If You are going to test a quite DC-DC convertes like low power buck or boost , the EMC should not be a problem.
With more noisy AC mains-driven converters, a poor PC USB connection noise immunity causes random PC communication failures. To solve this problem, You can try to use a good common-mode choke based USB filter.
 Also, AD2 has a slightly better USB noise immunity than AD1.

[jmw]

I got some time in the shop to drill enclosures for mine and a few of my other odd test tools. No nude virgins, but I'm calling this look "midcentury space-race brutalism"

[Wolfgang]

The proper term is early Hammond Era "Art Brut". My preferred building style. 
How did you get the letters on the boxes? I used an oversupply of overhead foils and a laser printer ...

[jmw]

There's a huge 120 W CO2 laser here, so I used a laser marking spray (https://enduramark.com/ - there are several others sold under CerMark, LaserBond). The results have been pretty nice and durable.

[RamboChen]

Hi, Thanks for your good job!
is there have any more detil information about the receiver channel(or some Clear images)? i‘m try to analysis it's mixer principle.